AI 기반 투자 의사결정 지원 챗봇에서 AI의 결과에 대한 설명 방법과 사용자의 위험 감수 수준의 영향 분석

초록

Objective: This study investigates the effects of explanation methods and users' risk-taking levels on trust, understanding, satisfaction, and investment behavior in AI-based investment chatbots. Background: AI-driven financial services increasingly require transparent explanations to build user trust and satisfaction. However, most explanation methods overlook user characteristics. This study examines how risk-taking levels impact users' perceptions and behaviors of AI explanations. Method: A 3 (explanation methods: factual, counterfactual, combined) × 2 (risk-taking levels: high, low) between-subject experiment was conducted using an AI chatbot developed with XGBoost and SHAP for explanations. 115 participants interacted with the chatbot in a P2P lending scenario. Trust, understanding, satisfaction, and investment amounts were measured. Results: Risk-taking levels significantly influenced trust, satisfaction, and investment amounts, while explanation methods affected understanding. High risk-takers showed higher satisfaction with factual explanations, despite lower understanding. Self-reported AI knowledge positively correlated with trust with AI results. Conclusion: Personalized explanation methods tailored to users' risk-taking levels and AI backgrounds enhance decision-making in AI-driven financial systems. Application: These findings guide the development of adaptive AI chatbots that personalize explanations, improving user experience and decision-making in realworld financial contexts.

키워드