MétaCan
Menu
Back to cohort
Record W4408084770 · doi:10.2196/69329

Perception of Empathy in Mental Health Care Through Voice-Based Conversational Agent Prototypes: Experimental Study

2025· article· en· W4408084770 on OpenAlexvenueno aff
Ruvini Sanjeewa, Ravi Iyer, Pragalathan Apputhurai, Nilmini Wickramasinghe, Denny Meyer

Bibliographic record

VenueJMIR Formative Research · 2025
Typearticle
Languageen
FieldPsychology
TopicDigital Mental Health Interventions
Canadian institutionsnot available
Fundersnot available
KeywordsPreprintEmpathyPerceptionPsychologyCognitive psychologySocial psychologyComputer scienceNeuroscienceWorld Wide Web

Abstract

fetched live from OpenAlex

Background: Empathy is a critical component of effective mental health care communication. Positive perceptions of empathy in conversational agents (CAs) operating in the health care domain are therefore needed to enhance the quality of care provided by these emerging technologies. However, research on how users perceive empathy in CAs is limited, particularly in voice-based prototypes. Objective: The objective of this study is to identify to what extent perceptions of empathy in CA prototypes correspond with the engineered empathy levels for these voice-based prototypes. In addition, as a secondary aim, this study investigates how the demographic characteristics of participants affect their perception of empathy in a mental health helpline service context. Methods: Swinburne University first-year psychology students (N=306) were presented with 9 CA prototypes engineered to portray low, medium, or high empathy levels, and their perceptions of empathy were collected via an electronic survey. Perceptions of empathy were rated using the Perceived Emotional Intelligence (PEI) Scale and the Raters' Scale (RS10). Results: Most participants were female (233/306, 76%) with a mean age of 30 (SD 10.69) years, while a majority (194/306, 63%) were of Australian and New Zealand background. A strong positive correlation between the PEI and RS10 ratings was observed (r=0.829, P<.001). The empathy ratings across the 3 engineered empathy levels showed significant differences when using both PEI (χ22=11.865, P=.003) and RS10 (χ22=19.737, P<.001) measures. A linear mixed model for PEI showed significantly higher ratings for high rather than low engineered empathy levels (t8=-2.34, P=.048). RS10 ratings were also significantly higher for high rather than low engineered empathy levels (t8=-2.45, P=.04). However, no significant differences were detected between the CAs with engineered medium-level empathy and the CAs with low or high engineered empathy levels. The linear mixed model for PEI showed significantly higher ratings for participants of the Asian and Other ethnic categories compared to the Oceanic category (t285=2.54, P=.01 and t286=2.25, P=.03 respectively). The RS10 ratings were also significantly higher for the Other category rather than for the Oceanic category (t284=2.24, P=.03). Women showed significantly higher RS10 ratings than men (t283=1.94, P=.05). Conclusions: Recognizing empathy levels in CA prototypes proved challenging, highlighting possible complexities involved with voice-based empathy detection. The perception of empathy may also be affected by different ethnic and gender-based factors. The study findings emphasize the importance of personalized communications by CAs, with expressions of empathy tailored to key demographic characteristics of users. Future studies in a similar context would benefit from the inclusion of participants who are end users of a mental health care service with more balanced gender and age distributions. Multimodal interactions could also be considered for CA prototype development.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.004
metaresearch head score (Gemma)0.015
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.005
Threshold uncertainty score0.023

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0040.015
Meta-epidemiology (narrow)0.0010.000
Meta-epidemiology (broad)0.0010.000
Bibliometrics0.0000.000
Science and technology studies0.0010.001
Scholarly communication0.0020.001
Open science0.0010.002
Research integrity0.0010.002
Insufficient payload (model declined to judge)0.0050.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.109
GPT teacher head0.542
Teacher spread0.432 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designBench or experimental
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations7
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueJMIR Formative ResearchSame topicDigital Mental Health InterventionsFrench-language works237,207