CoChat

Use and Effect of Embodied Conversational Agents for Improving Eating Behavior and Decreasing Loneliness Among Community-Dwelling Older Adults: Randomized Controlled Trial

Open access
Kramer, Lean L, van Velsen, Lex, Clark, Jenna L, Mulder, Bob C, de Vet, Emely · 2022 · JMIR Formative Research

Published version of a preprint

The DOI matches its Crossref record; 59 of the 59 references it cites were found in a registry or an index; 12 of the 22 sources its claims cite were checked for support.

Abstract

Background Embodied conversational agents (ECAs) have been proposed as a promising interaction modality for the delivery of programs focused on promoting lifestyle changes. However, it is not understood what factors influence the health effects of ECAs or their use. Objective We aimed to (1) identify whether ECAs could persuade community-dwelling older adults to change their dietary behavior and whether ECA use could decrease loneliness, (2) test the pathways to these effects, and (3) understand factors influencing the use of ECAs. Methods A randomized controlled trial was conducted. The intervention group received access to the PACO service for 8 weeks. The waitlist group started PACO use after waiting for 4 weeks. Two primary outcomes (eating behavior and loneliness) were assessed via online questionnaires at intake, upon joining the waitlist, after 4 weeks, and after 8 weeks. The third primary outcome (use) was assessed via data logs. Secondary outcomes were measured at the same time points, via questionnaires or an optional interview. Results In total, 32 participants completed the intervention. We found a significant correlation between use in minutes on the one hand, and perceived usefulness (r=0.39, P=.03) and enjoyment on the other (r=0.38, P=.03). However, these did not predict use in the full regression model (F2,29=1.98, P=.16, R2=0.12). Additionally, PACO use did not lead to improvement in eating behavior (χ22=0.34, P=.85) or a decrease in loneliness (χ22=0.02, P=.99). Conclusions Our study did not provide any concluding evidence about factors that are linked to the use or health effects of ECAs. Future service design could benefit from either creating a functional design catering to the predominant stage in the precaution adoption process model of the targeted population, or by personalizing the service based on an intake in which the end user’s stage is determined. Trial Registration ClinicalTrials.gov NCT04510883; https://clinicaltrials.gov/ct2/show/NCT04510883 International Registered Report Identifier (IRRID) RR2-10.2196/22186

Overview

CoChat Summary

## TL;DR

This randomized controlled trial of the PACO embodied conversational agent (ECA) in community-dwelling older adults found no evidence that the service improved eating behavior or reduced loneliness. Use was not explained in the full model by perceived usefulness or enjoyment, although both correlated with minutes of use.

## Key findings

  • Of the 51 participants who met inclusion criteria, 32 completed the intervention; the mean age was 73.00 years (SD 5.33, range 65–85) and 18 (56%) were women.
  • PACO use declined significantly over time: minutes per week dropped from a median of 69.66 in week 1 to 21.57 in week 8 (Friedman test, χ2_7=31.46, P<.001).
  • PACO did not improve eating behavior (χ2_2=0.34, P=.85) or decrease loneliness (χ2_2=0.02, P=.99); there were also no significant time effects for quality of life, autonomy, competence, or relatedness.
  • Perceived usefulness (r=0.39, P=.03) and enjoyment (r=0.38, P=.03) correlated with use in minutes, but neither predicted use in the full regression model (F2,29=1.98, P=.16, R2=0.12).

## Methods

  • Randomized controlled trial: cohort 1 received immediate access to the 8-week PACO intervention, and cohort 2 was placed on a 4-week waitlist before receiving access.
  • Sample: older adults aged 65 years or older, not in paid employment, living alone and independently at home, Dutch-speaking, able to use a tablet or computer, and with wireless internet; participants were recruited via research panels, flyers, newspapers, and social media.
  • Outcomes and analysis: use was logged from the PACO backend; eating behavior and loneliness were measured by questionnaires at intake, baseline, 4 weeks, and 8 weeks; analyses included Spearman correlations, Mann-Whitney U, Friedman tests, repeated-measures ANOVA, linear regression, and thematic analysis of interviews.

## Limitations

  • The authors state that response to recruitment was very low, the sample size was small, and overall statistical power was low.
  • The authors state they did not measure precaution adoption process model (PAPM) stage, which they think may have helped explain use and effectiveness.
  • The authors state the ongoing COVID-19 pandemic might have affected results and counteracted effects on loneliness; additionally, the text makes evident that the study relied on self-reported eating behavior and that participants were mostly not willing to pay for PACO.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.2196/33974

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Conceptual models explaining embodied conversational agent (ECA) use and health effects.
Figure 1 Conceptual models explaining embodied conversational agent (ECA) use and health effects.CC BY · PubMed Central
Figure 2 PACO home screen.
Figure 2 PACO home screen.CC BY · PubMed Central
Figure 3 Minutes per week.
Figure 3 Minutes per week.CC BY · PubMed Central
Figure 4 Relationship with the embodied conversational agents over time.
Figure 4 Relationship with the embodied conversational agents over time.CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 1 Not valid · 3 Not checked · 59 Own results and interpretations

30 more claims

Create a free account to see all 70 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 11
  • Probably not a substantial claim. 26
  • Borderline: one of the tests for a substantial claim is unclear. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

1 of 1

Citations

Cites

Show all 59 references

Research works printed but not linked: 1

  1. Gratch, J, 2007, IVA 2007, 125

Cited by

Metadata

Authors
Kramer, L. L., van Velsen, L., Clark, J. L., Mulder, B. C., & de Vet, E.
Affiliations
  • Wageningen University & Research (NL)
  • Duke University (US)
Corresponding author
Kramer, Lean L
Venue
JMIR Formative Research
Published
2022-04-11
Volume
6
Issue
4
Pages
e33974
ISSN
2561-326X (ISSN-L)
OpenAlex
W4213319785
Cited by
31
References
50
Citations by year
2026: 10 · 2025: 8 · 2024: 9 · 2023: 4
FWCI
4.96
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • JMIR Formative Research (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Socio-Environmental Systems Modeling (submittedVersion, cc-by)
  • University of Twente Research Information (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • Data Archiving and Networked Services (DANS) (submittedVersion, cc-by)
Topics
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Health Sciences › Medicine › Geriatrics and Gerontology › Technology Use by Older Adults
  • Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
Keywords
embodied conversational agents, dietary habits, loneliness, older adults, community-dwelling older adults, dietary behavior change, perceived usefulness, enjoyment, Paco, Precaution Adoption Process Model, lifestyle modification

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.