초거대 언어 모델 기반 생성형 AI의 정신건강 상담 가능성: 튜링 테스트를 통한 답변의 질과 신뢰성 평가

Abstract

AI has achieved groundbreaking advancements across various fields, opening new possibilities. In particular, generative AI based on large-scale language models has demonstrated exceptional performance in text generation, translation, and question-answering tasks, while also garnering attention for its potential in psychological counseling. Psychological counseling plays a vital role in addressing individuals' mental and emotional challenges, yet many people are unable to access proper assistance due to professional shortages, high costs, and limited accessibility. As an alternative to tackle these issues, AI-driven psychological counseling has been proposed, offering advantages such as cost efficiency, 24/7 availability, and enhanced privacy protection. However, questions remain about whether generative AI can provide counseling responses of quality and reliability comparable to human experts. This study aims to evaluate whether generative AI, leveraging large-scale language models, can deliver responses in youth counseling comparable to those provided by professional counselors. To this end, the study applies the Turing Test to assess the quality and reliability of AI-generated counseling responses and analyzes feedback from professional counselors. The research explores the practical potential of AI in the psychological counseling domain, highlighting its ability to complement or even substitute human experts.