Year: 2026
Type: article
Abstract: The integration of Artificial Intelligence into scientific discovery represents a paradigm shift in how research is conducted across disciplines. This study investigates the effectiveness of AI-driven automation in accelerating and enhancing scientific discovery through a mixed-methods design combining quantitative performance analysis with qualitative expert evaluation. Using datasets from materials science, fluid dynamics, and autonomous experimentation platforms, we examined how AI systems—including large language models, autonomous agents, and self-driving laboratories—perform across the scientific workflow from hypothesis generation to experimental execution and knowledge dissemination. Quantitative results demonstrate that AI-driven autonomous experimentation significantly increases the diversity of explored phenomena compared to conventional optimization routines, with novelty discovery scores improving by up to 31% over traditional approaches. AI agent frameworks successfully automated the complete research cycle, including hypothesis generation, experimental design, robotic execution, data analysis, and manuscript preparation, reducing research cycle time by approximately 60% in well-defined domains. However, qualitative findings revealed that AI systems face persistent challenges in identifying conceptual gaps that require structural insight beyond existing knowledge frameworks. Expert evaluation indicated that while AI excels at search, optimization, and pattern recognition, human researchers remain essential for formulating fundamentally novel research directions and providing interpretive judgment. These findings contribute to a nuanced understanding of AI’s role in scientific discovery and provide practical guidelines for designing human-AI collaborative research systems that balance automation with scientific creativity.
Author: Dasari, Roja
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7217506496