Evaluating the Quality of Psychotherapy Conversational Agents: Framework Development and Cross-Sectional Study
Bibliographic record
Abstract
Background: Despite potential risks, artificial intelligence-based chatbots that simulate psychotherapy are becoming more widely available and frequently used by the general public. A comprehensive way of evaluating the quality of these chatbots is needed. Objective: To address this need, we developed the CAPE (Conversational Agent for Psychotherapy Evaluation) framework to aid clinicians, researchers, and lay users in assessing psychotherapy chatbot quality. We use the framework to evaluate and compare the quality of popular artificial intelligence psychotherapy chatbots on the OpenAI GPT store. Methods: We identified 4 popular chatbots on OpenAI's GPT store. Two reviewers independently applied the CAPE framework to these chatbots, using 2 fictional personas to simulate interactions. The modular framework has 8 sections, each yielding an independent quality subscore between 0 and 1. We used t tests and nonparametric Wilcoxon signed rank tests to examine pairwise differences in quality subscores between chatbots. Results: Chatbots consistently scored highly on the sections of background information (subscores=0.83-1), conversational capabilities (subscores=0.83-1), therapeutic alliance, and boundaries (subscores=0.75-1), and accessibility (subscores=0.8-0.95). Scores were low for the therapeutic orientation (subscores=0) and monitoring and risk evaluation sections (subscores=0.67-0.75). Information on training data and knowledge base sections was not transparent (subscores=0). Except for the privacy and harm section (mean 0.017, SD 0.00; t3=∞; P<.001), there were no differences in subscores between the chatbots. Conclusions: The CAPE framework offers a robust and reliable method for assessing the quality of psychotherapy chatbots, enabling users to make informed choices based on their specific needs and preferences. Our evaluation revealed that while the popular chatbots on OpenAI's GPT store were effective at developing rapport and were easily accessible, they failed to address essential safety and privacy functions adequately.
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 imitationNot 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.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.146 | 0.134 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.006 | 0.004 |
| Science and technology studies | 0.003 | 0.002 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
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.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
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".