Acceptance and Use of Large Language Models in Mental Health Professionals: A Cross-Sectional Observational Study

Abstract

Introduction:Large language models (LLMs) could have high potential to improve mental health care, but their effective implementation depends on professional acceptance. This study examined mental health professionals’ acceptance and use of LLMs, and tested the applicability of the Unified Theory of Acceptance and Use of Technology (UTAUT2) model with additional determinants of acceptance.Methods:A cross-sectional online survey among mental health professionals in Germany was conducted. Participants reported LLM use, use intentions, and learning intentions across twelve application areas. Acceptance and its determinants were assessed using UTAUT2 constructs and additional constructs (loss of autonomy, legal and ethical ambiguity, client empowerment, client mistrust). Structural equation modelling with full information maximum likelihood estimation was applied.Results:Data from N = 269 professionals was analysed. Acceptance of LLMs in psychotherapy was low for 44% and moderate for 41% of participants. Private use (75.1%) exceeded professional use (18.6%), which was most common for documentation (30.9%). Use and learning intentions were moderate. The extended UTAUT2 model (RMSEA=0.06; SRMR=0.06; CFI=0.93; TLI=0.92) explained 92% of the variance in acceptance. Significant predictors were performance expectancy (γ=0.39, p<0.001), social influence (γ=0.11, p=0.034), price value (γ=0.30, p<0.001), habit (γ=0.14, p=0.011), loss of autonomy (γ=-0.16 , p=0.001), and client mistrust (γ=-0.12 , p=0.013).Conclusion:LLMs are already being integrated into clinical practice, yet acceptance among professionals remains moderate. Perceived usefulness emerged as key driver of acceptance, whereas concerns about autonomy and client trust represented barriers. Training, clear guidelines, and ethical standards may facilitate responsible integration.

Ähnliche Arbeiten