[4R1-OS-8a] AIとデモクラシー

[4R1-OS-8a-03] Assessing the Impact of AI Conversational Agents on the Mental Health and Self-Efficacy of Afghan Women

〇Tianchen Wang¹, Sofia Sahab¹, Jawad Haqbeen¹, Takayuki Ito¹ (1. Kyoto University)

キーワード：Conversational Agent, Mental Health, Self-efficacy

This study investigates the potential of AI conversational agents as a support tool for Afghan women facing mental health challenges post-2021.It assesses the effectiveness of these agents in providing a safe space for sharing experiences and improving mental health and self-efficacy. In a controlled experiment, 60 Afghan women interacted for an hour with one of three AI agents: a standard GPT-4 agent (Agent A), an instructional-enhanced agent (Agent B), and an empathy-enhanced agent (Agent C). Well-being and self-efficacy were measured using the WHO-5 Well-being Index and a General Self-Efficacy scale before the experiment and at a one-week follow-up. Preliminary findings indicate that Agent C significantly improves well-being and self-efficacy compared to Agents A and B. These initial insights highlight the potential of AI conversational agents as tools for psychological support.