2026 Volume 41 Issue 2 Pages 115-127
This paper addresses the reproducibility crisis in life science research by focusing on the critical role of tacit knowledge in experimental operations. In wet laboratories, expert researchers rely on unconscious, fine-tuned adjustments—such as modifying aspiration speeds based on liquid behavior or mitigating invisible air bubbles—that are rarely documented in SOPs. Conventional automation approaches often fail to ensure reproducibility because they merely mimic fixed procedures, overlooking these critical invisible adjustments. To address this challenge, we propose a four-layer model—Observation, Representation, Execution, and Verification—that redefines automation as a process of knowledge conversion. Using this framework, we analyze why tacit knowledge remains elusive, attributing it to a lack of descriptive language and insufficient sensing of physical states. Through case studies of automated Hi-C and RamDA-seq protocols, we argue that the essence of reproducibility lies not in repeating the Same Procedure, but in achieving the Same State Transition by dynamically adapting to variations in samples and environments. Finally, we discuss methodologies for formalizing tacit knowledge through the observation of physical proxies and future perspectives on reconstructing field knowledge in the era of generative AI.
