smart goals literature review

Being smart about writing SMART objectives

Affiliations.

• 1 University of North Dakota, School of Medicine & Health Sciences, Center for Rural Health Evaluation, 250 Centennial Dr. Stop 8138, Grand Forks, ND 58202-8138, United States. Electronic address: [email protected].
• 2 University of North Dakota, School of Medicine & Health Sciences, Center for Rural Health Evaluation, 250 Centennial Dr. Stop 8138, Grand Forks, ND 58202-8138, United States. Electronic address: [email protected].
• PMID: 28056403
• DOI: 10.1016/j.evalprogplan.2016.12.009

This article challenges the conventional wisdom in mainstream evaluation regarding the process for developing specific, measurable, attainable, relevant, and time-bound (SMART) objectives. The article notes several advantages of mainstreaming the SMART method including program capacity building and being able to independently monitor progress toward process and outcome objectives. It is argued the one size fits all approach for writing SMART objectives is misleading. The context in which the evaluation is conducted is a key deciding factor in how and when the SMART criteria should be applied. Without an appreciation of the evaluation context, mainstream users may be developing objectives that are far from smart. A case example is presented demonstrating a situation where a stepwise, rather than simultaneous application of the SMART criteria was necessary. Learning from this case, recommendations are forwarded for adjusting how SMART criteria should be presented in mainstream evaluation manuals/guides.

Keywords: Evaluation guidance; Mainstreaming; Objective development; SMART objectives.

Copyright © 2016 Elsevier Ltd. All rights reserved.

Publication types

• Research Support, Non-U.S. Gov't
• Health Knowledge, Attitudes, Practice
• Health Promotion / organization & administration
• Organizational Case Studies
• Out-of-Hospital Cardiac Arrest / diagnosis
• Out-of-Hospital Cardiac Arrest / therapy
• Program Development / methods*
• Program Development / standards*
• Program Evaluation / methods*
• Program Evaluation / standards*

Event Abstract

Are s.m.a.r.t goals really smart the psychological effects of goal-setting in a learning task.

• 1 Department of Psychological Science, School of Health and Human Sciences, Southern Cross University, Australia

Aim: Specific, Measurable, Achievable, Realistic, Time-bound (SMART) goals are commonly used in educational settings as a strategy to optimise learning. However, research and theory suggest that such goals may not benefit, and could even be detrimental to, learning. Based primarily on goal-setting theory, this study investigated the effects of different goal types on a wordlist learning task. Method: 56 university students (45% male), average age 28 years (SD = 10.9) were randomly assigned to different goal conditions: specific (e.g. remember 10 words), ‘do your best’, process (e.g. memorise and repeat the words) and ‘open’ goals (e.g. see how many words you can remember). The total number of words recalled across four wordlist learning trials (Wechsler Memory Scale-3rd Ed) was computed as the index of immediate memory performance and one week later. The Intrinsic Motivation Inventory (IMI) was used to assess perceived competence, effort/importance and pressure/tension. The Positive and Negative Affect Schedule (PANAS) was used to assess participants’ state affect. Results: Repeated measures analysis of covariance (ANCOVA), with age as covariate, indicated no main effects of goal type on performance. Goal type groups did not differ on PANAS, perceived performance and confidence ratings but there was a significant effect of goal type on perceived challenge. Participants with specific goals found their goals less challenging than participants with do-your-best goals in the second trial (p = .040). A significant effect of goal type was also noted for perceived pressure/tension subscale scores. Participants with specific goals experienced less pressure/tension than participants with process goals (p = .021). Preliminary findings also suggested a significant difference among goal types in their recall accuracy a week later. Participants with open goals (M = 4.85) recalled significantly more words than participants with specific goals (M = 3.10, p = .020). Conclusion: Findings suggest that setting specific goals elicited less challenge and pressure. Although such conditions may appear optimal in learning, goal-setting theory recommended a goal to be both specific and challenging for one to achieve the best performance. Thus, participants with specific goals may have performed poorly in their subsequent recall due to participants viewing specific goals as not adequately challenging. Additionally, findings suggest that there may be potential benefit in setting open goals in learning situations.

Keywords: Goal-setting theory, Educational learning, SMART Goals, Open goals, memory recall

Conference: 15th Annual Psychology Honours Research Conference , Coffs Harbour, Australia, 4 Oct - 5 Oct, 2018.

Presentation Type: Research

Topic: Abstract for 15th Annual Psychology Honours Research Conference

Citation: Chan M, Swann C and Donnelly JF (2019). Are S.M.A.R.T goals really smart? The psychological effects of goal-setting in a learning task. Front. Psychol. Conference Abstract: 15th Annual Psychology Honours Research Conference . doi: 10.3389/conf.fpsyg.2018.74.00020

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence ( https://creativecommons.org/licenses/by/4.0/ ) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions .

Received: 18 Sep 2018; Published Online : 27 Sep 2019.

* Correspondence: Miss. Mun Yu Chan, Department of Psychological Science, School of Health and Human Sciences, Southern Cross University, Coffs Harbour, Australia, [email protected]

Login Required

This action requires you to be registered with Frontiers and logged in. To register or login click here.

• Mun Yu Chan
• Christian Swann
• James F Donnelly

Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Using SMART Goals to Make Scientific Progress

By Alex Szatmary

Thursday, July 14, 2016

My well-organized desk, where SMART goals get completed.

A timer is always running on my desk, as I try to complete my current task before it beeps. This makes me sound like an organized, competitive, methodical person—not so! My desk is covered with mugs, stacks of paper, and cables unneeded for years. Email sits unreplied-to for days.

As a theoretician in the lab of Dr. Ralph Nossal (NICHD), I use mathematical modeling to study how cells get to places in the body. Most of my time is focused on completing clearly written goals born from project plans. A system of timers, project plans, and goals keeps me on track to do what I need to do so that I can get back to the fun part of my job that I would happily do for free. 

Here is some background information on my research project: When a person experiences inflammation, the body directs white blood cells to the inflammation site. In a process called chemotaxis, white blood cells migrate along a gradient of signaling molecules. During a bacterial inflammation, molecules made by the bacteria spread from the inflammation site. When white blood cells sense bacterial molecules, they secrete their own molecule called leukotriene B4 (LTB 4 ). White blood cells can navigate based on signals received from bacterial molecules, as well as by responding to LTB 4 . My overarching goal as a member of Dr. Nossal’s team is to determine how LTB 4 helps white blood cells coordinate their motion, which may clarify the ways various moving cells communicate in other contexts, such as during development and cancer metastasis.

Upon entering the body, bacterial molecules (black circles) stimulate a nearby population of white blood cells (grey objects) to generate LTB 4 (red triangles); the LTB 4 reaches other white blood cells that are too far from the bacteria to sense them directly, enhancing the immune system’s response.

A typical research project requires me to:

• read scientific literature on cell migration and signaling
• write computer code to model cell motion and communication
• communicate with experimentalists to determine what measurements we need
• gather, plot, and analyze the data
• collaborate with colleagues to write a paper that will undergo peer review
• manage the paper through the peer-review process

One of my challenges is that I like writing code more than I like writing prose, and so I can spend a long time working with nothing publishable to show. Writing goals and scheduling my time helps me bridge that gap between purpose and results, to make the most productive use of my time as an IRP postdoc. I wanted to make sure that I limit my coding efforts to what will be relevant to the paper and make steady progress on writing the paper.

Writing and pursuing goals can waste valuable time if the goals are not good. What makes a goal ‘good’? When writing goals, I’ve learned to use the ‘SMART’ criteria to ensure that it’s posed in a way that will move our research efforts forward.

SMART goals are:

In the business world, George T. Dolan pioneered the idea of setting SMART goals back in 1981 (1) . Since then, multiple authors have adapted his concepts to setting objectives for project management and personal development (2) .

Examples of how I employ SMART goals in scientific research:

Goals should not be ambiguous. First, I write down my overall goal and describe precisely what I’m trying to achieve. What do I want to accomplish?

I avoid goals like “Make plots,” because that’s a big, complex goal. It’s better for me to focus by breaking goals down into smaller, targeted parts:

• “Make plots showing how LTB4 concentration varies over time.”
• “Arrange plots to compare cell motion with and without LTB4.”
• “Fix problems noted by colleagues in draft.”

I clarify my specific outcome before I start, which lets me focus on the “what” rather than the “why” of what I’m doing, while I’m doing it.

How do I know when a goal is complete? By evaluating my progress. Each goal I write has a series of objectives that help me make small steps toward achieving the overall goal. These objectives are precise, concrete, and measurable.

Questions I ask when writing my goals include:

• “How am I going to accomplish this goal?”
• “What will I do or learn in the process?”
• To figure out if a goal is measurable, I ask, “How will I know when this is done?”

A goal like “Read papers on chemotaxis” can never be completed— a quick search of PubMed for ‘chemotaxis’ pulls up 36,039 papers . On the other hand, “Read three review papers on chemotaxis” is something I can do this afternoon if I start now. In scientific writing, goals that include word counts can help, because they’re objective and precise, but I prefer goals like, “Write paragraph on results for cell recruitment in early inflammation.”

Measurable doesn’t have to mean completely objective; a goal only has to be clear enough for me to know when I’m making progress on it and when it’s time to stop and do something else.

Who is responsible for making the goal happen? Are expectations clear and agreed upon by all interested parties? On a team collecting and analyzing data, it’s important to identify not just who has which role, but what condition the data should be in when it’s passed from the collectors to the analysts.

Most of the goals I write are assigned to me, but I also record to-dos to remind me to check on things I have asked others to do:

• “Who said they would give me feedback on my paper, and by when? Do they have everything they need?”
• “Did that order for printer toner get made? If not, what needs to happen?”
• “Has my summer student completed a draft of his poster?”

Can I achieve this goal with my current skills and resources? If not, is it feasible to acquire the necessary skills and resources in the goal’s established time frame? Is the time frame appropriate to the complexity and amount of effort the goal requires?

Goals are made to be achieved:

• “Write subsection on modeling LTB 4 transport today” is doable.
• “Write methods section this week” could be realistic, but “Write results and discussion today” probably is not.

Writing unrealistic goals leaves me discouraged when I don’t meet them, so I write goals that I’m confident that I can accomplish. Having a realistic plan lets me tell my collaborators when they can count on having things finished.

I establish a timeline for completing each goal and assessing progress. Is the timeline relevant to my current deadlines, and does it reflect my long-term objectives?

Many goals have a deadline built in, sometimes recurring. For example, I have a poster session coming up at the NIH Research Festival , and I need to prepare weekly lessons for a class I’m teaching. Some projects don’t have a hard deadline, but if I feel like I need to rush to finish a paper before I send out a grant application, it might be too late already.

Similarly, making a career move takes lots of preparation. I’m in the middle of my search for a tenure-track position at a predominantly undergraduate institution. To figure out my career goal, I arranged informational interviews with people in my network. I took workshops on teaching and then taught a class twice through FAES . Then, I prepared a job package and improved it with feedback from friends and mentors. All that could not have been done in the last few months of a fellowship.

I regularly review my plans to clarify what I need to do now rather than next week, even for projects that seem open-ended. It’s fine to have big goals like “submit paper before November,” but I usually break large goals down into things that can be finished in 30 minutes to four hours of work. Tasks much shorter than half an hour can actually take more time to keep track of than to do, so I group related short tasks into a single goal. On the other hand, gauging progress on goals that take more than half a day can be difficult, unless broken down into smaller steps.

Below is an example of my goal tracking in practice, with some notes included on why one of my goals was not achievable as written:

I don’t plan projects and keep track of goals because this way of thinking comes easily to me; I have to track goals explicitly, because I don’t automatically know what needs doing. Benefits of using the SMART criteria when planning and assessing goals include:

• Making it easy for me to figure out what to do next
• Determining what doesn’t really need doing, or what doesn’t need doing right now
• Managing expectations with my mentor and co-workers
• Sensing when it’s time to take a break or work on a fun side project

Research feels slow sometimes. It can also feel intimidating to start writing a manuscript. One of the most satisfying things to me about tracking goals is that, when I feel like I’m not making progress fast enough, I can look at my records and see how much I’ve actually accomplished.

This spring, I used SMART goals to lay out what is needed to turn my modeling work into a paper. Specific, measurable, and assignable goals helped my collaborators understand what data I needed from them on how neutrophils secrete LTB 4 , which also helped them predict what data they would have to present at a conference. Realistic and time-bound goals clarified our options as we decided which hypotheses to test. Two months later, we have drafted a paper we are close to submitting for review.

Thank you to Jennifer Patterson-West for contributing significant efforts to this post.

1. Doran, G. T. (1981). "There's a S.M.A.R.T. Way to Write Management's Goals and Objectives", Management Review, Vol. 70, Issue 11, pp. 35-36. 2. Fuhrmann, C.N., Hobin, J.A., Clifford, P.S., and Lindstaedt, B. (2013) “Goal-Setting Strategies for Scientific and Career Success.” Science Careers. http://sciencecareers.sciencemag.org/career_magazine/previous_issues/articles/2013_12_03/caredit.a1300263

Additional Resources

• http://cds.sdce.edu/decision-making/SMART-Goal-Setting
• This webpage has some useful worksheets for setting S.M.A.R.T. goals and provides a more detailed description of the concept.
• http://professional.opcd.wfu.edu/files/2012/09/Smart-Goal-Setting.pdf

Related Blog Posts

• Six Paths Forward in Biomedical Research
• Find and Replace: DNA Editing Tool Shows Gene Therapy Promise
• Mouth Microbes Turn Treasonous in Gum Disease
• Inflammation Cuts Lifeline for Blood-Producing Stem Cells
• Top-of-the-Line Supercomputer Turbocharges NIH Research

This page was last updated on Monday, January 29, 2024

• Methodology
• Open access
• Published: 19 October 2019

Smart literature review: a practical topic modelling approach to exploratory literature review

• Claus Boye Asmussen   ORCID: orcid.org/0000-0002-2998-2293 1 &
• Charles Møller 1  

Journal of Big Data volume  6 , Article number:  93 ( 2019 ) Cite this article

46k Accesses

183 Citations

3 Altmetric

Metrics details

Manual exploratory literature reviews should be a thing of the past, as technology and development of machine learning methods have matured. The learning curve for using machine learning methods is rapidly declining, enabling new possibilities for all researchers. A framework is presented on how to use topic modelling on a large collection of papers for an exploratory literature review and how that can be used for a full literature review. The aim of the paper is to enable the use of topic modelling for researchers by presenting a step-by-step framework on a case and sharing a code template. The framework consists of three steps; pre-processing, topic modelling, and post-processing, where the topic model Latent Dirichlet Allocation is used. The framework enables huge amounts of papers to be reviewed in a transparent, reliable, faster, and reproducible way.

Introduction

Manual exploratory literature reviews are soon to be outdated. It is a time-consuming process, with limited processing power, resulting in a low number of papers analysed. Researchers, especially junior researchers, often need to find, organise, and understand new and unchartered research areas. As a literature review in the early stages often involves a large number of papers, the options for a researcher is either to limit the amount of papers to review a priori or review the papers by other methods. So far, the handling of large collections of papers has been structured into topics or categories by the use of coding sheets [ 2 , 12 , 22 ], dictionary or supervised learning methods [ 30 ]. The use of coding sheets has especially been used in social science, where trained humans have created impressive data collections, such as the Policy Agendas Project and the Congressional Bills Project in American politics [ 30 ]. These methods, however, have a high upfront cost of time, requiring a prior understanding where papers are grouped by categories based on pre-existing knowledge. In an exploratory phase where a general overview of research directions is needed, many researchers may be dismayed by having to spend a lot of time before seeing any results, potentially wasting efforts that could have been better spent elsewhere. With the advancement of machine learning methods, many of the issues can be dealt with at a low cost of time for the researcher. Some authors argue that when human processing such as coding practice is substituted by computer processing, reliability is increased and cost of time is reduced [ 12 , 23 , 30 ]. Supervised learning and unsupervised learning, are two methods for automatically processing papers [ 30 ]. Supervised learning relies on manually coding a training set of papers before performing an analysis, which entails a high cost of time before a result is achieved. Unsupervised learning methods, such as topic modelling, do not require the researcher to create coding sheets before an analysis, which presents a low cost of time approach for an exploratory review with a large collection of papers. Even though, topic modelling has been used to group large amounts of documents, few applications of topic modelling have been used on research papers, and a researcher is required to have programming skills and statistical knowledge to successfully conduct an exploratory literature review using topic modelling.

This paper presents a framework where topic modelling, a branch of the unsupervised methods, is used to conduct an exploratory literature review and how that can be used for a full literature review. The intention of the paper is to enable the use of topic modelling for researchers by providing a practical approach to topic modelling, where a framework is presented and used on a case step-by-step. The paper is organised as follows. The following section will review the literature in topic modelling and its use in exploratory literature reviews. The framework is presented in “ Method ” section, and the case is presented in “ Framework ” section. “ Discussion ” and “ Conclusion ” sections conclude the paper with a discussion and conclusion.

Topic modelling for exploratory literature review

While there are many ways of conducting an exploratory review, most methods require a high upfront cost of time and having pre-existent knowledge of the domain. Quinn et al. [ 30 ] investigated the costs of different text categorisation methods, a summary of which is presented in Table  1 , where the assumptions and cost of the methods are compared.

What is striking is that all of the methods, except manually reading papers and topic modelling, require pre-existing knowledge of the categories of the papers and have a high pre-analysis cost. Manually reading a large amount of papers will have a high cost of time for the researcher, whereas topic modelling can be automated, substituting the use of the researcher’s time with the use of computer time. This indicates a potentially good fit for the use of topic modelling for exploratory literature reviews.

The use of topic modelling is not new. However, there are remarkably few papers utilising the method for categorising research papers. It has been predominantly been used in the social sciences to identify concepts and subjects within a corpus of documents. An overview of applications of topic modelling is presented in Table  2 , where the type of data, topic modelling method, the use case and size of data are presented.

The papers in Table  2 analyse web content, newspaper articles, books, speeches, and, in one instance, videos, but none of the papers have applied a topic modelling method on a corpus of research papers. However, [ 27 ] address the use of LDA for researchers and argue that there are four parameters a researcher needs to deal with, namely pre-processing of text, selection of model parameters and number of topics to be generated, evaluation of reliability, and evaluation of validity. The uses of topic modelling are to identify themes or topics within a corpus of many documents, or to develop or test topic modelling methods. The motivation for most of the papers is that the use of topic modelling enables the possibility to do an analysis on a large amount of documents, as they would otherwise have not been able to due to the cost of time [ 30 ]. Most of the papers argue that LDA is a state-of-the-art and preferred method for topic modelling, which is why almost all of the papers have chosen the LDA method. The use of topic modelling does not provide a full meaning of the text but provides a good overview of the themes, which could not have been obtained otherwise [ 21 ]. DiMaggio et al. [ 12 ] find a key distinction in the use of topic modelling is that its use is more of utility than accuracy, where the model should simplify the data in an interpretable and valid way to be used for further analysis They note that a subject-matter expert is required to interpret the outcome and that the analysis is formed by the data.

The use of topic modelling presents an opportunity for researchers to add a tool to their tool box for an exploratory and literature review process. Topic modelling has mostly been used on online content and requires a high degree of statistical and technical skill, skills not all researchers possess. To enable more researchers to apply topic modelling for their exploratory literature reviews, a framework will be proposed to lower the requirements for technical and statistical skills of the researcher.

Topic modelling has proven itself as a tool for exploratory analysis of a large number of papers [ 14 , 24 ]. However, it has rarely been applied in the context of an exploratory literature review. The selected topic modelling method, for the framework, is Latent Dirichlet Allocation (LDA), as it is the most used [ 6 , 12 , 17 , 20 , 32 ], state-of-the-art method [ 25 ] and simplest method [ 8 ]. While other topic modelling methods could be considered, the aim of this paper is to enable the use of topic modelling for researchers. For enabling topic modelling for researchers, ease of use and applicability are highly rated, where LDA is easily implemented and understood. Other topic modelling methods could potentially be used in the framework, where reviews of other topic models is presented in [ 1 , 26 ].

The topic modelling method LDA is an unsupervised, probabilistic modelling method which extracts topics from a collection of papers. A topic is defined as a distribution over a fixed vocabulary. LDA analyses the words in each paper and calculates the joint probability distribution between the observed (words in the paper) and the unobserved (the hidden structure of topics). The method uses a ‘Bag of Words’ approach where the semantics and meaning of sentences are not evaluated. Rather, the method evaluates the frequency of words. It is therefore assumed that the most frequent words within a topic will present an aboutness of the topic. As an example, if one of the topics in a paper is LEAN, then it can be assumed that the words LEAN, JIT and Kanban are more frequent, compared to other non-LEAN papers. The result is a number of topics with the most prevalent topics grouped together. A probability for each paper is calculated for each topic, creating a matrix with the size of number of topics multiplied with the number of papers. A detailed description of LDA is found in [ 6 ].

The framework is designed as a step-by-step procedure, where its use is presented in a form of a case where the code used for the analysis is shared, enabling other researchers to easily replicate the framework for their own literature review. The code is based on the open source statistical language R, but any language with the LDA method is suitable for use. The framework can be made fully automated, presenting a low cost of time approach for exploratory literature reviews. An inspiration for the automation of the framework can be found in [ 10 ], who created an online-service, towards processing Business Process Management documents where text-mining approaches such as topic modelling are automated. They find that topic modelling can be automated and argue that the use of a good tool for topic modelling can easily present good results, but the method relies on the ability of people to find the right data, guide the analytical journey and interpret the results.

The aim of the paper is to create a generic framework which can be applied in any context of an exploratory literature review and potentially be used for a full literature review. The method provided in this paper is a framework which is based upon well-known procedures for how to clean and process data, in such a way that the contribution from the framework is not in presenting new ways to process data but in how known methods are combined and used. The framework will be validated by the use of a case in the form of a literature review. The outcome of the method is a list of topics where papers are grouped. If the grouping of papers makes sense and is logical, which can be evaluated by an expert within the research field, then the framework is deemed valid. Compared to other methods, such as supervised learning, the method of measuring validity does not produce an exact degree of validity. However, invalid results will likely be easily identifiable by an expert within the field. As stated by [ 12 ], the use of topic modelling is more for utility than for accuracy.

The developed framework is illustrated in Fig.  1 , and the R-code and case output files are located at https://github.com/clausba/Smart-Literature-Review . The smart literature review process consists of the three steps: pre-processing, topic modelling, and post-processing.

Process overview of the smart literature review framework

The pre-processing steps are getting the data and model ready to run, where the topic-modelling step is executing the LDA method. The post-processing steps are translating the outcome of the LDA model to an exploratory review and using that to identify papers to be used for a literature review. It is assumed that the papers for review are downloaded and available, as a library with the pdf files.

Pre-processing

The pre-processing steps consist of loading and preparing the papers for processing, an essential step for a good analytical result. The first step is to load the papers into the R environment. The next step is to clean the papers by removing or altering non-value-adding words. All words are converted to lower case, and punctuation and whitespaces are removed. Special characters, URLs, and emails are removed, as they often do not contribute to identification of topics. Stop words, misread words and other non-semantic contributing words are removed. Examples of stop words are “can”, “use”, and “make”. These words add no value to the aboutness of a topic. The loading of papers into R can in some instances cause words to be misread, which must either be rectified or removed. Further, some websites add a first page with general information, and these contain words that must be removed. This prevents unwanted correlation between papers downloaded from the same source. Words are stemmed to their root form for easier comparison. Lastly, many words only occur in a single paper, and these should be removed to make computations easier, as less frequent words will likely provide little benefit in grouping papers into topics.

The cleansing process is often an iterative process, as it can be difficult to identify all misread and non-value adding-words a priori. Different papers’ corpora contain different words, which means that an identical cleaning process cannot be guaranteed if a new exploratory review is conducted. As an example, different non-value-adding words exist for the medical field compared to sociology or supply chain management (SCM). The cleaning process is finished once the loaded papers mainly contain value-adding words. There is no known way to scientifically evaluate when the cleaning process is finished, which in some instances makes the cleaning process more of an art than science. However, if a researcher is technically inclined methods, provided in the preText R-package can aid in making a better cleaning process [ 11 ].

LDA is an unsupervised method, which means we do not, prior to the model being executed, know the relationship between the papers. A key aspect of LDA is to group papers into a fixed number of topics, which must be given as a parameter when executing LDA. A key process is therefore to estimate the optimal number of topics. To estimate the number of topics, a cross-validation method is used to calculate the perplexity, as used in information theory, and it is a metric used to evaluate language models, where a low score indicates a better generalisation model, as done by [ 7 , 31 , 32 ]. Lowering the perplexity score is identical to maximising the overall probability of papers being in a topic. Next, test and training datasets are created: the LDA algorithm is run on the training set, and the test set is used to validate the results. The criteria for selecting the right number of topics is to find the balance between a useable number of topics and, at the same time, to keep the perplexity as low as possible. The right number of topics can differ greatly, depending on the aim of the analysis. As a rule of thumb, a low number of topics is used for a general overview and a higher number of topics is used for a more detailed view.

The cross-validation step is used to make sure that a result from an analysis is reliable, by running the LDA method several times under different conditions. Most of the parameters set for the cross-validation should have the same value, as in the final topic modelling run. However, due to computational reasons, some parameters can be altered to lower the amount of computation to save time. As with the number of topics, there is no right way to set the parameters, indicating a trial-and-error process. Most of the LDA implementations have default values set, but in this paper’s case the following parameters were changed: burn-in time, number of iterations, seed values, number of folds, and distribution between training and test sets.

• Topic modelling

Once the papers have been cleaned and a decision has been made on the number of topics, the LDA method can be run. The same parameters as used in the cross-validation should be used as a guidance but for more precise results, parameters can be changed such as a higher number of iterations. The number of folds should be removed, as we do not need a test set, as all papers will be used to run the model. The outcome of the model is a list of papers, a list of probabilities for each paper for each topic, and a list of the most frequent words for each topic.

If an update to the analysis is needed, new papers simply have to be loaded and the post-processing and topic modelling steps can be re-run without any alterations to the parameters. Thus, the framework enables an easy path for updating an exploratory review.

Post-processing

The aim of the post-processing steps is to identify and label research topics and topics relevant for use in a literature review. An outcome of the LDA model is a list of topic probabilities for each paper. The list is used to assign a paper to a topic by sorting the list by highest probability for each paper for each topic. By assigning the papers to the topics with the highest probability, all of the topics contain papers that are similar to each other. When all of the papers have been distributed into their selected topics, the topics need to be labelled. The labelling of the topics is found by identifying the main topic of each topic group, as done in [ 17 ]. Naturally, this is a subjective matter, which can provide different labelling of topics depending on the researcher. To lower the risk of wrongly identified topics, a combination of reviewing the most frequent words for each topic and a title review is used. After the topics have been labelled, the exploratory search is finished.

When the exploratory search has finished, the results must be validated. There are three ways to validate the results of an LDA model, namely statistical, semantic, or predictive [ 12 ]. Statistical validation uses statistical methods to test the assumptions of the model. An example is [ 28 ], where a Bayesian approach is used to estimate the fit of papers to topics. Semantic validation is used to compare the results of the LDA method with expert reasoning, where the results must make semantic sense. In other words, does the grouping of papers into a topic make sense, which ideally should be evaluated by an expert. An example is [ 18 ], who utilises hand coding of papers and compare the coding of papers to the outcome of an LDA model. Predictive validation is used if an external incident can be correlated with an event not found in the papers. An example is in politics where external events, such as presidential elections which should have an impact on e.g. press releases or newspaper coverage, can be used to create a predictive model [ 12 , 17 ].

The chosen method for validation in this framework is semantic validation. The reason is that a researcher will often be or have access to an expert who can quickly validate if the grouping of papers into topics makes sense or not. Statistical validation is a good way to validate the results. However, it would require high statistical skills from the researchers, which cannot be assumed. Predictive validation is used in cases where external events can be used to predict the outcome of the model, which is seldom the case in an exploratory literature review.

It should be noted that, in contrast to many other machine learning methods, it is not possible to calculate a specific measure such as the F-measure or RMSE. To be able to calculate such measures, there must exist a correct grouping of papers, which in this instance would often mean comparing the results to manually created coding sheets [ 11 , 19 , 20 , 30 ]. However, it is very rare that coding sheets are available, leaving the semantic validation approach as the preferred validation method. The validation process in the proposed framework is two-fold. Firstly, the title of the individual paper must be reviewed to validate that each paper does indeed belong in its respective topic. As LDA is an unsupervised method, it can be assumed that not all papers will have a perfect fit within each topic, but if the majority of papers are within the theme of the topic, it is evaluated to be a valid result. If the objective of the research is only an exploratory literature review, the validation ends here. However, if a full literature review is conducted, the literature review can be viewed as an extended semantic validation method. By reviewing the papers in detail within the selected topics of research, it can be validated if the vast majority of papers belong together.

Using the results from the exploratory literature review for a full literature review is simple, as all topics from the exploratory literature review will be labelled. To conduct the full literature review, select the relevant topics and conduct the literature review on the selected papers.

To validate the framework, a case will be presented, where the framework is used to conduct a literature review. The literature review is conducted in the intersection of the research fields analytics, SCM, and enterprise information systems [ 3 ]. As the research areas have a rapidly growing interest, it was assumed that the number of papers would be large, and that an exploratory review was needed to identify the research directions within the research fields. The case used broadly defined keywords for searching for papers, ensuring to include as many potentially relevant papers as possible. Six hundred and fifty papers were found, which were heavily reduced by the use of the smart literature review framework to 76 papers, resulting in a successful literature review. The amount of papers is evaluated to be too time-consuming for a manual exploratory review, which provides a good case to test the smart literature review framework. The steps and thoughts behind the use of the framework are presented in this case section.

The first step was to load the 650 papers into the R environment. Next, all words were converted to lowercase and punctuation, whitespaces, email addresses, and URLs were removed. Problematic words were identified, such as words incorrectly read from the papers. Words included in a publisher’s information page were removed, as they add no semantic value to the topic of a paper. English stop words were removed, and all words were stemmed. As a part of an iterative process, several papers were investigated to evaluate the progress of cleaning the papers. The investigations were done by displaying words in a console window and manually evaluating if more cleaning had to be done.

After the cleaning steps, 256,747 unique words remained in the paper corpus. This is a large number of unique words, which for computational reasons is beneficial to reduce. Therefore, all words that did not have a sparsity or likelihood of 99% to be in any paper were removed. The operation lowered the amount of unique words to 14,145, greatly reducing the computational needs. The LDA method will be applied on the basis of the 14,145 unique words for the 650 papers. Several papers were manually reviewed, and it was evaluated that removal of the unique words did not significantly worsen the ability to identify main topics of the paper corpus.

The last step of pre-processing is to identify the optimal number of topics. To approximate the optimal number of topics, two things were considered. The perplexity was calculated for different amounts of topics, and secondly the need for specificity was considered.

At the extremes, choosing one topic would indicate one topic covering all papers, which will provide a very coarse view of the papers. On the other hand, if the number of topics is equal to the number of papers, then a very precise topic description will be achieved, although the topics will lose practical use as the overview of topics will be too complex. Therefore, a low number of topics was preferred as a general overview was required. Identifying what is a low number of topics will differ depending on the corpus of papers, but visualising the perplexity can often provide the necessary aid for the decision.

The perplexity was calculated over five folds, where each fold would identify 75% of the papers for training the model and leave out the remaining 25% for testing purposes. Using multiple folds reduces the variability of the model, ensuring higher reliability and reducing the risk of overfitting. For replicability purposes, specific seed values were set. Lastly, the number of topics to evaluate is selected. In this case, the following amounts of topics were selected: 2, 3, 4, 5, 10, 20, 30, 40, 50, 75, 100, and 200. The perplexity method in the ‘topicmodels’ R library is used, where the specific parameters can be found in the provided code.

The calculations were done over two runs. However, there is no practical reason for not running the calculations in one run. The first run included all values of number of topics below 100, and the second run calculated the perplexity for 100 and 200 number of topics. The runtimes for the calculations were respectively 9 and 10 h on a standard issue laptop. The combined results are presented in Fig.  2 , and the converged results can be found in the shared repository.

5-Fold cross-validation of topic modelling. Results of cross-validation

The goal in this case is to find the lowest number of topics, which at the same time have a low perplexity. In this case, the slope of the fitted line starts to gradually decline at twenty topics, which is why the selected number of topics is twenty.

Case: topic modelling

As the number of topics is chosen, the next step is to run the LDA method on the entire set of papers. The full run of 650 papers for 20 topics took 3.5 h to compute on a standard issue laptop. An outcome of the method is a 650 by 20 matrix of topic probabilities. In this case, the papers with the highest probability for each topic were used to allocate the papers. The allocation of papers to topics was done in Microsoft Excel. An example of how a distribution of probabilities is distributed across topics for a specific paper is depicted in Fig.  3 . Some papers have topic probability values close to each other, which could indicate a paper belonging to an intersection between two or more topics. These cases were not considered, and the topic with the highest probability was selected.

Example of probability distribution for one document (Topic 16 selected)

The allocation of papers to topics resulted in the distribution depicted in Fig.  4 . As can be seen, the number of papers varies for each topic, indicating that some research areas have more publications than others do.

Distribution of papers per topic

Next step is to process the findings and find an adequate description of the topics. A combination of reviewing the most frequent words and a title review was used to identify the topic names. Practically, all of the paper titles and the most frequent words for each topic, were transferred to a separate Excel spreadsheet, providing an easy overview of paper titles. An example for topic 17 can be seen in Table  3 . The most frequent words for the papers in topic 17 are “data”, “big” and “analyt”. Many of the paper titles also indicate usage of big data and analytics for application in a business setting. The topic is named “Big Data Analytics”.

The process was repeated for all other topics. The names of the topics are presented in Tables  4 and 5 .

Based on the names of the topics, three topics were selected based on relevancy for the literature review. Topics 5, 13, and 17 were selected, with a total of 99 papers. In this specific case, it was deemed that there might be papers with a sub-topic that is not relevant for the literature review. Therefore, an abstract review was conducted for the 99 papers, creating 10 sub-topics, which are presented in Table  6 .

The sub-topics RFID, Analytical Methods, Performance Management, and Evaluation and Selection of IT Systems were evaluated to not be relevant for the literature review. Seventy-six papers remained, grouped by sub-topics.

The outcome of the case was an overview of the research areas within the paper corpus, represented by the twenty topics and the ten sub-topics. The selected sub-topics were used to conduct a literature review. The validation of the framework consisted of two parts. The first part addressed the question of whether the grouping of papers, evaluated by the title and keywords, makes sense and the second part addressed whether the literature review revealed any misplaced papers. The framework did successfully place the selected papers into groups of papers that resemble each other. There was only one case where a paper was misplaced, namely that a paper about material informatics was placed among the papers in the sub-topic EIS and Analytics. The grouping and selection of papers in the literature review, based on the framework, did make semantic sense and was successfully used for a literature review. The framework has proven its utility in enabling a faster and more comprehensive exploratory literature review, as compared to competing methods. The framework has increased the speed for analysing a large amount of papers, as well as having increased the reliability in comparison with manual reviews as the same result can be obtained by running the analysis once again. The transparency in the framework is higher than in competing methods, as all steps of the framework are recorded in the code and output files.

This paper presents an approach not often found in academia, by using machine learning to explore papers to identify research directions. Even though the framework has its limitations, the results and ease of use leave a promising future for topic-modelling-based exploratory literature reviews.

The main benefit of the framework is that it provides information about a large number of papers, with little effort on the researcher’s part, before time-costly manual work is to be done. It is possible, by the use of the framework, to quickly navigate many different paper corpora and evaluate where the researchers’ time and focus should be spent. This is especially valuable for a junior researcher or a researcher with little prior knowledge of a research field. If default parameters and cleaning settings can be found for the steps in the framework, a fully automatic grouping of papers could be enabled, where very little work has to be done to achieve an overview of research directions. From a literature review perspective, the benefit of using the framework is that the decision to include or exclude papers for a literature review will be postponed to a later stage where more information is provided, resulting in a more informed decision-making process. The framework enables reproducibility, as all of the steps in the exploratory review process can be reproduced, and enables a higher degree of transparency than competing methods do, as the entire review process can, in detail, be evaluated by other researchers.

There is practically no limit of the number of papers the framework is able to process, which could enable new practices for exploratory literature reviews. An example is to use the framework to track the development of a research field, by running the topic modelling script frequently or when new papers are published. This is especially potent if new papers are automatically downloaded, enabling a fully automatic exploratory literature review. For example, if an exploratory review was conducted once, the review could be updated constantly whenever new publications are made, grouping the publications into the related topics. For this, the topic model has to be trained properly for the selected collection of papers, where it can be assumed that minor additions of papers would likely not warrant any changes to the selected parameters of the model. However, as time passes and more papers are processed, the model will learn more about the collection of papers and provide a more accurate and updated result. Having an automated process could also enable a faster and more reliable method to do post-processing of the results, reducing the post-analysis cost identified for topic modelling by [ 30 ], from moderate to low.

The framework is designed to be easily used by other researchers by designing the framework to require less technical knowledge than a normal topic model usage would entail and by sharing the code used in the case work. The framework is designed as a step-by-step approach, which makes the framework more approachable. However, the framework has yet not been used by other researchers, which would provide valuable lessons for evaluating if the learning curve needs to be lowered even further for researchers to successfully use the framework.

There are, however, considerations that must be addressed when using the smart literature review framework. Finding the optimal number of topics can be quite difficult, and the proposed method of cross-validation based on the perplexity presented a good, but not optimal, solution. An indication of why the number of selected topics is not optimal is the fact that it was not possible to identify a unifying topic label for two of the topics. Namely topics 12 and 20, which were both labelled miscellaneous. The current solution to this issue is to evaluate the relevancy of every single paper of the topics that cannot be labelled. However, in future iterations of the framework, a better identification of the number of topics must be developed. This is a notion also recognised by [ 6 ], who requested that researchers should find a way to label and assign papers to a topic other than identifying the most frequent words. An attempt was made by [ 17 ] to generate automatic labelling on press releases, but it is uncertain if the method will work in other instances. Overall, the grouping of papers in the presented case into topics generally made semantic sense, where a topic label could be found for the majority of topics.

A consideration when using the framework is that not all steps have been clearly defined, and, e.g., the cleaning step is more of an art than science. If a researcher has no or little experience in coding or executing analytical models, suboptimal results could occur. [ 11 , 25 , 27 ] find that especially the pre-processing steps can have a great impact on the validity of results, which further emphasises the importance of selecting model parameters. However, it is found that the default parameters and cleaning steps set in the code provided a sufficiently valid and useable result for an exploratory literature analysis. Running the code will not take much of the researcher’s time, as the execution of code is mainly machine time, and verifying the results takes a limited amount of a researcher time.

Due to the semantic validation method used in the framework, it relies on the availability of a domain expert. The domain expert will not only validate if the grouping of papers into topics makes sense, but it is also their responsibility to label the topics [ 12 ]. If a domain expert is not available, it could lead to wrongly labelled topics and a non-valid result.

A key issue with topic modelling is that a paper can be placed in several related topics, depending on the selected seed value. The seed value will change the starting point of the topic modelling, which could result in another grouping of papers. A paper consists of several sub-topics and depending on how the different sub-topics are evaluated, papers can be allocated to different topics. A way to deal with this issue is to investigate papers with topic probabilities close to each other. Potential wrongly assigned papers can be identified and manually moved if deemed necessary. However, this presents a less automatic way of processing the papers, where future research should aim to improve the assignments of papers to topics or create a method to provide an overview of potentially misplaced papers. It should be noted that even though some papers can be misplaced, the framework provides outcome files than can easily be viewed to identify misplaced papers, by a manual review.

As the smart literature review framework heavily relies on topic modelling, improvements to the selected topic model will likely present better results. The results of the LDA method have provided good results, but more accurate results could be achieved if the semantic meaning of the words would be considered. The framework has only been tested on academic papers, but there is no technical reason to not include other types of documents. An example is to use the framework in a business context to analyse meeting minutes notes to analyse the discussion within the different departments in a company. For this to work, the cleaning parameters would likely have to change, and another evaluation method other than a literature review would be applicable. Further, the applicability of the framework has to be assessed on other streams of literature to be certain of its use for exploratory literature reviews at large.

This paper aimed to create a framework to enable researchers to use topic modelling to, do an exploratory literature review, decreasing the need for manually reading papers and, enabling the possibility to analyse a greater, almost unlimited, amount of papers, faster, more transparently and with greater reliability. The framework is based upon the use of the topic model Latent Dirichlet Allocation, which groups related papers into topic groups. The framework provides greater reliability than competing exploratory review methods provide, as the code can be rerun on the same papers, which will provide identical results. The process is highly transparent, as most decisions made by the researcher can be reviewed by other researchers, unlike, e.g., in the creation of coding sheets. The framework consists of three main phases: Pre-processing, Topic Modelling, and Post-Processing. In the pre-processing stage, papers are loaded, cleaned, and cross-validated, where recommendations to parameter settings are provided in the case work, as well as in the accompanied code. The topic modelling step is where the LDA method is executed, using the parameters identified in the pre-processing step. The post-processing step creates outputs from the topic model and addresses how validity can be ensured and how the exploratory literature review can be used for a full literature review. The framework was successfully used in a case with 650 papers, which was processed quickly, with little time investment from the researcher. Less than 2 days was used to process the 650 papers and group them into twenty research areas, with the use of a standard laptop. The results of the case are used in the literature review by [ 3 ].

The framework is seen to be especially relevant for junior researchers, as they often need an overview of different research fields, with little pre-existing knowledge, where the framework can enable researchers to review more papers, more frequently.

For an improved framework, two main areas need to be addressed. Firstly, the proposed framework needs to be applied by other researchers on other research fields to gain knowledge about the practicality and gain ideas for further development of the framework. Secondly, research in how to automatically identity model parameters could greatly improve the usability for the use of topic modelling for non-technical researchers, as the selection of model parameters has a great impact on the result of the framework.

Availability of data and materials

https://github.com/clausba/Smart-Literature-Review (No data).

Abbreviations

• Latent Dirichlet Allocation

supply chain management

Alghamdi R, Alfalqi K. A survey of topic modeling in text mining. Int J Adv Comput Sci Appl. 2015;6(1):7. https://doi.org/10.14569/IJACSA.2015.060121 .

Article   Google Scholar  

Ansolabehere S, Snowberg EC, Snyder JM. Statistical bias in newspaper reporting on campaign finance. Public Opin Quart. 2003. https://doi.org/10.2139/ssrn.463780 .

Asmussen CB, Møller C. Enabling supply chain analytics for enterprise information systems: a topic modelling literature review. Enterprise Information Syst. 2019. (Submitted To) .

Atteveldt W, Welbers K, Jacobi C, Vliegenthart R. LDA models topics… But what are “topics”? In: Big data in the social sciences workshop. 2015. http://vanatteveldt.com/wp-content/uploads/2014_vanatteveldt_glasgowbigdata_topics.pdf .

Baum D. Recognising speakers from the topics they talk about. Speech Commun. 2012;54(10):1132–42. https://doi.org/10.1016/j.specom.2012.06.003 .

Blei DM. Probabilistic topic models. Commun ACM. 2012;55(4):77–84. https://doi.org/10.1145/2133806.2133826 .

Blei DM, Lafferty JD. A correlated topic model of science. Ann Appl Stat. 2007;1(1):17–35. https://doi.org/10.1214/07-AOAS114 .

Article   MathSciNet   MATH   Google Scholar  

Blei DM, Ng AY, Jordan MI. Latent Dirichlet Allocation. J Mach Learn Res. 2003;3:993–1022. https://doi.org/10.5555/944919.944937 .

Article   MATH   Google Scholar  

Bonilla T, Grimmer J. Elevated threat levels and decreased expectations: how democracy handles terrorist threats. Poetics. 2013;41(6):650–69. https://doi.org/10.1016/j.poetic.2013.06.003 .

Brocke JV, Mueller O, Debortoli S. The power of text-mining in business process management. BPTrends.

Denny MJ, Spirling A. Text preprocessing for unsupervised learning: why it matters, when it misleads, and what to do about it. Polit Anal. 2018;26(2):168–89. https://doi.org/10.1017/pan.2017.44 .

DiMaggio P, Nag M, Blei D. Exploiting affinities between topic modeling and the sociological perspective on culture: application to newspaper coverage of U.S. government arts funding. Poetics. 2013;41(6):570–606. https://doi.org/10.1016/j.poetic.2013.08.004 .

Elgesem D, Feinerer I, Steskal L. Bloggers’ responses to the Snowden affair: combining automated and manual methods in the analysis of news blogging. Computer Supported Cooperative Work: CSCW. Int J. 2016;25(2–3):167–91. https://doi.org/10.1007/s10606-016-9251-z .

Elgesem D, Steskal L, Diakopoulos N. Structure and content of the discourse on climate change in the blogosphere: the big picture. Environ Commun. 2015;9(2):169–88. https://doi.org/10.1080/17524032.2014.983536 .

Evans MS. A computational approach to qualitative analysis in large textual datasets. PLoS ONE. 2014;9(2):1–11. https://doi.org/10.1371/journal.pone.0087908 .

Ghosh D, Guha R. What are we “tweeting” about obesity? Mapping tweets with topic modeling and geographic information system. Cartogr Geogr Inform Sci. 2013;40(2):90–102. https://doi.org/10.1080/15230406.2013.776210 .

Grimmer J. A Bayesian hierarchical topic model for political texts: measuring expressed agendas in senate press releases. Polit Anal. 2010;18(1):1–35. https://doi.org/10.1093/pan/mpp034 .

Grimmer J, Stewart BM. Text as data: the promise and pitfalls of automatic content analysis methods for political texts. Polit Anal. 2013;21(03):267–97. https://doi.org/10.1093/pan/mps028 .

Guo L, Vargo CJ, Pan Z, Ding W, Ishwar P. Big social data analytics in journalism and mass communication. J Mass Commun Quart. 2016;93(2):332–59. https://doi.org/10.1177/1077699016639231 .

Jacobi C, Van Atteveldt W, Welbers K. Quantitative analysis of large amounts of journalistic texts using topic modelling. Digit J. 2016;4(1):89–106. https://doi.org/10.1080/21670811.2015.1093271 .

Jockers ML, Mimno D. Significant themes in 19th-century literature. Poetics. 2013;41(6):750–69. https://doi.org/10.1016/j.poetic.2013.08.005 .

Jones BD, Baumgartner FR. The politics of attention: how government prioritizes problems. Chicago: University of Chicago Press; 2005.

Google Scholar  

King G, Lowe W. An automated information extraction tool for international conflict data with performance as good as human coders: a rare events evaluation design. Int Org. 2008;57:617–43. https://doi.org/10.1017/s0020818303573064 .

Koltsova O, Koltcov S. Mapping the public agenda with topic modeling: the case of the Russian LiveJournal. Policy Internet. 2013;5(2):207–27. https://doi.org/10.1002/1944-2866.POI331 .

Lancichinetti A, Irmak Sirer M, Wang JX, Acuna D, Körding K, Amaral LA. High-reproducibility and high-accuracy method for automated topic classification. Phys Rev X. 2015;5(1):1–11. https://doi.org/10.1103/PhysRevX.5.011007 .

Mahmood A. Literature survey on topic modeling. Technical Report, Dept. of CIS, University of Delaware Newark, Delaware. http://www.eecis.udel.edu/~vijay/fall13/snlp/lit-survey/TopicModeling-ASM.pdf . 2009.

Maier D, Waldherr A, Miltner P, Wiedemann G, Niekler A, Keinert A, Adam S. Applying LDA topic modeling in communication research: toward a valid and reliable methodology. Commun Methods Meas. 2018;12(2–3):93–118. https://doi.org/10.1080/19312458.2018.1430754 .

Mimno D, Blei DM. Bayesian checking for topic models. In: EMLP 11 proceedings of the conference on empirical methods in natural language processing. 2011. p 227–37. https://doi.org/10.5555/2145432.2145459

Parra D, Trattner C, Gómez D, Hurtado M, Wen X, Lin YR. Twitter in academic events: a study of temporal usage, communication, sentimental and topical patterns in 16 Computer Science conferences. Comput Commun. 2016;73:301–14. https://doi.org/10.1016/j.comcom.2015.07.001 .

Quinn KM, Monroe BL, Colaresi M, Crespin MH, Radev DR. How to analyze political attention. Am J Polit Sci. 2010;54(1):209–28. https://doi.org/10.1111/j.1540-5907.2009.00427.x .

Xu Z, Raschid L. Probabilistic financial community models with Latent Dirichlet Allocation for financial supply chains. In: DSMM’16 proceedings of the second international workshop on data science for macro-modeling. 2016. https://doi.org/10.1145/2951894.2951900 .

Zhao W, Chen JJ, Perkins R, Liu Z, Ge W, Ding Y, Zou W. A heuristic approach to determine an appropriate number of topics in topic modeling. BMC Bioinform. 2015;16(13):S8. https://doi.org/10.1186/1471-2105-16-S13-S8 .

Download references

Acknowledgements

Not applicable.

Author information

Authors and affiliations.

Department of Materials and Production, Center for Industrial Production, Aalborg University, Fibigerstræde 16, 9220, Aalborg Øst, Denmark

Claus Boye Asmussen & Charles Møller

You can also search for this author in PubMed   Google Scholar

Contributions

CBA wrote the paper, developed the framework and executed the case. CM Supervised the research and developed the framework. Both authors read and approved the final manuscript.

Corresponding author

Correspondence to Claus Boye Asmussen .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and permissions

About this article

Cite this article.

Asmussen, C.B., Møller, C. Smart literature review: a practical topic modelling approach to exploratory literature review. J Big Data 6 , 93 (2019). https://doi.org/10.1186/s40537-019-0255-7

Download citation

Received : 26 July 2019

Accepted : 02 October 2019

Published : 19 October 2019

DOI : https://doi.org/10.1186/s40537-019-0255-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Supply chain management
• Automatic literature review

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Geriatrics (Basel)

Healthy Lifestyle Behavior, Goal Setting, and Personality among Older Adults: A Synthesis of Literature Reviews and Interviews

Ming yu claudia wong.

1 Department of Health and Education, The Education University of Hong Kong, Hong Kong, China

Kai-ling Ou

2 Department of Sport, Physical Education and Health, Hong Kong Baptist University, Hong Kong, China

Pak Kwong Chung

Associated data.

Not applicable.

Background: Despite the well-known health benefits of adopting a healthy lifestyle, older adults’ self-determination, goals, and motivation, as well as other personality factors, are known to influence their healthy lifestyle behaviors, yet these interactions have rarely been discussed. Method: The literature that investigated and discussed the interaction of personality, goals, and healthy lifestyle behaviors among older adults was reviewed. In addition, interview responses from older adults regarding their experiences in participating in a real-life physical activity intervention and its relationship with their personality traits and goal setting were synthesized using content analysis. Results: The current review highlights the relationship between healthy living practices, goal setting, and personalities, and it is backed up and expanded upon by interviews with participants. People with different personality types are likely to have diverse views on HLBs. Individuals who are more conscientiousness or extraverted are more likely to adopt HLBs than those who are not. Discussion: It is suggested that a meta-analysis should be conducted on the relationship between personality, goal setting, and physical exercise or other specific HLBs. In addition, future research should focus on various types of HLB therapies that take into account personality and goal setting.

1. Highlights

• The current review highlights the relationship between healthy living practices, goal setting, and personality.
• Conscientiousness, neuroticism, and extraversion were shown to have significant effects on health-related behaviors as well as actual healthy lifestyle behaviors.
• Future research should focus on various types of HLB therapies that take into account personality and goal-setting processes.

2. Background

Despite the well-known benefits of engaging in a healthy lifestyle [ 1 , 2 , 3 ], older adults are expected to have a less goal-oriented mindset than teenagers and younger adults. This is because older adults have a weaker sense of pursuing life goals or they have a scheduled routine for their life that facilitates self-regulation. Therefore, older adults’ self-determination, goals, and motivation to engage in a healthy lifestyle, as well as other personality factors, are seen to influence their healthy lifestyle behaviors. A few well-documented factors have been proven to have an impact on older adults’ adherence to a healthy lifestyle or engagement in physical activities in particular, including believing in the benefits of exercising and having experience with exercise, setting goals, and having certain personality traits [ 4 ]. Considering the unchangeability of older adults’ past experiences, setting appropriate goals based on individual personalities is expected to be useful in fostering older adults’ participation in physical activity or adherence to other healthy lifestyle behaviors. The existing literature describes the relationship between personality and goals among older adults, with their outcomes associated with subjective well-being, quality of life, and other mental benefits [ 5 , 6 ]. Previous research also indicated that various sociocultural variables, such as gender, education level, the sense of autonomy, and the physical capability of older adults, should be taken into account when determining their attitudes toward and intention to participate in physical activities. It is believed that various factors will negatively affect older adults’ personality, autonomy, and awareness of physical activity goal setting, thus affecting their healthy lifestyle behaviors [ 7 , 8 ].

According to the literature, life goals are considered as an internal mental conception of desired outcomes or activities that a person aims to pursue in their daily life [ 9 ]. Based on the self-determination theory, individuals emphasize the importance of goal content rather than only having objectives and claim that these types of goals are important. Personality is seen as a factor influencing the construction of goals [ 10 ]. Psychologists view personality traits and life goals as part of a distinct concept or theory, with personality traits being comparatively stable and consistent, whereas life goals might fluctuate due to life changes. Other psychologists [ 11 , 12 ] claim that goals or life tasks are the “doing” or “acting” side of the personality. Moreover, goals are described as dynamic components of personality that reflect one’s interaction with the environment across time. As a result, goals serve as a link between personality traits and behavior. Based on establishing the concept of goals as personality-in-context [ 13 ], a research study showed that extraversion was associated with understanding the importance of intrinsic goals, having good health, and making progress in social goals among older adults [ 6 , 14 ]. Conscientiousness results in higher levels of health and social progress, while neuroticism only results in lower progress toward social goals and no progress in health goals [ 14 ]. Furthermore, individuals with higher neuroticism are more sensitive to stress and have significantly moderate progress in health and social goals, while conscientiousness individuals who are sensitive to stress have significantly moderate progress only in health goals, and for extraverts, there is no moderation effect on any goal progress [ 14 ]. Nevertheless, optimism is associated with goal attainment and continuity and leads to good health [ 5 , 15 ]. Unsurprisingly, goal continuity showed a negative correlation with neuroticism, and a positive correlation with conscientiousness [ 15 ]. However, a research study also claimed that the relationship between goal orientation and proactive health-related coping behavior depends on the stressors, yet this has not been completely explained by individual differences [ 16 ].

Although the relationship between goals and personality has been investigated in previous studies, to the best of our best knowledge, the relationship between personality and particular health goals among older adults has been less discussed. With regard to the importance of healthy and active aging, pursuing health-related goals is essential for older adults. Meanwhile, the interaction between personality and appropriate health goals among older adults is not yet known. Therefore, a discussion on the interaction of personality, goals, and healthy lifestyle behaviors among older adults should be presented.

3. Objectives

The purpose of this review is to emphasize the importance of personality and goal setting in older adults’ participation in healthy lifestyle behaviors. The body of research on this topic has been limited; therefore, as part of this study, we reviewed the prior literature to examine the interactions between personality, goals, and healthy lifestyle behaviors among older adults. We synthesized the literature review to address the first research question (RQ1): How does the interaction between personality, goals, and healthy lifestyle behaviors (e.g., physical activity and eating habits) manifest among adults? Second, we analyzed interviews with older adults who participated in the authors’ previous physical activity intervention program to determine how their participation was associated with their personality traits and goals. To achieve this objective, we formulated the second research question (RQ2): How do interviewees describe their most satisfying goals for facilitating participation in healthy lifestyle behaviors?

4.1. Literature Search

The two authors, a post-doctoral research fellow and a PhD student, each with extensive knowledge on physical health and developmental psychology, conducted a literature search on the body of research that investigates the relationship between personality and health-related goals. The Scopus and Web of Science databases were searched using the keywords (“Older Adults” AND “Goals” AND “Personality) AND (“Healthy Lifestyle Behaviour” OR “Physical Activity” OR “Eating Habits”). The extracted research articles were imported to EndNote for management. After removing duplicates, the authors screened the titles and abstracts of the retrieved studies independently to identify relevant research articles. Disagreements were settled through discussions. Considering the generalizability of the literature synthesis covering adults to young, old, and older adults, the only inclusion/exclusion criterion for screening relevant papers was to exclude studies that targeted the samples of subjects aged 18 or below. Studies that mentioned older adults were also included. Appropriate papers were then identified and processed for a full-text review. The literature synthesis mainly targeted the following information: (1) types of personality traits, (2) types of goals, (3) content of goals, (4) types of healthy lifestyle behaviors, and (5) how they interact.

4.2. Interview

A total of 20 participants from the authors’ previous intervention program [ 17 ] were invited to participate in semi-structured in-depth interviews after completing the program. These 20 people participated in either tai chi or resistance training interventions during three 1 h sessions per week over 18 weeks. Questions addressed participants’ goals in the intervention program and for daily healthy living, their perceived impact of personality on goal setting, forms of preferred goals, and the attractiveness of extrinsic rewards. The interview guide included the following questions:

• 1. Did you set any goals for yourself before this program? What were they?
• 2. Think about your overall health and fitness and ability to get around and do the things you want to do. What are your wishes and hopes for that in the future?
• 2a. Do you think this program/participating in physical activity has facilitated you to do so, or at least achieve a certain extent of that?
• 3. Again, think about your health and fitness and ability to get around and do the things you want to do. What are your fears and worries about the future?
• 3a. Do you think this program/participating in physical activity could eliminate these fears or worries?
• 4. What kinds of goals do you tend to set for yourself, in terms of physical activity/health-related factors (e.g., diet, exercise)?
• 5. What kind of personality do you think you have?
• 5a. Do you think your personality affects your goals for physical activity and overall health? How?
• 6. Do you prefer having standardized static goals or personalized goals when participating in an exercise program with coach supervision?
• 7. Do you prefer having a single ultimate goal or integrated smaller goals (goal phrase) before achieving the ultimate goal?
• 8. To what extent could rewards prompt you to achieve your goals?

4.3. Data Analysis

The literature review was synthesized and interview responses were analyzed by first creating an open-coding label to identify relevant information and factors, individual differences, types of goals, and types of healthy lifestyle behaviors, as well as potential interactions between factors. The above-mentioned study objectives provided the basis for the initial categorization. Then, the relevant open codes were refined and combined by labelling with an “analytical” theme. After this process, the second coder went through the same process together, expressing agreement with the first coder’s labels, indicating inter-rater reliability. The audio recordings of the interviews were transcribed verbatim and translated from Chinese to English. The verbatim transcripts were screened by another author to eliminate translation variations, strengthening the qualitative research’s credibility. The data were managed by using NVivo 12 (QSR, 2020) using memo writing, coding, and categorizing until data saturation was reached. A similar inter-rater reliability process was used for analyzing the interview’s transcripts.

5.1. Literature Review and Interview Demographics

A total of 22 research articles were identified during the database search. Figure 1 displays the flowchart of the search. The research articles that were retrieved and included in the synthesis mentioned the effects of personalities on healthy lifestyle behavior (HLB)-related goals among older adults as well as younger adults. A total of 20 people participated in the in-depth interviews:17 women and 3 men; the age range for 17 of them was 65–69 years, and for 3 of them, it was 70–74 years. The inter-rater reliability showed 95% agreement in both the literature synthesis and interview transcript analysis.

Flowchart of the literature search [ 18 ].

5.1.1. Types of Healthy Lifestyle Behavior

Most of the 22 research articles mentioned the types of healthy lifestyle behavior that may be affected by personality traits and goal achievement. They include exercising, participating in sports, eating a healthy diet, participating in physical activity using electronic devices [ 19 , 20 ], participating in leisure-time physical activity that is “planned, structured, repeated and with maintenance” [ 21 ], controlling one’s weight, managing one’s nutrition [ 22 ], engaging in social activities, and having intentions and beliefs about physical activity [ 23 ]. The literature indicated that people who engage in these HLBs tend to have better physical, mental, emotional, and cognitive health [ 24 ].

5.1.2. Personality Types and Healthy Lifestyle Behaviors

Among the 22 research articles, personality traits were mostly represented by the “Big Five” traits [ 21 , 23 , 25 , 26 , 27 , 28 , 29 , 30 , 31 ] and measured by the Big Five Inventory [ 32 ]. Research has shown that personality traits play an important role in influencing people’s health and HLB choices throughout their lives [ 28 ]. Studies have shown that people who are highly extraverted, highly conscientious, and not neurotic tend to have higher levels of self-efficacy, self-motivation, and self-control [ 21 ] and, thus, engage in more leisure-based physical activities. An intervention study showed that for people with a high level of conscientiousness, their step count was positively predicted from a pre-test, while under monitoring during the intervention, people with a high level of neuroticism showed a significant increase in their daily step counts as well [ 30 ]. Meanwhile, people with a sense of openness tended to be slightly more active, but there were no significant differences in the level of leisure-based physical activity among people with agreeableness [ 31 , 33 , 34 ]. In addition to people with high neuroticism, people with other types of personality traits were associated with a high level of engagement in leisure activities, including social and physical activities, which indicates successful active aging [ 29 ].

Other than the Big Five personality traits, the relationship between conscientiousness and HLB was shown using the effect of perfectionism [ 35 , 36 ] and measured using self-oriented and socially prescribed perfectionism subscales [ 37 ]. Despite its association with a high level of self-discipline, perfectionism was also associated with extreme behaviors such as binge eating, which is considered to have a deleterious effect on physical and mental health [ 36 ]. The opposite of perfectionism, according to Sirois [ 24 ], may be chronic procrastination, which is defined as a trait-like personal characteristic. It was shown to have an association with difficulties in self-regulation and the avoidance of engaging in HLBs such as physical activity and healthy eating habits, thus resulting in poor physical health and well-being [ 38 , 39 , 40 ]. Some research studies [ 41 , 42 ] discussed personalities in a more general manner by dividing them into optimistic and pessimistic, as measured by the Life Orientation Test Revised (LOT-R) [ 43 ]. They showed that optimistic individuals tend to have a higher mental health-related quality of life [ 42 ], as well as favorable physical health outcomes [ 44 , 45 ]. Another study [ 22 ] investigated personality measurements in terms of the Type D personality, which is correlated with obesity, unhealthy lifestyle behavior, and vulnerability [ 22 , 46 ].

5.1.3. Healthy Lifestyle Behavior, Goals, and Personality Traits

There are multiple types of goals, including achievement, maintenance, disengagement, engagement, and compensation goals [ 47 ], and the retrieved research studies mostly focused on achievement and disengagement goals. Achievement goals were examined and applied mostly in the context of healthy lifestyle behavior interventions, which consisted of physical activity, nutrition management, stress management, and cognitive function training. Achievement goals in interventions tended to be mutual goals regardless of personality traits [ 22 ]. On the other hand, research on M-health indicated that electronic devices, such as a Fitbit, were considered to be effective in facilitating goal setting and achievements. This is because the planning, learning [ 19 ], rewarding, comparing, and sharing functions in electronic devices can trigger individuals’ motivation to achieve their desired goals [ 30 , 48 ]; it is also noteworthy that electronic devices can support personalized goals based on individual personality traits.

Engagement goals involve more inflexible pursuits, and research has shown that disengagement goals with more flexibility (e.g., short-term or long-term goals) for adjustment were also important for achieving a healthy life [ 49 , 50 ]. Apart from theoretically supported goals, qualitative research has also revealed other forms of desire goals or preventive goals related to achieving HLBs, which were labelled as “hoped-for possible self” and “feared possible self” [ 24 ]. The “hoped-for” goals include being physically active, being a vegetarian, reducing weight, etc. Fear-related goals include preventing oneself from becoming obese, preventing heart disease, preventing weak lower limbs, etc. These kinds of goals can also encourage individuals to engage in HLBs. They are also considered to be similar to desired goals and “anti-goals” [ 25 ]. Research has shown that individuals stick with HLBs when they have goal-relevant information, such as a concept of their ideal self (both physically and psychologically) and health-related information [ 31 ]. Goals, with meaningful purpose and rewards for achievement, were also considered as effective in achieving a healthy lifestyle.

Persistent and perceived goal progress can also be influenced by individual personality traits [ 47 ]; for instance, people with a higher level of conscientiousness tend to show better self-control and self-regulation and engage in self-corrective actions when pursuing goals [ 21 , 51 ]. Moreover, conscientiousness is strongly associated with a high level of perfectionism, which also correlates with higher personal standards and high-order goals [ 36 ], as well as the high-level self-evaluation of goal achievement [ 26 ], which might mean vulnerability in terms of the mental status or even health (e.g., eating disorders or extreme eating as related to exercise goals) [ 35 , 52 ].

5.1.4. Interaction between Personality, Goals, and Healthy Lifestyle Behaviors

In general, the retrieved studies demonstrated the inter-relationships between personality, goals, and HLBs. Some of the studies also mentioned the lack of a discussion on the role of personality in health-related goal settings and HLBs, which has been shown to have implications in terms of affecting the motivation and perception of HLB among older adults [ 27 , 47 , 50 ].

The studies indicated that personality plays a role in health-related goal setting and HLBs. Only people with a high level of conscientiousness are significantly affected by the product judgments and health-relevant information that they receive, which affects their decision making with respect to dietary choices and physical activity related to their overall goal of a healthy lifestyle [ 31 ]. As mentioned above, conscientiousness has been associated with higher levels of self-control, and as discussed in [ 21 ], people with self-control tend to give priority to long-term goals over short-term goals, can resist goal-disrupting temptations [ 53 , 54 ], and have strong goal-striving abilities [ 28 ], such as avoiding junk food or staying engaged in healthy acts on a daily basis, in order to pursue their long-terms goals consistently. Hence, people with a higher level of self-control, similarly to conscientiousness, are significantly associated with positive subjective well-being and physical activity, mediated by high levels of perceived goal progress and self-efficacy [ 21 , 23 ]. Similarly, Briki and Dagot [ 25 ] demonstrated that dispositional self-control and perceived goal progress were negatively associated with neurotic self-attentiveness, thus negatively predicting subjective well-being. Furthermore, people with lower levels of conscientiousness and optimism are associated with low goal re-engagement, higher mental fatigability, and poorer physical and cognitive health maintenance [ 33 ]. Meanwhile, optimism and conscientiousness are positively associated with goal achievement, adjustments, and re-engagement and thus are related to older adults’ participating in physical activity and having a healthy quality of life [ 30 ]. However, people who exhibit extreme conscientiousness, or perfectionism, tend to overvaluate and think dichotomously about their goal progress, such as weight and body shape, which might lead to unhealthy eating habits or even an eating disorder [ 35 , 36 ]. On the other hand, people who procrastinate tend to have lower expectations for “hoped-for goals” and place no importance on avoiding “feared-for” possible outcomes; thus, they have less intention to change their health behaviors [ 24 ].

Intervention studies [ 22 ] showed that intervention programs that followed an adaptation of goal attainment theories facilitating the development of participants’ mutual goals were able to improve the HLBs of those with Type D personality, reducing their mental vulnerability. Moreover, the vulnerability and established healthy habits of Type D participants were also shown to be influenced by the social support gained from the intervention. Another study [ 41 ] demonstrated that people who expressed optimism had higher levels of health-related self-efficacy after the intervention and were able to reduce their waist circumference (lose weight) even one year after the intervention. This indicates that optimistic people have better self-control than pessimistic people with regard to health-related goals and behaviors.

There was a trend of increasing research investigating the self-determination effect of using electronic health devices, such as smart watches, on goal monitoring and physical activity levels. Studies indicated that even though the functions of these electronic devices could trigger the self-monitoring and self-regulation of HLBs by individuals, including older adults, the research results showed that personality also plays a role in people’s levels of regulation and HLBs. Hence, it is recommended that smart devices provide personalized goal setting systems in which big data can be used to provide goals or HLB plans that are tailored to users’ individual differences and personalities [ 20 ]. Bischoff [ 19 ] highlighted the importance of goal-specific apps in smart devices in being able to identify individual differences, including personality traits, levels of vulnerability, and usual health behaviors, and being able to provide suitable goal planning and management strategies, as well as relevant features for pursuing HLBs. In addition, based on the differences in physical activity outcomes among people with different personality traits in intervention studies, including personalized interventions and individualized goals based on people’s personality traits, priorities, and attitudes is also suggested, instead of standardized goals and procedures, in order to achieve positive and successful aging [ 48 , 55 , 56 ]. In addition to individualized goals, promoting programs that offer individual care, verbal encouragement from coaches or instructors, and personalized regular schedules in both research and community settings is suggested to enhance older adults’ HLB [ 56 ].

5.2. Participant Interviews

5.2.1. personality.

After the interview, all participants were asked to fill in a 16-item personality test based on a framework that evolved from the Myers–Briggs Type Indicator (MBTI) [ 57 ]. It indicates personality traits in four dimensions: (1) preferred orientation to life: extraversion (E) or introversion (I); (2) preferred way of perceiving things: sensing (S) or intuition (N); (3) preferred way of making decisions: thinking (T) or feeling (F); and (4) preferred way of dealing with the world: judging (J) or perceiving (P) [ 58 ].

To be more consistent with the literature synthesis, the relationship between MBTI and the Big Five personality traits was determined. A previous study found that the T-F dimension was correlated with agreeableness and the J-P dimension correlated with conscientiousness; the E-I dimension was strongly correlated with extraversion, and neuroticism was not correlated with any MBTI dimensions [ 59 ]. However, when separating the subfactors, researchers [ 60 ] later found that E was correlated with extraversion. Openness was significantly associated with N and inversely correlated with S, which involves a direct involvement with information. Agreeableness was negatively correlated with T but positively correlated with F, which means more concern with feelings. Conscientiousness was positively correlated with J and negatively correlated with P, which indicates people who are orderly, deliberate, and self-disciplined. Neuroticism was positively correlated with I but negatively correlated with E; the mental processes of such people are more oriented toward their inner world, which is highly associated with self-consciousness, depression, and anxiety [ 59 ].

The results showed that the participants were mainly divided into four types: ESFJ (12), ISFJ (6), INFP (1), and ESTJ (1). According to the findings and combining them with the Big Five personality types, we can clarify that the personality types of these participants are as follows: ESFJ—extroversion, based on information perception, agreeableness, and conscientiousness; ISFJ—introversion (or neuroticism), which relies on information perception, agreeableness, and conscientiousness; INFP—introversion (or neuroticism), openness, agreeableness, and conscientiousness; and ESTJ—extroversion, based on information perception, logical thinking, and conscientiousness.

5.2.2. Goals and Healthy Lifestyle Behavior

• Hoped-for possible self

Participants recalled that before they joined the intervention program, their expectations (goals) for the program were mainly to maintain their physical activity level, improve their physical health, expand their social networks, learn knowledge, and feel a sense of persistence. Only three participants said that they did not set any goals.

“My goal in retirement is to do more exercises, stay healthy and have fewer doctor visits" (F, R, 10, ISFJ). “Reducing fatty liver” (F, T, 12, ESFJ). “Improve balance and body cold” (F, T,6, ISFJ). “Flexibility in arms and legs and good health” (F, T,14,ESFJ).

They want to use what they have learned to educate others: “My aim is not only to help myself but also others, as well as teach my friends Tai Chi together” (F, T, 14, ESFJ). Another participant agreed: “I can also teach my family, I will be happy and initiate other older adults” (F, T, 12,ESFJ).

Some participants wanted to learn about exercise theory to help them maintain their physical activity levels for their entire lives: “I hope that the intervention can open up the path to understanding sports and learn some sports that I can do at home so that I can do them regularly without the help of others” (F, R, 7, ESTJ); “I hope the knowledge of Tai Chi taught by my teacher can be my lifelong asset” (F, T, 2, ESFJ); “I usually do exercise, but I don’t have enough theoretical knowledge, so I sometimes overexert myself and hurt my knee” (F, R, 15, ESFJ).

One participant noted that perseverance was the goal of program participation: “Try not to miss a single day of training” (F, R, 2, ESFJ).

Those with no goals were mainly inactive and had no sports experience: “I haven’t tried resistance training, I don’t have any goal in doing sports” (F, R, 18, ESFJ); “I don’t have special goals, I usually do stretching at home, I haven’t participated in any other sports activity” (F, R, 19, ESFJ).

• Fear of possible self

The worries expressed by participants were mainly having limited physical fitness, becoming injured, and a lack of determination: “I was worried that my muscles were not strong enough or that I was not able to do the movements in the class” (F, R, 11, ESFJ); “I was afraid of Tai Chi standing might hurt my knee” (F, T, 14, ESFJ); “I was afraid I wouldn’t be able to do it [resistance] as I was almost 70 years old and I was afraid it would be too drastic” (F, R, 19, ESFJ); “I was worried that I would give up halfway because I don’t know if the exercise is too difficult” (F, R, 7, ESTJ).

However, all participants said that the intervention eventually eliminated their concerns. Because some participants believed in the professionalism of the instructors, their stereotypes regarding physical activity changed after learning from the intervention: “Because the instructor is very experienced and knows what age we are, they won’t force us to do it if we can’t” (F, R, 18, ESFJ). A female participant (F, T, 14, ESFJ), after practicing tai chi, said the following: “Standing on one leg in Tai Chi helps me to maintain my balance and I know how to adjust when I fall”.

• Long term health-related goals

Participants said that they set long-term health-related goals after the intervention and they reported being more conscious of healthy eating habits after participating: “I will pay attention to health information, such as the ingredients of food, to avoid food allergies” (F, R, 10, ISFJ); “I will maintain my weight and take foods that maintain muscle, such as protein” (F, R, 11, ESFJ); “Eat less fried and sweet foods” (M, R, 1, ESFJ); “Eat lighter, less diet out, make breakfast cooked at home” (M, T, 4, ISFJ); “Eat more vegetables, grains and cereals” (F, R, 15, ESFJ).

Regarding physical health, participants stated that the recovering function was their long-term health goal: “I hope my lower limb function can be improved, stronger feet and bones” (F, T, 9, ESFJ); “reducing muscle loss” (F, R, 10, ISFJ).

Meanwhile, participants had a clearer vision of further goal setting for physical activity. A male participant was determined to use technology to monitor his health: “I will be more aware of different activities. … I use a pedometer to set goals for myself and recently I have been using another function to lift myself up to drink” (M, R, 3, ISFJ). In addition, the majority emphasized cultivating the habit of engaging in physical activity, having perseverance when performing exercises, and reducing physical deterioration as their long-term aims: “The long-term goal is to exercise two to three days a week for one to one and a half hours each time” (F, R, 2, ESFJ); “My new goal is how to remain motivated to exercise” (F, R, 17, ISFJ); “I hope I can practice Tai Chi every day, I am satisfied if my body deterioration is slowed down” (F, T, 6, ISFJ). Moreover, one participant (F, T, 12, ESFJ) believed that achieving long-term health-related goals needed time for learning and building up: “It needs time to build up. I want to further learn Tai Chi stances, change the eating pattern and focus on improving healthy lifestyles”.

Importantly, a female participant (F, R, 18, ESFJ) mentioned that health is not the same as longevity and that being able to live independently is critical: “The goal is not to live a long life, but to be able to walk even at an advanced age, without the need for assistance or difficulty in walking”.

Finally, a participant said that a further goal was not only to promote health for herself but also to encourage others: “The goal is now to encourage other people to do exercise together, not just myself” (F, R, 11, ESFJ).

• Goal preference for intervention/program: generalized vs. personalized aims

Participants’ opinions were mixed, with 10 preferring generalized aims, 9 preferring personalized aims, and 1 saying both were important.

Participants believe generalized aims make them more motivated, encourage them, make them happy, and make teaching easier: “It is motivating to work together as a team towards a goal” (M, R, 1, ESFJ); “Generalized aims are more encouraging” (F, T, 9, ESFJ); “We are happier to have generalized aims because they can last for a longer time” (F, R, 17, ISFJ); “The course was too short with only 4 months. It would be too difficult for the instructor to get to know everyone in a short period of time if there was another opportunity to set up personalized aims for each person in the future” (F, R, 7, ESTJ); “Prefer generalized aims, because the instructor doesn’t know everyone’s abilities and physique” (M, T, 4, ISFJ). One participant said that generalized goals can protect their self-esteem: "Personalized goals disguised as one-to-one can be inferior or embarrassing. It’s embarrassing to see that others can but you can’t, so it’s better to have the same goal but give modifications at the same time” (F, T, 14, ESFJ).

On the other hand, one participant preferred personalized aims since they were more in line with individual needs: “There was a big difference in the physical ability of the group for this program, so it would have been boring if I had to do the movements of the less fit students” (F, R, 15, ESFJ).

• Goal preference for intervention/program: ultimate aim vs. progressive small aims

There were 14 participants who supported progressive small goals, whereas only 5 preferred ultimate goals, and 1 person chose both.

Progressive smaller goals will give participants a sense of success and satisfaction: "Smaller goals are more likely to have a sense of success and will be more intentional, large goals take a long time to see results. I am worried that something else is going on in the meantime that will affect the results, therefore less motivation and commitment” (F, R, 10, ISFJ). Some participants believe that small aims can make it easier for them to control their progress and are more suitable for their age: “For example, if I can’t achieve it today, I will try the next day and it will be easy to know if I have done it” (F, R, 15, ESFJ); “I can know the progress in each class and if go in the right direction” (F, R, 18); “It is better to take a gradual approach as you are getting older” (F, T, 5, ESFJ).

Other participants believe that ultimate aims are more flexible: “Because we may not master the movements she [the instructor] taught us in each class, I hope that we will be given a target in the end, because if we have a target in every class, we will be under pressure and we will be afraid of being compared to others. I would rather have a long-term goal so that we can go home and practice even if we don’t do well in that class, and a small goal like asking us to hand in our homework”(F, R, 19, ESFJ).

• Personality, goals, and healthy lifestyle behavior

In addition to filling out personality questionnaires, participants were also asked to describe their personalities. Those categorized as ESFJ, who comprised the majority, described themselves as positive, extroverted, agreeable, open, and conscientious. They like taking the initiative and trying new things, are open-minded about life, and enjoy team activities.

“I am open-minded, and always find activities and sports to participate in” (F, T, 9, ESFJ). “I like to interact and share with others, so I want to find activities and meet other people” (F, T, 12, ESFJ). “I learn spontaneously, and if it is a sport I am interested in, I will participate in it” (F, T, 14, ESFJ). “In the sports program, I was responsible for creating a WhatsApp group to keep my group members in communication” (F, T, 16, ESFJ).

Those categorized as ISFJ described themselves as introverted, curious, anxious, and impatient as well as open-minded to changing themselves to maintain a healthy lifestyle.

“I am a person who likes to get to the bottom of things. I want to find theories to prove, to accept different opinions and then synthesize” (M, R, 3, ISFJ). “After participating in the program, I made new friends, contact with people makes me happier, let the introverted self become extroverted” (M, T, 4, ISFJ). “Actively looking for activities/groups sports can make myself cheerful … avoid overthinking” (F, R, 11, ISFJ). “Originally, I am a person who is in a hurry, but by participating in Tai Chi, I can slow myself down” (F, T, 6, ISFJ).

Only two participants fell into the INFP and ESTJ categories. The participant assessed as INFP described herself as passive and overthinking, which hindered her from connecting with others. The participant assessed as ESTJ believed she was an active, positive, self-disciplined person, which has a positive effect on maintaining a healthy lifestyle.

5.3. Motivations and Intentions

• Theory-based program

The participants reported that being given regular health information was helpful, including the introductory definitions of physical and mental health, how to exercise, how to prevent injuries and reduce the risk of falling, how to eat healthily, etc.: “Giving participants advice on what level of physical or mental health they have, then explaining the means to them, and increasing their awareness of their own health” (F, T, 12, ESFJ); “Distributing exercise practices in the group, such as the movements, let us review, otherwise we will forget” (F, R, 15, ESFJ); “Holding health seminars on how to prevent injuries rather than only sport, provide us with knowledge such as fall prevention, injury triggers, diet and so on” (F, R, 18, ESFJ).

• Exercise encouragement

Participants reported that verbal encouragement, coupons, coach supervision, and social support inspire them to maintain exercise. “Verbal compliments with a few material rewards” (F, R, 11, ESFJ). “Having coaches who can correct movement mistakes and encourage them to keep doing it” (F, T, 14, ESFJ). “Group sports can encourage and help each other” (F, R, 18, ESFJ).

• Perceived constraints on exercise

Participants report that due to the COVID-19 pandemic, they had to stop all exercise programs. When the interviewer asked whether they used Zoom as a platform to participate in online exercise programs, some of them said that they did not because of privacy, limited space at home, or family issues. They also said that social interactions in face-to-face programs cannot be replaced by Zoom. Moreover, a few participants said that they could not adhere to exercise without a regular exercise program due to family, social matters, lack of discipline, and inertia.

“Because of the pandemic, my exercise classes were stopped, and I’m lazy so I don’t do any exercise anymore” (F, R, 2, ESFJ).

“Using Zoom is not good enough, because of the different housing environment” (M, T, 4, ISFJ). “Zoom class is actually fine, but it’s a little more scattered because at home you’re thinking about other things and it’s hard to concentrate” (F, R, 8, ISFJ). “Because Zoom classes can only be conducted at home, there are no sports facilities at my home, especially full body mirror, I cannot see whether my movements are correct, however in the face-to-face program, I can do exercises with familiar classmates, and we can rectify each other’s mistakes in performing exercises, which cannot replaceable by Zoom” (F, R, 20, INFP); she added, “I am afraid I am not able to use the recording function in Zoom”.

“I am not very disciplined. Without an exercise program my exercises become irregular, some time watching TV, some time talking on the phone, sometimes shopping for food and cooking” (F, R, 15, ESFJ).

Table 1 and Table 2 provide summaries of the retrieved literature and the categories enriched by the literature and interviews.

Summary of the retrieved literature.

Summary of categories and coding for literature synthesis and interviews.

6. Discussion

The current review summarizes the interaction between healthy lifestyle behaviors, goal setting, and personality, and the results are supported and extended by interviews with participants. The study results demonstrate that healthy lifestyle behaviors mainly consist of participating in physical activity, adopting healthy eating habits, and engaging in social activities. There is a tendency for people with different personality traits to have different attitudes toward engaging in HLB. In addition to progress in achieving goals, both the literature review and interviews provide evidence for the use of different types of goal setting plans, including personalized goals, short-term and long-term goals, generalized goals, and progressive goals.

Based on the review of the literature and interviews with participants, conscientiousness and neuroticism are the two personality traits that showed a significant interaction with personality, goals, and HLBs. Among those studies that focused on it, conscientiousness was shown to have a significant effect on the intention to protect one’s health, as well as the self-reported practice of health-protecting activities (e.g., exercise), according to the theory of planned behavior [ 61 ]. Furthermore, a study that used electronic devices in a physical activity intervention revealed that people with a high level of conscientiousness showed greater improvement in their daily steps, while people with health neuroticism and who have high levels of both neuroticism and conscientiousness had much higher daily steps [ 30 ]. It can be presumed that people with both neuroticism and conscientiousness have a strong fear of their possible future self, thus forcing them to develop greater self-regulation and self-control in order to prevent negative health-related outcomes, and it makes them more eager to strive for HLBs. However, the level of conscientiousness was shown to be changeable [ 62 , 63 ]. Nonetheless, studies documented that one’s level of conscientiousness increases with age [ 64 ]; hence, we acknowledge that most of the literature reviewed in the current study and the interviews with older adults tended to focus on the characteristic of conscientiousness. Studies have further indicated that a high level of conscientiousness is associated with more beneficial health behaviors and health-related outcomes [ 65 , 66 ]. Conversely, it is important to note that harmful behaviors such as drug and alcohol abuse do not seem to improve with conscientiousness over time or with age [ 67 , 68 ]. Therefore, this provides further evidence that conscientious people tend to take preventative actions to achieve health protection outcomes. Aside from age, social environment, and personal experience, psychological interventions such as behavioral cognitive therapy, mindfulness, and mental contrasting were found to be effective in increasing people’s level of conscientiousness by enhancing their commitment to goals and improving effective goal selection and goal striving, thus cultivating behavioral changes [ 63 ]. Based on this, interventions that include both physical activity and psychological elements should be promoted to cultivate conscientiousness and goal setting not only in older adults specifically but also in people in general, thus resulting in better HLB engagement.

7. Future Implications

It is shown that extroverts and conscientious people are more capable of achieving goals, adjusting their goals, and re-engaging with goals compared to people with neuroticism; however, the effect of personalities on the types of goals preferred by older adults has been scarcely discussed. While the reviewed literature pointed out that the use of an electronic device during exercise, such as a Fitbit, can cultivate personalized goal setting by using the device’s planning, regulation, and monitoring functions, the effect of personalized goals showed a greater effect among people who were conscientious and introverted [ 30 , 48 ]. This was also indicated in the participants’ interviews. The interviews indicated that introverts mostly avoid social gatherings or group activities because they would like the activity to be individualized. Attending group activities, including physical activity interventions, can cause worry and anxiety, such as worrying about not being as capable as others or not being able to meet the standard of the group. On the contrary, extroverts are more likely to attend group activities, because they like to share their experiences with others or within the group, thus they prefer non-personalized goals. In addition, no discernible personality differences were found between the preference for ultimate goals and progressive goals.

Based on the interview content, it is suggested that for physical activity or other HLB interventions in the future, the personality should be considered as a covariant and be controlled, or its possible effects on intervention outcomes should be examined during data analyses. Interventions that highlight individual differences could be offered, for instance, by recruiting participants with a particular personality type to improve their goal-setting skills and HLBs. Furthermore, different types of goals should be included in HLB inventions to investigate their effects on different personality traits when cultivating HLB. These suggestions are not exclusive to older adults but are also applicable to different populations and age groups. In addition, with an understanding of the importance of goal setting for healthy lifestyle behaviors, coaches and other health professionals can be given practical suggestions to enhance older adults’ self-understanding and literacy regarding healthy lifestyle behaviors (i.e., help them better understand their own physical capability, provide health information, or encourage them to set goals for themselves) before participating in exercise classes or undergoing medical treatments.

8. Strengths and Limitations of the Study

In the current study, we reviewed and summarized the interaction between goal setting, personality, and HLBs in the literature, further supported by qualitative interviews. Based on this review, it was found that conscientiousness, neuroticism, and extraversion had a significant effect on health-related behaviors as well as actual healthy lifestyle behaviors. Although personality traits such as agreeableness and openness were less likely to be examined or showed no significant outcomes, from the qualitative interviews, we are able to further conclude that older adults with all types of personalities, except neuroticism, tended to take the initiative to try new things, participate in physical activity interventions, and learn new knowledge. They were more open-minded toward unknown life and health-related knowledge, which facilitated their development and maintenance of a healthy lifestyle. On the other hand, older adults with neuroticism tended to overthink the unknown in the form of, for example, fear over their possible self or anxiety about the outcomes of new physical activities, thus holding them off from engaging in healthy behaviors.

Given that the literature review was not a systematic review and might have neglected the quality of retrieved studies, it also could not indicate and quantify the interaction effects between personalities, goal setting, and healthy lifestyle behaviors. Moreover, the interviewees were recruited by using convenience sampling from a previous physical activity intervention; hence, it could be assumed that they have already engaged in regular HLBs. This represents a limitation in that we were not able to distinguish potential differences between older adults who do not engage in regular HLBs and those who are not willing to engage in group-based physical activity interventions. Additionally, in the current study, we adopted the Myers–Briggs Type Indicator (MBTI) instead of the Big Five Inventory, because we considered the difficulty of distinguishing the participants’ personalities based on the latter. This indicates the interaction effect between two or more strong traits within an individual. The interaction effect between personality traits can make it difficult to identify the specific effect of personality traits on HLBs and goal setting. Therefore, we adopted the MBTI, with the understanding that the Big Five Inventory would be more suitable for quantitative analyses. Therefore, a meta-analysis of the interaction between personalities, goal setting, and physical activities or other specific HLBs is suggested. In addition, examining different types of HLB interventions that consider personality and goal setting strategies in future research is highly encouraged.

Funding Statement

This research received no external funding.

Author Contributions

Conceptualization, M.Y.C.W., K.-l.O. and P.K.C.; methodology, M.Y.C.W.; software, M.Y.C.W. and K.-l.O.; validation, M.Y.C.W. and K.-l.O.; formal analysis, M.Y.C.W. and K.-l.O.; investigation, M.Y.C.W. and K.-l.O.; resources, M.Y.C.W. and K.-l.O.; data curation, M.Y.C.W. and K.-l.O.; writing—original draft preparation, M.Y.C.W. and K.-l.O.; writing—review and editing, P.K.C.; visualization, M.Y.C.W. and K.-l.O.; supervision, P.K.C.; project administration, P.K.C.; funding acquisition, P.K.C. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Research Ethics Committee (REC) at the Hong Kong Baptist University (Ref: REC/18–19/0149).

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

Conflicts of interest.

The authors declare no conflict of interest.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.