상세 보기
초록
This study investigates GPT-4’s ability to assess scientific inquiry writing tasks, focusing on its capacity to justify evaluations with evidence. Scientific inquiry involves problem definition, hypothesis formulation, data analysis, and experimental design. Using scenario-based scientific questions, GPT-4’s responses were compared with human evaluators to analyze its reasoning and limitations. Results indicate that GPT-4 can provide valid reasoning under specific conditions but struggles with complex data interpretation. This research highlights GPT-4’s potential and limitations as an automated evaluation tool, addressing inconsistencies in human evaluation and suggesting future directions for AI-assisted educational assessments.
키워드
Automated Agent; Large Language Model; ChatGPT; Scientific Inquiry; Capability Evaluation; Evidence Extraction
- 제목
- GPT-4기반 과학적 탐구 글쓰기 역량의 평가 자동화를 위한 언어 모델의 투명성과 시각화에 관한 연구
- 제목 (타언어)
- Automated Robot Agent for Evaluating the Scientific Inquiry Capabilities via GPT-4: A Study on the Transparency and Visualization in Large Language Model
- 저자
- 윤지나; 이효정; 이성혜; 박소영; 동서연; 심주용
- 발행일
- 2025-06
- 저널명
- 로봇학회 논문지
- 권
- 20
- 호
- 2
- 페이지
- 200 ~ 209