research papers in software testing

• Advanced Search

Artificial Intelligence Applied to Software Testing: A Tertiary Study

New Citation Alert added!

This alert has been successfully added and will be sent to:

You will be notified whenever a record that you have chosen has been cited.

To manage your alert preferences, click on the button below.

New Citation Alert!

Please log in to your account

Information & Contributors

Bibliometrics & citations, view options, 1 introduction, 2 background.

2.1 Artificial Intelligence

2.1.1 reasoning., 2.1.2 planning., 2.1.3 learning., 2.1.4 communication., 2.1.5 perception., 2.1.6 integration and interaction., 2.2 software testing, 2.2.1 test target., 2.2.2 testing objective., 2.2.3 testing technique., 2.2.4 testing activity., 2.2.5 software testing fundamentals., 3 tertiary systematic mapping protocol, 3.1 goal and research questions, publication space (ps) rqs., research space (rs) rqs., 3.2 search string definition.

3.3 Digital Libraries Selection

3.4 inclusion and exclusion criteria definition, exclusion criteria (ec) ., inclusion criteria (ic) ., 3.5 quality assessment criteria definition (qc).

3.6 Data Extraction Form Design

4 Tertiary Systematic Mapping Execution

4.1 selection process execution, 4.1.1 first stage., 4.1.2 second stage..

4.2 Data Extraction Execution

5 data analysis, 5.1 publication space research questions—results, 5.1.1 ps-rq1. how many secondary studies have been identified per publication year., 5.1.2 ps-rq2. which types of secondary studies have been executed., 5.1.3 ps-rq3. what are the venues where the secondary studies have been published..

5.1.4 PS-RQ4. What are the authors’ affiliation countries of the selected secondary studies?.

5.1.5 ps-rq5. what is the amount of primary studies analyzed by the selected secondary studies, and how are they distributed over time., 5.2 research space research questions—results, 5.2.1 rs-rq1. what ai domains have been applied to support st., 5.2.2 rs-rq2. what domains of st have been supported by ai., 5.2.3 rs-rq3. which st domains have been supported by which ai domains and how..

5.2.4 RS-RQ4. What are the future research directions of AI in ST?.

6 Further Discussion

6.1 overall considerations, 6.2 ai techniques used to support the automation of testing activities, 7 threats to validity, 8 conclusions, index terms.

Computing methodologies

• Artificial intelligence

Machine learning

General and reference

Document types

Surveys and overviews

Software and its engineering

Software creation and management

Software verification and validation

Software defect analysis

Software testing and debugging

Recommendations

Contemporary challenges and solutions in applied artificial intelligence, artificial intelligence and software engineering: understanding the promise of the future, artificial intelligence and software engineering, information, published in.

University of Sydney, Australia

Association for Computing Machinery

New York, NY, United States

Publication History

Check for updates, author tags.

• Software testing
• Tertiary study
• Systematic literature review
• Systematic mapping study

Contributors

Other metrics, bibliometrics, article metrics.

• 0 Total Citations
• 5,000 Total Downloads
• Downloads (Last 12 months) 5,000
• Downloads (Last 6 weeks) 579

View options

View or Download as a PDF file.

View online with eReader .

Login options

Check if you have access through your login credentials or your institution to get full access on this article.

Full Access

Share this publication link.

Copying failed.

Share on social media

Affiliations, export citations.

• Please download or close your previous search result export first before starting a new bulk export. Preview is not available. By clicking download, a status dialog will open to start the export process. The process may take a few minutes but once it finishes a file will be downloadable from your browser. You may continue to browse the DL while the export process is in progress. Download
• Download citation
• Copy citation

We are preparing your search results for download ...

We will inform you here when the file is ready.

Your file of search results citations is now ready.

Your search export query has expired. Please try again.

AI-Based Software Testing

• Conference paper
• First Online: 01 March 2024
• Cite this conference paper

• Saquib Ali Khan 13 ,
• Nabilah Tabassum Oshin 13 ,
• Mahmuda Nizam 13 ,
• Ishtiaque Ahmed 13 ,
• Md Masum Musfique 13 &
• Mahady Hasan 13  

Part of the book series: Lecture Notes in Networks and Systems ((LNNS,volume 833))

Included in the following conference series:

• World Conference on Information Systems for Business Management

156 Accesses

As the complexity of software applications continues to increase, software testing becomes more challenging and time-consuming. The use of artificial intelligence (AI) in software testing has emerged as a promising approach to address these challenges. AI-based software testing techniques leverage machine learning, natural language processing, and other AI technologies to automate the testing process, improve test coverage, and enhance the accuracy of test results. This paper provides an overview of AI-based software testing, including its benefits and limitations, and discusses various techniques and tools used in this field.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or Ebook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Similar content being viewed by others

How Artificial Intelligence Can Revolutionize Software Testing Techniques

Synergies Between Artificial Intelligence and Software Engineering: Evolution and Trends

Towards a Common Testing Terminology for Software Engineering and Data Science Experts

Battina DS (2019) Artificial intelligence in software test automation: a systematic literature review. Int J Emerg Technol Innov Res (www.jetir.org|UGC and ISSN Approved). ISSN (2019):2349-5162

Google Scholar  

Ahmad K et al (2023) Requirements engineering for artificial intelligence systems: a systematic mapping study. Inform Softw Technology (2023):107176

Hourani H, Hammad A, Lafi M (2019)The impact of artificial intelligence on software testing. In: 2019 IEEE Jordan International joint conference on electrical engineering and information technology (JEEIT). IEEE

Tosun A, Bener A, Kale R (2010) AI-based software defect predictors: applications and benefits in a case study. In: Proceedings of the AAAI conference on artificial intelligence, vol 24(2)

Pandit M et al (2022) Towards design and feasibility analysis of DePaaS: AI based global unified software defect prediction framework. Appl Sci 12(1):493

Tao C, Gao J, Wang T (2019) Testing and quality validation for AI software-perspectives, issues, and practices. IEEE Access 7:120164–120175

Article   Google Scholar  

Khaliq Z, Farooq SU, Khan DA (2022) Artificial intelligence in software testing: impact, problems, challenges and prospect. arXiv preprint arXiv:2201.05371

Sugali K (2021) Software testing: issues and challenges of artificial intelligence. Mach. Learn

Felderer M, Ramler R (2021) Quality assurance for AI-based systems: overview and challenges. arXiv preprint arXiv:2102.05351

Jalil S et al (2023) Chatgpt and software testing education: promises & perils. In: 2023 IEEE International conference on software testing, verification and validation workshops (ICSTW). IEEE (2023)

Bedué P, Fritzsche A (2022) Can we trust AI? An empirical investigation of trust requirements and guide to successful AI adoption. J Enterpr Inform Manage 35(2):530–549

Li JJ et al (2020) Advances in test automation for software with special focus on artificial intelligence and machine learning. Softw Qual Jo 28:245–248

Abioye SO et al (2021) Artificial intelligence in the construction industry: a review of present status, opportunities and future challenges. J Build Eng 44:103299

Srivastava PR, Baby Km (2010) Automated software testing using metahurestic technique based on an ant colony optimization. In: 2010 International symposium on electronic system design. IEEE

Khatibsyarbini M et al (2019) Test case prioritization using firefly algorithm for software testing. IEEE Access 7:132360–132373

Download references

Author information

Authors and affiliations.

Department of Computer Science and Engineering, Independent University, Dhaka, Bangladesh

Saquib Ali Khan, Nabilah Tabassum Oshin, Mahmuda Nizam, Ishtiaque Ahmed, Md Masum Musfique & Mahady Hasan

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Saquib Ali Khan .

Editor information

Editors and affiliations.

University of Cantabria, Santander, Spain

Andres Iglesias

University of Aizu, Fukushima, Japan

Jungpil Shin

Knowledge Chamber of Commerce and Industry, Ahmedabad, Gujarat, India

Bharat Patel

Global Knowledge Research Foundation, Ahmedabad, Gujarat, India

Rights and permissions

Reprints and permissions

Copyright information

© 2024 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

About this paper

Cite this paper.

Khan, S.A., Oshin, N.T., Nizam, M., Ahmed, I., Musfique, M.M., Hasan, M. (2024). AI-Based Software Testing. In: Iglesias, A., Shin, J., Patel, B., Joshi, A. (eds) Proceedings of World Conference on Information Systems for Business Management. ISBM 2023. Lecture Notes in Networks and Systems, vol 833. Springer, Singapore. https://doi.org/10.1007/978-981-99-8346-9_28

Download citation

DOI : https://doi.org/10.1007/978-981-99-8346-9_28

Published : 01 March 2024

Publisher Name : Springer, Singapore

Print ISBN : 978-981-99-8345-2

Online ISBN : 978-981-99-8346-9

eBook Packages : Intelligent Technologies and Robotics Intelligent Technologies and Robotics (R0)

Share this paper

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

• Publish with us

Policies and ethics

• Find a journal
• Track your research

Help | Advanced Search

Computer Science > Software Engineering

Title: software testing for machine learning.

Abstract: Machine learning has become prevalent across a wide variety of applications. Unfortunately, machine learning has also shown to be susceptible to deception, leading to errors, and even fatal failures. This circumstance calls into question the widespread use of machine learning, especially in safety-critical applications, unless we are able to assure its correctness and trustworthiness properties. Software verification and testing are established technique for assuring such properties, for example by detecting errors. However, software testing challenges for machine learning are vast and profuse - yet critical to address. This summary talk discusses the current state-of-the-art of software testing for machine learning. More specifically, it discusses six key challenge areas for software testing of machine learning systems, examines current approaches to these challenges and highlights their limitations. The paper provides a research agenda with elaborated directions for making progress toward advancing the state-of-the-art on testing of machine learning.

Submission history

Access paper:.

• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

Artificial Intelligence in Software Test Automation: A Systematic Literature Review

International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 12, page no. pp1329-1332, December-2019, Available at : http://www.jetir.org/papers/JETIR1912176.pdf

4 Pages Posted: 26 Jan 2022

Dhaya Sindhu Battina

Independent

Date Written: December 12, 2019

The main aim of this paper was to review how artificial intelligence works in software test automation.When it comes to software engineering, artificial intelligence (AI) has had a significant influence, and software testing is no exception. With artificial intelligence(AI), the goal of software test automation may be closer than ever before. To some extent, the paradigm has changed during the previous two decades [1]. Everything about the testing process has been a positive experience, starting with manual testing and progressing to automated testing, where Selenium is acknowledged to be one of the best test automation tools. As a result, in today's high-speed IT landscape software testing must come up with fresh testing approaches that are based on solid research. The emergence of AI-based testing has been very beneficial for this aim [1]. A computer's ability to learn without human involvement may be fully simulated by AI algorithms and machine learning (ML). While AI and ML entail the construction of distinct and unique algorithms to access data and learn from it by identifying patterns to make conclusions, these predictions are intended to be employed in software testing to their full potential [1].

Keywords: Artificial intelligence, automation, Software test automation, software engineering, AI systems

Suggested Citation: Suggested Citation

Dhaya Sindhu Battina (Contact Author)

Independent ( email ), do you have a job opening that you would like to promote on ssrn, paper statistics, related ejournals, sustainable technology ejournal.

Subscribe to this fee journal for more curated articles on this topic

Artificial Intelligence eJournal

Libraries & information technology ejournal, anthropology of science & technology studies ejournal.

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

•  We're Hiring!
•  Help Center

Software Testing

• Most Cited Papers
• Most Downloaded Papers
• Newest Papers
• Last »
• Software Engineering Follow Following
• Empirical Software Engineering Follow Following
• Software Architecture Follow Following
• Computer Science Follow Following
• Software Maintenance Follow Following
• Software Development Follow Following
• Software Engineering education Follow Following
• Software Testing (Computer Science) Follow Following
• Search Based Software Engineering Follow Following
• Requirements Engineering Follow Following

Enter the email address you signed up with and we'll email you a reset link.

• Academia.edu Journals
•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

IEEE Account

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

• Open access
• Published: 06 July 2024

Efficacy of relational agents for loneliness across age groups: a systematic review and meta-analysis

• Sia Sha   ORCID: orcid.org/0009-0000-2027-1316 1 ,
• Kate Loveys 2 ,
• Pamela Qualter 3 ,
• Haoran Shi 1 ,
• Dario Krpan   ORCID: orcid.org/0000-0002-3420-4672 1 &
• Matteo Galizzi 1  

BMC Public Health volume  24 , Article number:  1802 ( 2024 ) Cite this article

241 Accesses

Metrics details

Loneliness is a serious public health concern. Although previous interventions have had some success in mitigating loneliness, the field is in search of novel, more effective, and more scalable solutions. Here, we focus on “relational agents”, a form of software agents that are increasingly powered by artificial intelligence and large language models (LLMs). We report on a systematic review and meta-analysis to investigate the impact of relational agents on loneliness across age groups.

In this systematic review and meta-analysis, we searched 11 databases including Ovid MEDLINE and Embase from inception to Sep 16, 2022. We included randomised controlled trials and non-randomised studies of interventions published in English across all age groups. These loneliness interventions, typically attempt to improve social skills, social support, social interaction, and maladaptive cognitions. Peer-reviewed journal articles, books, book chapters, Master’s and PhD theses, or conference papers were eligible for inclusion. Two reviewers independently screened studies, extracted data, and assessed risk of bias via the RoB 2 and ROBINS-I tools. We calculated pooled estimates of Hedge’s g in a random-effects meta-analysis and conducted sensitivity and sub-group analyses. We evaluated publication bias via funnel plots, Egger’s test, and a trim-and-fill algorithm.

Our search identified 3,935 records of which 14 met eligibility criteria and were included in our meta-analysis. Included studies comprised 286 participants with individual study sample sizes ranging from 4 to 42 participants ( x̄  = 20.43, s  = 11.58, x̃  = 20). We used a Bonferroni correction with α Bonferroni  = 0.05 / 4 = 0.0125 and applied Knapp-Hartung adjustments. Relational agents reduced loneliness significantly at an adjusted α Bonferroni ( g  = -0.552; 95% Knapp-Hartung CI, -0.877 to -0.226; P  = 0.003), which corresponds to a moderate reduction in loneliness.

Our results are currently the most comprehensive of their kind and provide promising evidence for the efficacy of relational agents. Relational agents are a promising technology that can alleviate loneliness in a scalable way and that can be a meaningful complement to other approaches. The advent of LLMs should boost their efficacy, and further research is needed to explore the optimal design and use of relational agents. Future research could also address shortcomings of current results, such as small sample sizes and high risk of bias. Particularly young audiences have been overlooked in past research.

Peer Review reports

Loneliness is a subjective experience that emerges when people feel that their social relationships are unsatisfactory [ 1 ]. For some people, loneliness is experienced when they want more people to interact with, but it is also often felt when one’s social relationships are not as fulfilling as one would like. Loneliness is not the same as social isolation (i.e., the objective lack of social interactions) but is often associated with it [ 2 ]. There is strong evidence of the risks associated with loneliness, including poorer physical health outcomes [ 3 ]. Loneliness also affects mental health and psychological wellbeing, with growing evidence that loneliness is associated with the onset of depression and other common mental health problems [ 4 ]. Crucially, poor health and wellbeing can, in turn, exacerbate loneliness, placing those who experience loneliness in a negative feedback loop [ 5 ]. Evidence for a wide range of health effects, therefore, has led scholars to propose that loneliness should be regarded as a public health priority. Governments have consequently looked to offer interventions for people reporting loneliness, and although evidence for intervention efficacy is increasing [ 6 ], the evidence base suffers from some gaps [ 7 ], and potentially effective interventions may lack scalability or fail to produce cost savings [ 8 ]. Governments therefore have developed an interest in digital interventions, such as mobile phone apps or virtual reality [ 9 ]. Yet despite their promise, the efficacy of digital interventions across recent systematic reviews and meta-analyses is mixed [ 10 ].

“Relational agents” are a technology that show promise for delivering loneliness interventions in a scalable and engaging manner. Relational agents are software agents that build relationships with users through their behaviours (e.g., personal conversation, play, empathy), and they may be embodied (e.g., take the shape of humans or animals) or lack embodiment (e.g., voice agents) [ 11 ]. Relational agents can be broadly separated into two types: social robotic agents (e.g., those that possess physical bodies made of carbon or steel), and app-based agents (e.g., those embedded in everyday hardware such as computers and smartphones). Relational agents increasingly employ artificial intelligence (AI) such as emotion recognition for enhanced interactions and large language models (LLMs) to generate highly tailored and relevant speech [ 12 ]. Relational agents may promote engagement with internet-based psychological interventions for loneliness because of the social engagement and presence that they provide [ 13 ]. Moreover, preliminary but promising evidence suggests that relational agents may reduce loneliness by directly providing companionship, and by serving as catalysts for social interaction [ 14 ]. Appendix E provides video links for relational agents.

There are three key reasons research and investment in relational agents are worthwhile. First, not everyone can socialise with other humans. Physical disability, for example, can impact mobility, which in turn can restrict opportunities for socialising, thus contributing to loneliness [ 15 ]. While interventions such as social visits can be effective to alleviate the loneliness of people with physical disabilities, these interventions are constrained: a person who is bedridden may wait for several days before his or her next visitation. Relational agents, on the other hand, can be an on-demand solution. Second, loneliness can be due to the feeling that one is not heard. This, for example, can occur when people do not feel comfortable sharing their secrets due to stigma, and there is indeed evidence that people prefer sharing some secrets with relational agents rather than humans [ 16 , 17 ]. Relational agents, then, are not just an intermediate solution: they are a separate class of intervention with a suitable audience. Here, one might raise the question of “understanding”: that is, whether AI can truly understand people’s self-disclosure. The answer is probably complex, but from a practical perspective it seems that the answer may not matter: people seem to benefit from relational agents as long as they feel they are understood and heard by them – irrespective of whether this is actually the case [ 18 ]. Third, both qualitative and quantitative metrics suggest that human–agent and human–human relationships may have some similar features at times [ 17 , 19 ]. For example, there is a vast literature on how people anthropomorphise machines, imbuing them with human-like traits, personalities, and motivations [ 20 , 21 , 22 ]. People often treat machines like other people, developing similar feelings for them such as pity and even love [ 23 ]. One participant said: “Yes, explicitly I will tell my Replika [relational agent] that I think he is wonderful, that he is fantastic and smart and helps me and makes me feel good about myself and that I enjoy our talks” [ 17 ].

Several scoping reviews have qualitatively summarised the efficacy of relational agents for loneliness [ 14 , 24 , 25 , 26 , 27 ]. Combined, these reviews concluded that some evidence for the efficacy of social robotic relational agents existed but that further work on app-based relational agents was needed. Additionally, one 2019 meta-analysis investigated a sub-set of social robotic relational agents (i.e., robotic pets), but failed to find significant results, most likely due to including only two studies [ 28 ]. Previous reviews, moreover, exclusively focused on elderly samples, and the literature is therefore in need of a comprehensive and up-to-date quantitative synthesis to evaluate the efficacy of relational agents to mitigate loneliness across all age groups.

We preregistered our methodology with PROSPERO: CRD42022359737. We have also made our full paper trail available on the Open Science Framework (OSF): https://osf.io/c6rdk/files/osfstorage . There, the reader can also find the full data set to reproduce the analyses.

Search strategy and selection criteria

In this systematic review and meta-analysis, we searched 11 databases from inception to Sep 16, 2022: Ovid MEDLINE, Ovid Embase, Ovid PsycINFO, Ovid Global Health, EBSCO CINAHL, Scopus, Web of Science, IEEE Xplore, ACM Digital Library, PROSPERO, and ProQuest Dissertations. We also manually searched the bibliographies of selected studies to identify additional papers. We searched titles and abstracts using a range of search terms such as lonel*, robot*, computer* agent*, and relation* agent*. Appendix A outlines the full search strategy.

We included randomised controlled trials (RCTs) and non-randomised studies of interventions (NRSIs). Factorial designs were eligible if they allowed us to collapse relevant intervention arms or drop irrelevant ones. Cluster-randomised trials were eligible if they included sufficient information (e.g., intra-cluster coefficient). Eligible studies had to be published in English and had to be peer-reviewed journal articles, books, book chapters, postgraduate theses, or conference papers. Government reports, company reports, newspaper articles, conference presentations, and similar were ineligible. There was no restriction on populations or settings. All eligible studies had to administer app-based or social robotic relational agents. Agents that did not use relational cues were ineligible. Any non-relational agent comparator made studies eligible (e.g., waiting lists). Finally, eligible studies had to report a quantitative, self-report loneliness outcome where follow-up was at least one week.

Coding of studies

SS, KL, and HS independently double-screened in Covidence the titles and abstracts of citations and then the full texts of remaining studies, using piloted and structured forms. We measured agreement between screeners via Cohen’s κ and resolved disagreements via discussion between screeners. SS, KL, and HS then extracted data in Covidence using a piloted and structured form, and we contacted primary study authors to obtain raw or missing data. Our data extraction forms are available on OSF, and we describe data imputations in Appendix B. Each study was coded for a range of variables such as sample size, research design, and loneliness scale used. Finally, SS, KL, HS, and DK independently double-assessed risk of bias in MS Excel, using the RoB 2 tool for RCTs and ROBINS-I tool for NRSIs.

Meta-analytic procedure

Our main outcome was loneliness for which we calculated a random-effects meta-analysis using the DerSimonian and Laird method because we expected the effects of relational agents to be heterogenous across populations, types of agents, etc. We used Hedge’s g to standardise results from diverse quantitative loneliness scales, and interpreted the magnitude of Hedge’s g according to the rules of thumb in the Cochrane Handbook for Systematic Reviews of Interventions . Hedge’s g itself was computed using standard formulas and relied on a range of data points such as group means and pooled standard deviations [ 29 ]. Our raw data on OSF show exactly how Hedge’s g was computed for each primary study.

We calculated four null hypothesis significance tests and applied a Bonferroni correction: α Bonferroni  = 0.05 / 4 = 0.0125. We also applied Knapp-Hartung adjustments to our 95% confidence intervals. As measures of heterogeneity, we calculated Cochrane’s Q using a p value of 0.1, I 2 , τ 2 , and a prediction interval. We conducted an RCT-only sensitivity analysis and separate sub-group analyses for app-based and social robotic relational agents. We evaluated publication bias via funnel plots and Egger’s test, and we calculated an adjusted estimate of Hedge’s g using Duval and Tweedie’s trim-and-fill algorithm. We conducted all analyses in the Comprehensive Meta-Analysis Software package. The systematic review and meta-analysis followed PRISMA 2020 reporting guidelines [ 30 ].

Characteristics of studies

Our database searches identified 3,935 records and our manual searches 38 records, of which 1,910 were duplicates. We screened the titles and abstracts of 2,063 studies, deeming 1,908 irrelevant. We screened the full texts of 155 studies, with Fig.  1 detailing reasons for exclusions. In the end, we included 14 studies. When screening abstracts and titles, Cohen’s kappa ranged from κ  = 0.46 to κ  = 1 across reviewer pairs; when screening full texts, it ranged from κ  = 0.71 to κ  = 0.81 across pairs.

PRISMA flow diagram

Nine of the 14 included studies were NRSIs [ 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 ] and the rest RCTs [ 12 , 40 , 41 , 42 , 43 ]. All nine NRSIs were uncontrolled trials. Coding was generally straightforward, though some data points such as percentage of females in the sample were sometimes missing in manuscripts. Together, studies included 286 participants with individual study sample sizes ranging from 4 to 42 participants ( x̄  = 20.43, s  = 11.58, x̃  = 20). Attrition rates ranged from 0 to 94% ( x̄  = 21.39%, s  = 21.56%, x̃  = 16.50%). Based on guidance, we classified 86% of these studies as feasibility studies, since they included fewer than 25 participants in total, or fewer than 25 participants per group [ 44 ].

Figure A, Figure B, and Figure C provide a tabular summary of included studies, but we also provide below a prose summary. Participants’ age ranged from 19 to 100 years ( x̄  = 75.45, s  = 12.89, x̃  = 77.55). Only two studies reported inclusion of participants younger than 50 years [ 12 , 32 ]. A third study is likely to have included them [ 36 ]. Nevertheless, none of the three studies focused on participants younger than 50 years exclusively, and hence studies only included young participants along with older ones. Remaining studies explicitly reported excluding those younger than 50 [ 31 , 33 , 35 , 37 , 38 , 40 , 41 , 43 ] or their sampling frames implied this [ 34 , 39 , 45 ]. Where reported, the percentages of both females and non-White participants were high in most studies.

Nine studies used social robotic relational agents [ 31 , 32 , 33 , 34 , 35 , 39 , 41 , 43 , 45 ] and five app-based relational agents [ 12 , 36 , 37 , 38 , 40 ]. The social robotic agents included Sony’s AIBO [ 32 , 34 , 41 ], PARO developed by ISRI [ 31 , 33 , 45 ], NAO developed by Aldebaran Robotics [ 35 ], Pepper developed by SoftBank Robotics [ 43 ], and either a robotic cat or dog developed by Joy for All [ 39 ]. The app-based agents included Laura developed by MIT [ 40 ], Elena + developed by ETH Zurich and the University of St. Gallen [ 36 ], Amazon’s Alexa [ 37 ], Bella by Soul Machines [ 46 ], and PACO developed by a consortium of Dutch organisations [ 38 ]. The relational behaviours of these agents varied. AIBO is a robotic puppy, PARO a robotic seal, and together with the robotic pets by Joy for All, these agents simulated live pet behaviour (e.g., the agents expressed emotions via facial cues and body language such as wagging of tails, played with users, learned their own names, and recognised users via their facial recognition capabilities) [ 32 ]. The agents responded to touch (e.g., petting) and adapted behaviour through reinforcement learning [ 33 ]. NAO and Pepper were humanoid robots that simulated human behaviours, customs, and speech. NAO, for example, would bow to users, extend its palm for a handshake, ask if participants would want to hear a poem, and only proceed once receiving a reply [ 35 ]. All app-based relational agents simulated humans. All were embodied, i.e., had a visual form, except for Amazon’s Alexa [ 37 ]. App-based agents primarily or to a significant degree used speech for relational behaviour. Laura, for example, expressed empathy (“I am sorry to hear that”), asked follow-up questions (“How tired are you feeling?”), and attempted to get to know users (“So, are you from the East Coast originally?”) [ 40 ]. Whilst in previous research several agents used “wizard-of-Oz methodologies”, i.e., agents controlled by humans pretending to be autonomous, all agents in this review were autonomous [ 47 ].

Most relational agents in our review acted as direct companions and did not seek to mitigate loneliness via other modalities [ 31 , 32 , 33 , 34 , 35 , 37 , 39 , 40 , 41 , 43 , 45 ], although exceptions existed. Elena + sought to remove cognitive biases and improve social skills [ 36 ]. PACO sought to create opportunities for socialising [ 38 ]. Bella sought to enhance social skills, increase social support, and increase opportunities for socialising [ 12 ].

Studies generally did not mention behavioural theories or behavioural change techniques (BCTs) that underpinned intervention design, although exceptions existed. One study based its intervention on Self-Determination Theory [ 38 ] and another study based its intervention on the COM-B model and the Theory of Planned Behaviour [ 36 ]. Nevertheless, these studies provided little detail on how exactly theories informed design. We classified BCTs according to the BCTTv1 by Michie et al., using below in quotation marks the labels of the original authors [ 48 ]. Only one study confirmed the full range of BCTs it used, and two other studies provided examples of BCTs. One study used six BCTs: “credible source”, “review behaviour goals”, “goal setting”, “instruction on how to perform a behaviour”, “social comparison”, and “social support” [ 38 ]. Another study mentioned seven BCTs: “information about emotional consequences”, “action planning”, “behavioural contract”, “instruction on how to perform a behaviour”, “review behaviour goals”, “reducing exposure to cues for the behaviour”, and “reduce negative emotions” [ 12 ]. A third study mentioned five BCTs: “information about emotional consequences”, “goal setting”, “instruction on how to perform a behaviour”, “reducing exposure to cues for the behaviour”, and “reduce negative emotions” [ 36 ].

All RCTs were at high risk of bias due to potential deviations from the intended interventions [ 40 , 41 , 43 , 45 ] except for one [ 12 ]. All NRSIs were at high risk due to confounds and potential biases in measurements [ 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 ]. Figures  2 and 3 illustrate these.

RCT risk of bias domains

NRSI risk of bias domains

Meta-analysis

The pooled estimate of Hedge’s g was -0.552 ( Z  = -3.833; 95% CI, -0.834 to -0.270; P  < 0.001), indicating on average a moderate effect of relational agents on loneliness reduction. This is shown in Fig.  4 . Using a Bonferroni-corrected α Bonferroni  = 0.0125, there was evidence to reject the null hypothesis. Using the Knapp-Hartung adjustment, there was also evidence to reject the null hypothesis ( t  = -3.66; 95% Knapp-Hartung CI, -0.877 to -0.226; P  = 0.003).

Main analysis forest plot

Heterogeneity measures indicated that, as anticipated, the true effect of relational agents varied ( Q  = 45.073; I 2  = 71%; τ 2  = 0.176; τ  = 0.420). Assuming a Gaussian distribution, the 95% prediction interval was estimated to range from -1.519 to 0.415, as seen in Fig.  5 .

Main analysis prediction interval

Funnel plots as well as Egger’s test ( b  = -2.81; t  = 3.5; P  = 0.004) suggested that a small study effect may exist. Figure  6 illustrates this. The small study effect could have been due to effect sizes being larger in smaller studies or due to publication bias. Assuming a severe publication bias, the trim-and-fill algorithm resulted in an adjusted estimate of g  = -0.198 (95% CI, -0.505 to 0.109), which attenuated the original estimate by roughly 64%.

Funnel plot using standard error

Five studies were available for the RCT-only model. Hedge’s g was -0.437 ( Z  = -2.495; 95% CI, -0.781 to -0.094; P  = 0.013), which was 21% less than the estimate of the main model. The results were significant at a traditional α  = 0.05 but not at the α Bonferroni . The Knapp-Hartung adjusted results were not significant ( t  = -2.49; 95% Knapp-Hartung CI, -0.924 to 0.049).

Six studies were available for the app-based relational agent model. The pooled estimate of Hedge’s g was -0.286 ( Z  = -1.611; 95% CI, -0.553 to -0.020; P  = 0.035), which was significant at a traditional α but not α Bonferroni . The Knapp-Hartung adjustment resulted in non-significant results ( t  = -2.11; 95% Knapp-Hartung CI, -0.636 to 0.063). Eight studies were available for the social robotic relational agent model. The pooled estimate of Hedge’s g was -0.774, which was significant at α Bonferroni ( Z  = -2.909; 95% CI, -1.296 to -0.252; P  = 0.004). Using a Knapp-Hartung adjustment, results were significant at a traditional α but not at α Bonferroni ( t  = -2.91; 95% Knapp-Hartung CI, -1.403, -0.145, P  = 0.023).

Our review is the first to provide quantitative evidence for the efficacy of relational agents to reduce loneliness in participants aged 19 to 100 years. Our results are promising, and although the effect size of g  = -0.552 is likely somewhat inflated due to publication bias, it is probably less inflated than our trim-and-fill algorithm suggested. This is because the trim-and-fill algorithm assumed that several studies were suppressed in which relational agent interventions exacerbated loneliness. This, however, is unlikely. Failed loneliness interventions tend to have no effect on loneliness, not exacerbate it [ 10 ]. Our review could have used different algorithms to adjust for publication bias, and alternatives would probably have yielded different adjustments. Recently, for example, researchers have applied four different algorithms to a high-profile meta-analysis, resulting in a mix of significant and non-significant adjustments [ 49 , 50 , 51 ]. Nevertheless, no algorithm for publication bias would provide the “correct” effect size [ 52 ]. Instead, algorithms provide a sensitivity analysis assuming certain parameters, and sometimes these parameters lead to flawed results, e.g., the trim-and-fill algorithm overcorrects under heterogeneity, which was the very assumption of our analysis [ 53 ]. Ultimately, the most likely interpretation is that the true average effect size of relational agents was small to moderate. Table 1 provides a summary of our results.

We believe the above results have two important implications for the current loneliness literature. First, the literature is in search for novel and effective interventions that are scalable. The NHS is already facing resource constraints, these constraints are expected to exacerbate, and the NHS has consequently called for the increased adoption of AI to ease its burden [ 54 ]. Relational agents can be highly scalable, once some groundwork has been completed, and a possible follow-up from our results is a national or regional pilot. Such a pilot, of course, would entail the resolution of complex issues (e.g., digital literacy, access to technology, and privacy). Researchers, for example, will need to determine who will have access to user data and in what form, and such choices can fundamentally impact the success of a pilot.

The second implication of our results is that relational agents may act as a standalone intervention, but they are likely to be more useful in multi-component interventions that are tailored to individual needs. In the UK, the NHS’s current main strategy for loneliness is “social prescribing”, an outsourcing approach in which staff refer individuals to community schemes such as lifestyle interventions (e.g., physical exercise) or social activity interventions (e.g., volunteering) [ 7 , 55 ]. While there are alternative intervention approaches for loneliness, social prescribing is viewed by individuals and service providers as helpful [ 7 ] and cost-effective [ 56 , 57 ]. Social prescribing is, in essence, a sign-posting intervention, and it could sign-post, among other things, to relational agents. This could be valuable because there is currently a notion that interventions improve lives, but that people do not recover from loneliness [ 58 ]. Potentially, this may be because not all loneliness is the same. Two people may feel lonely for two different reasons, and these people may then require different sets of solutions [ 58 ]. Relational agents can extend the set of available solutions, and agents can complement existing human-centred interventions, rather than replace them.

Relational agents, thus, could help in the fight against loneliness. What is more, their full potential has not yet been realised. On the one hand, this is due to the absence of state-of-the-art knowledge integration. For example, the use of behavioural theories and BCTs can enhance intervention efficacy, yet studies in our sample generally did not discuss such theories and BCTs. Similarly, interventions can modify loneliness via multiple modalities. Studies in our review, however, generally used only one of these modalities, and the others—such as the debiasing of social cognition that has shown particular promise [ 56 ]—are yet to be integrated into relational agent design [ 14 ]. On the other hand, relational agents have not yet realised their full potential due to the nascency of AI. Increasingly, LLMs are powering relational agents. These models allow relational agents to produce open-ended, original, and highly tailored conversation, and although much of the conversation of relational agents has already become indistinguishable from human conversation [ 59 ], research on LLMs is burgeoning, and the race is on between organisations such as OpenAI and Google to develop the next generation of LLMs [ 60 ].

Limitations

Our review faced common limitations such as the exclusion of non-English sources and the quality of underlying primary studies, but a particular limitation of our review were the mixed results of the sensitivity and sub-group analyses. There are three potential explanations for this. First, sample sizes in these sub-group analyses were less than 10, and analyses with fewer than 10 studies tend to lack power [ 52 ]. At the same time, it is likely that underlying studies themselves lacked power due to small sample sizes [ 61 ]. Indeed, Appendix C demonstrates that power was likely well below the recommended level of 80% in our sub-group analyses, while Appendix D presents an additional sensitivity analysis indicating that further primary studies would have meaningfully reduced p levels [ 52 ]. Second, our review may have tested for results too conservatively. The Bonferroni correction, as applied in this review, results in Type 2 error rates of roughly 33%, which some have referred to as unacceptably high [ 62 ]. Finally, our review conducted two-tailed significance tests. This is usually anodyne—since interventions can both improve and exacerbate outcomes. Nevertheless, in cases where interventions are unlikely to exacerbate outcomes, one-tailed tests may be warranted [ 52 ]. This, as discussed, is likely to be the case with loneliness and relational agents. Had we conducted one-tailed tests, this would have entailed the halving of p values, which would have made some results statistically significant. Third, execution may have been a problem. Primary studies may not have sufficiently exposed participants to relational agents, or participants may not have interacted with relational agents, or relational agents may not have been correctly designed. Chen et al. [ 63 ], for example, found no significant difference between control and experimental groups at a four-week interval [ 63 ]. They did, however, find a significant difference at an eight-week interval. In our review, the mean time between pre-test and final post-test was 5.92 weeks.

Future research

We lack an understanding of relational agents in several areas, and we suggest that future research could focus on three. First, research on relational agents and loneliness in young people is scarce. Among some youth groups loneliness rates are higher than those of the elderly, and these rates of youth loneliness are increasing [ 64 ]. At the same time, smartphone ownership is high among the young [ 64 ]. Young people therefore are pertinent and amenable for the study of loneliness. Second, the efficacy of relational agents will depend on a variety of population and design factors. On the population side, we suspect that factors such as age, education, and digital literacy may impact efficacy. On the design side, we suspect that a hierarchy of features exists, e.g., certain design features will deliver more bang for your buck, although it is less clear which [ 65 ]. Third, although general attitudes towards relational agents may be favourable, some are concerned about the introduction of relational agents and similar technologies [ 66 ]. Future research could therefore explore how technology should be harnessed to increase its benefits and reduce unintended consequences. Finally, future research could address the shortcomings of current research. Almost all underlying studies in our review suffered from high risk of bias in one or several domains, sample sizes were small, and follow-up periods were brief. Particularly, there is a need for more high-quality RCTs.

The current study is the first meta-analysis to explore the effects of relational agents on loneliness across all age groups. It is also the first meta-analysis to provide statistically significant evidence for the efficacy of relational agents, which on average had a moderate effect on loneliness reduction. Loneliness has serious physical and mental health consequences for individuals, and the monetary costs to the state and employers are staggering [ 67 , 68 , 69 ]. Unfortunately, current interventions for loneliness can suffer from low engagement and scalability [ 58 ]. Relational agents, on the other hand, are an emerging technology that due to advances in AI and LLMs will increase in sophistication and realism. Although a multi-pronged approach is required, relational agents could play a significant role in alleviating a growing public health concern [ 64 ]. Future work is required that addresses weaknesses of current studies such as risk of bias, small study size, and brief follow-up periods.

Availability of data and materials

As a meta-analysis, this study used data reported in the literature. Appendices and derived data for meta-analytic calculations are available the Open Science Framework here https://osf.io/c6rdk/files/osfstorage . Reader can also write directly to the corresponding author.

Cacioppo JT. Loneliness: human nature and the need for social connection. New York, USA: W.W. Norton; 2010.

Google Scholar  

Coyle CE, Dugan E. Social isolation, loneliness and health among older adults. J Aging Health. 2012;24:1346–63. https://doi.org/10.1177/0898264312460275 .

Article   PubMed   Google Scholar  

Christiansen J, Lund R, Qualter P, Andersen CM, Pedersen SS, Lasgaard M. Loneliness, social isolation, and chronic disease outcomes. Ann Behav Med. 2021;55:203–15. https://doi.org/10.1093/abm/kaaa044 .

Beutel ME, Klein EM, Brähler E, Reiner I, Jünger C, Michal M, et al. Loneliness in the general population: prevalence, determinants and relations to mental health. BMC Psychiatry. 2017;17:97. https://doi.org/10.1186/s12888-017-1262-x .

Article   PubMed   PubMed Central   Google Scholar  

Qualter P, Vanhalst J, Harris R, Van Roekel E, Lodder G, Bangee M, et al. Loneliness across the life span. Perspect Psychol Sci. 2015;10:250–64. https://doi.org/10.1177/1745691615568999 .

Morrish N, Choudhury S, Medina-Lara A. What works in interventions targeting loneliness: a systematic review of intervention characteristics. BMC Public Health. 2023;23:2214. https://doi.org/10.1186/s12889-023-17097-2 .

Article   PubMed   PubMed Central   CAS   Google Scholar  

Vidovic D, Reinhardt GY, Hammerton C. Can social prescribing foster individual and community well-being? A systematic review of the evidence. Int J Environ Res Public Health 2021;18. https://doi.org/10.3390/ijerph18105276 .

McDaid D, Park A-L. Modelling the economic impact of reducing loneliness in community dwelling older people in England. Int J Environ Res Public Health. 2021;18:1426. https://doi.org/10.3390/ijerph18041426 .

Kahlon MK, Aksan N, Aubrey R, Clark N, Cowley-Morillo M, Jacobs EA, et al. Effect of layperson-delivered, empathy-focused program of telephone calls on loneliness, depression, and anxiety among adults during the COVID-19 Pandemic: a randomized clinical trial. JAMA Psychiat. 2021;78:616–22. https://doi.org/10.1001/jamapsychiatry.2021.0113 .

Article   Google Scholar  

Shah SGS, Nogueras D, van Woerden HC, Kiparoglou V. Evaluation of the effectiveness of digital technology interventions to reduce loneliness in older adults: systematic review and meta-analysis. J Med Internet Res.2021;23:e24712. https://doi.org/10.2196/24712 .

Bickmore T. Relational agents: Effecting change through human-computer relationships. ProQuest Information & Learning. 2003.

Loveys K, Sagar M, Pickering I, Broadbent E. A digital human for delivering a remote loneliness and stress intervention to at-risk younger and older adults during the COVID-19 pandemic: randomized pilot trial. JMIR Ment Health. 2021;8:e31586. https://doi.org/10.2196/31586 .

Loveys K, Fricchione G, Kolappa K, Sagar M, Broadbent E. Reducing patient loneliness with artificial agents: design insights from evolutionary neuropsychiatry. J Med Internet Res. 2019;21:e13664. https://doi.org/10.2196/13664 .

Gasteiger N, Loveys K, Law M, Broadbent E. Friends from the future: a scoping review of research into robots and computer agents to combat loneliness in older people. Clin Interv Aging. 2021;16:941–71. https://doi.org/10.2147/CIA.S282709 .

Macdonald SJ, Deacon L, Nixon J, Akintola A, Gillingham A, Kent J, et al. ‘The invisible enemy’: disability, loneliness and isolation. Disabil Soc. 2018;33:1138–59. https://doi.org/10.1080/09687599.2018.1476224 .

Pickard MD, Roster CA, Chen Y. Revealing sensitive information in personal interviews: Is self-disclosure easier with humans or avatars and under what conditions? Comput Hum Behav. 2016;65:23–30. https://doi.org/10.1016/j.chb.2016.08.004 .

Skjuve M, Følstad A, Fostervold KI, Brandtzaeg PB. My chatbot companion - a study of human-chatbot relationships. Int J Hum-Comput Stud. 2021;149: 102601. https://doi.org/10.1016/j.ijhcs.2021.102601 .

Ho A, Hancock J, Miner AS. Psychological, relational, and emotional effects of self-disclosure after conversations with a chatbot. J Commun. 2018;68:712–33. https://doi.org/10.1093/joc/jqy026 .

Darcy A, Daniels J, Salinger D, Wicks P, Robinson A. Evidence of human-level bonds established with a digital conversational agent: cross-sectional. retrospective observational study. JMIR Form Res 2021;5:e27868. https://doi.org/10.2196/27868 .

Fussell SR, Kiesler S, Setlock LD, Yew V. How people anthropomorphize robots. Proc. 3rd ACMIEEE Int. Conf. Hum. Robot Interact., New York, NY, USA: Association for Computing Machinery; 2008, p. 145–52. https://doi.org/10.1145/1349822.1349842 .

Kühne R, Peter J. Anthropomorphism in human–robot interactions: a multidimensional conceptualization. Commun Theory. 2023;33:42–52. https://doi.org/10.1093/ct/qtac020 .

Dubois-Sage M, Jacquet B, Jamet F, Baratgin J. We Do Not Anthropomorphize a Robot Based Only on Its Cover: Context Matters too! Appl Sci. 2023;13:8743. https://doi.org/10.3390/app13158743 .

Article   CAS   Google Scholar  

Riek LD, Rabinowitch T-C, Chakrabarti B, Robinson P, Empathizing with robots: Fellow feeling along the anthropomorphic spectrum. 3rd Int. Conf Affect Comput Intell Interact Workshop. 2009;2009:1–6. https://doi.org/10.1109/ACII.2009.5349423 .

Watford-Spence ATR. The effectiveness and usability of companion robots in combating loneliness in the elderly in mental healthcare : a scoping literature review 2022.

Bemelmans R, Gelderblom GJ, Jonker P, de Witte L. Socially assistive robots in elderly care: a systematic review into effects and effectiveness. J Am Med Dir Assoc. 2012;13:114-120.e1. https://doi.org/10.1016/j.jamda.2010.10.002 .

Pu L, Moyle W, Jones C, Todorovic M. The effectiveness of social robots for older adults: a systematic review and meta-analysis of randomized controlled studies. Gerontologist. 2019;59:e37-51. https://doi.org/10.1093/geront/gny046 .

Bousardt HM. Using embodied conversational agents to combat loneliness in older adults : a scoping review 2022.

Abbott R, Orr N, McGill P, Whear R, Bethel A, Garside R, et al. How do “robopets” impact the health and well-being of residents in care homes? A systematic review of qualitative and quantitative evidence. Int J Older People Nurs. 2019;14: e12239. https://doi.org/10.1111/opn.12239 .

Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to meta-analysis. Wiley: Hoboken; 2021.

Book   Google Scholar  

Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372: n71. https://doi.org/10.1136/bmj.n71 .

Chen S-C, Moyle W, Jones C, Petsky H. A social robot intervention on depression, loneliness, and quality of life for Taiwanese older adults in long-term care. Int Psychogeriatr. 2020;32:981–91. https://doi.org/10.1017/S1041610220000459 .

Rahatul A. Can Robots Help with Loneliness? University of Manitoba, 2020.

Bennett CC, Sabanovic S, Piatt JA, Nagata S, Eldridge L, Randall NA, Robot a Day Keeps the Blues Away. IEEE Int. Conf Healthc Inform ICHI. 2017;2017:536–40. https://doi.org/10.1109/ICHI.2017.43 .

Kanamori M, Suzuki M, Tanaka M. Maintenance and improvement of quality of life among elderly patients using a pet-type robot. Jpn J Geriatr. 2002;39:214–8. https://doi.org/10.3143/geriatrics.39.214 .

Fields N, Xu L, Greer J, Murphy E. Shall I compare thee...to a robot? An exploratory pilot study using participatory arts and social robotics to improve psychological well-being in later life. Aging Ment Health 2021;25:575–84. https://doi.org/10.1080/13607863.2019.1699016 .

Ollier J, Neff S, Dworschak C, Sejdiji A, Santhanam P, Keller R, et al. Elena+ Care for COVID-19, a Pandemic Lifestyle Care Intervention (preprint) Preprint.

Jones VK, Hanus M, Yan C, Shade MY, Blaskewicz Boron J, Maschieri BR. Reducing loneliness among aging adults: the roles of personal voice assistants and anthropomorphic interactions. Front Public Health. 2021;9:750736. https://doi.org/10.3389/fpubh.2021.750736 .

Kramer LL, van Velsen L, Clark JL, Mulder BC, de Vet E. Use and effect of embodied conversational agents for improving eating behavior and decreasing loneliness among community-dwelling older adults: randomized controlled trial. JMIR Form Res. 2022;6: e33974. https://doi.org/10.2196/33974 .

Fogelson DM, Rutledge C, Zimbro KS. The Impact of Robotic Companion Pets on Depression and Loneliness for Older Adults with Dementia During the COVID-19 Pandemic. J Holist Nurs Off J Am Holist Nurses Assoc 2021:8980101211064605. https://doi.org/10.1177/08980101211064605 .

Bickmore T, Caruso L, Clough-Gorr K, Heeren T. ‘It’s just like you talk to a friend’ relational agents for older adults. Interact Comput. 2005;17:711–35. https://doi.org/10.1016/j.intcom.2005.09.002 .

Banks MR, Willoughby LM, Banks WA. Animal-assisted therapy and loneliness in nursing homes: use of robotic versus living dogs. J Am Med Dir Assoc. 2008;9:173–7. https://doi.org/10.1016/j.jamda.2007.11.007 .

Robinson NL, Cottier TV, Kavanagh DJ. Psychosocial health interventions by social robots: systematic review of randomized controlled trials. J Med Internet Res. 2019;21: e13203. https://doi.org/10.2196/13203 .

Papadopoulos C, Castro N, Nigath A, Davidson R, Faulkes N, Menicatti R, et al. The CARESSES randomised controlled trial: exploring the health-related impact of culturally competent artificial intelligence embedded into socially assistive robots and tested in older adult care homes. Int J Soc Robot. 2022;14:245–56. https://doi.org/10.1007/s12369-021-00781-x .

Billingham SA, Whitehead AL, Julious SA. An audit of sample sizes for pilot and feasibility trials being undertaken in the United Kingdom registered in the United Kingdom Clinical Research Network database. BMC Med Res Methodol. 2013;13:104. https://doi.org/10.1186/1471-2288-13-104 .

Robinson H, Macdonald B, Kerse N, Broadbent E. The psychosocial effects of a companion robot: a randomized controlled trial. J Am Med Dir Assoc. 2013;14:661–7. https://doi.org/10.1016/j.jamda.2013.02.007 .

Loveys K. Developing engaging digital humans for psychotherapeutic applications. Thesis. ResearchSpace@Auckland, 2021.

Vardoulakis LP, Ring L, Barry B, Sidner CL, Bickmore T. Designing relational agents as long term social companions for older adults. In: Nakano Y, Neff M, Paiva A, Walker M, editors. Intell. Virtual Agents, Berlin, Heidelberg: Springer; 2012, p. 289–302. https://doi.org/10.1007/978-3-642-33197-8_30 .

Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, et al. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Ann Behav Med. 2013;46:81–95. https://doi.org/10.1007/s12160-013-9486-6 .

Mertens S, Herberz M, Hahnel UJJ, Brosch T. The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains. Cogn Sci n.d.:10.

Maier M, Bartoš F, Stanley TD, Shanks DR, Harris AJL, Wagenmakers E-J. No evidence for nudging after adjusting for publication bias. Proc Natl Acad Sci. 2022;119:e2200300119. https://doi.org/10.1073/pnas.2200300119 .

Szaszi B, Higney A, Charlton A, Gelman A, Ziano I, Aczel B, et al. No reason to expect large and consistent effects of nudge interventions. Proc Natl Acad Sci. 2022;119: e2200732119. https://doi.org/10.1073/pnas.2200732119 .

Borenstein M. Common mistakes in meta-analysis and how to avoid them. Englewood, NJ, USA: Biostat, Inc.; 2019.

Vevea JL, Woods CM. Publication bias in research synthesis: sensitivity analysis using a priori weight functions. Psychol Methods. 2005;10:428–43. https://doi.org/10.1037/1082-989X.10.4.428 .

NHS. The Topol Review: Preparing the healthcare workforce to deliver the digital future. London: Health Education England; 2019.

Husk K, Elston J, Gradinger F, Callaghan L, Asthana S. Social prescribing: where is the evidence? Br J Gen Pract. 2019;69:6–7. https://doi.org/10.3399/bjgp19X700325 .

Masi CM, Chen H-Y, Hawkley LC, Cacioppo JT. A Meta-Analysis of Interventions to Reduce Loneliness. Personal Soc Psychol Rev Off J Soc Personal Soc Psychol Inc 2011;15: https://doi.org/10.1177/1088868310377394 . https://doi.org/10.1177/1088868310377394 .

Kimberlee R, Bertotti M, Dayson C, Asthana S, Polley M, Burns L, et al. The economic impact of social prescribing 2022.

Akhter-Khan SC, Au R. Why Loneliness Interventions Are Unsuccessful: A Call for Precision Health. Adv Geriatr Med Res 2020;4. https://doi.org/10.20900/agmr20200016 .

Dale R. GPT-3: What’s it good for? Nat Lang Eng. 2021;27:113–8. https://doi.org/10.1017/S1351324920000601 .

The Economist. Is Google’s 20-year dominance of search in peril? The Economist 2023.

Quintana DS. A guide for calculating study-level statistical power for meta-analyses. Adv Methods Pract Psychol Sci. 2023;6:25152459221147260. https://doi.org/10.1177/25152459221147260 .

Nakagawa S. A farewell to Bonferroni: the problems of low statistical power and publication bias. Behav Ecol. 2004;15:1044–5. https://doi.org/10.1093/beheco/arh107 .

Chen S-C, Jones C, Moyle W. The Impact of Engagement with the PARO Therapeutic Robot on the Psychological Benefits of Older Adults with Dementia. Clin Gerontol 2022;0:1–13. https://doi.org/10.1080/07317115.2022.2117674 .

Buecker S, Mund M, Chwastek S, Sostmann M, Luhmann M. Is loneliness in emerging adults increasing over time? A preregistered cross-temporal meta-analysis and systematic review. Psychol Bull. 2021;147:787–805. https://doi.org/10.1037/bul0000332 .

ter Stal S, Kramer LL, Tabak M, op den Akker H, Hermens H. Design features of embodied conversational agents in ehealth: a literature review. Int J Hum-Comput Stud 2020;138:102409. https://doi.org/10.1016/j.ijhcs.2020.102409 .

Turkle S. Reclaiming Conversation: The Power of Talk in a Digital Age. New York: New York: Penguin Publishing Group; 2015.

Jeffrey K, Abdallah S, Michaelson J. The Cost of Loneliness to UK Employers. London: New Economics Foundations; 2017.

Mihalopoulos C, Le LK-D, Chatterton ML, Bucholc J, Holt-Lunstad J, Lim MH, et al. The economic costs of loneliness: a review of cost-of-illness and economic evaluation studies. Soc Psychiatry Psychiatr Epidemiol 2020;55:823–36. https://doi.org/10.1007/s00127-019-01733-7 .

Peytrignet S, Garforth-Bles S, Keohane K. Loneliness monetisation report. London: Simetrica; 2020.

Ollier J, Suryapalli P, Fleisch E, Wangenheim F von, Mair JL, Salamanca-Sanabria A, et al. Can digital health researchers make a difference during the pandemic? Results of the single-arm, chatbot-led Elena+: Care for COVID-19 interventional study. Front Public Health 2023;11. https://doi.org/10.3389/fpubh.2023.1185702 .

Download references

Acknowledgements

We would like to thank LSE staff for additional feedback, including Alina Velias, Andra Fry, Jessica Kong, and Georgia Nichols. We thank Nina Shahrizad for proofreading. We would also like to thank all primary authors who made this systematic review and meta-analysis possible.  After our analysis was completed, a final version of one of our included studies [ 36 ] was published, which was a preprint. This final version is available here [ 70 ].

Not applicable.

Author information

Authors and affiliations.

Department of Psychological and Behavioural Science, LSE, London, UK

Sia Sha, Haoran Shi, Dario Krpan & Matteo Galizzi

Department of Psychological Medicine, The University of Auckland, Auckland, New Zealand

Kate Loveys

Manchester Institute of Education, University of Manchester, Manchester, UK

Pamela Qualter

You can also search for this author in PubMed   Google Scholar

Contributions

SS drafted the first version of the protocol and search strategy. All other authors provided feedback and approved the final protocol. SS, KL, and HS screened the titles and abstracts and full texts of citations, and these authors also extracted data from included studies. SS, KL, HS, and DK conducted risk of bias assessment. SS completed the data analysis. SS, KL, and PQ co-wrote the first version of the final manuscript. All authors provided feedback and approved the final manuscript. All authors had access to the underlying data. SS, KL, and HS verified the data.

Corresponding author

Correspondence to Sia Sha .

Ethics declarations

Ethics approval and consent to participate.

Since the systematic review and meta-analysis synthesised publicly available data, ethics approval was not required.

Consent for publication

Competing interest.

The authors declare no competing interests.

Additional information

Publisher’s note.

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

Supplementary Information

Supplementary material 1., supplementary material 2., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Sha, S., Loveys, K., Qualter, P. et al. Efficacy of relational agents for loneliness across age groups: a systematic review and meta-analysis. BMC Public Health 24 , 1802 (2024). https://doi.org/10.1186/s12889-024-19153-x

Download citation

Received : 11 October 2023

Accepted : 14 June 2024

Published : 06 July 2024

DOI : https://doi.org/10.1186/s12889-024-19153-x

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

• Relational agents
• Social robots

BMC Public Health

ISSN: 1471-2458

• General enquiries: [email protected]

NTRS - NASA Technical Reports Server

Filter results, collections, top results, subject category, distribution.