Development of an Autonomous Scientific Discovery System Based on Generative Artificial Intelligence

The rapid growth of scientific publications and the increasing complexity of multidisciplinary research have made conventional scientific discovery slower, less scalable, and less effective in identifying hidden knowledge across research domains. Recent advances in Generative Artificial Intelligence (Generative AI), Large Language Models (LLMs), and knowledge graphs offer new opportunities to automate scientific discovery. This study aims to develop and evaluate an Autonomous Scientific Discovery System Based on Generative AI capable of integrating scientific knowledge, understanding literature, generating hypotheses, planning experiments, performing scientific reasoning, and producing structured research reports. The study adopts the Design Science Research (DSR) methodology to design and evaluate an intelligent prototype integrating knowledge graphs, literature understanding, Generative AI, Retrieval-Augmented Generation (RAG), scientific reasoning, hypothesis generation, experiment planning, evidence evaluation, and automated report generation. System performance is evaluated using expert assessment, scientific validation, and quantitative metrics, including Hypothesis Quality Score, Scientific Validity, Novelty, Precision, Recall, F1-score, BERTScore, Explainability Score, Time Reduction, and Discovery Success Rate. The results demonstrate that the proposed system successfully automates literature analysis, knowledge synthesis, hypothesis generation, and experiment planning, significantly improving research efficiency, hypothesis quality, interdisciplinary knowledge integration, and explainability compared with conventional research workflows. These findings demonstrate the feasibility of autonomous AI-supported scientific discovery while emphasizing the importance of rigorous validation, explainable reasoning, ethical governance, and human oversight for future deployment.