In scientific research, reproducing experimental results often proves more complicated than following written protocols, a reality that highlights the crucial role of tacit knowledge—skills and insights that are learned through experience but difficult to document. This challenge has spurred efforts to capture the subtle, often unconscious actions of researchers that influence experimental outcomes.

Scientists routinely rely on lab notebooks, published papers, and protocols to communicate methods, but these records rarely capture all the nuanced decisions made during experiments. For instance, subtle variations in how a tube is shaken or the timing of reagent addition can significantly affect results. Researchers with so-called “magic hands” consistently achieve reliable outcomes, though they may not fully understand or articulate the reasons behind their success.

Tacit knowledge, a concept identified in the 1960s by Hungarian chemist Michael Polanyi, acknowledges that “we know more than we can tell.” While this unspoken expertise is accepted as integral to laboratory work, it also complicates replication efforts. Attempts to reproduce studies sometimes fail because essential details are missing or because replicators inadvertently deviate from original procedures.

This issue gained prominence in various fields, including psychology. The 1993 study suggesting that listening to Mozart temporarily boosts reasoning skills, known as the Mozart effect, was difficult to replicate. While the original researcher attributed failures to replicators’ lack of expertise, subsequent comprehensive reviews found little evidence supporting the phenomenon.

To bridge this gap, some initiatives have aimed to make tacit knowledge more accessible. For example, the Journal of Visualized Experiments (JoVE) provides video demonstrations of complex experiments, and platforms like Protocols.io facilitate protocol sharing and discussion among researchers, helping to onboard new lab members more effectively.

Building on these efforts, the biotechnology company Transfyr has developed a system that captures extensive data during experiments—including video, audio, sensor readings, and supply chain information such as lot numbers—to identify subtle factors influencing results. Using artificial intelligence to analyze equipment use and researcher movements, Transfyr has uncovered significant procedural variations among experienced scientists performing the same task.

In one case, technicians were found to be allowing a chemical reaction to proceed 30 seconds longer than intended, a mistake that paradoxically enhanced the outcome. The system also documented differences in total protocol durations between researchers, revealing inefficiencies or hidden optimizations not easily apparent from written procedures.

Transfyr has partnered with a diagnostics company and other clients interested in applying this technology to track and improve their research processes. Experts suggest that this approach could eventually detect both errors and unexpected breakthroughs by reviewing detailed, time-stamped records of experimental work alongside AI analysis.

However, some caution that while the approach shows promise—particularly in mature fields like molecular biology—it is not a guaranteed solution across all scientific disciplines. Complex fields with less consensus on methodology may not benefit as readily from intensive data capture and analysis.

Further validation through controlled studies comparing research progress with and without the technology will be necessary to determine its impact on the reproducibility and advancement of scientific knowledge. Nonetheless, efforts to quantify and communicate tacit scientific skills represent a significant step toward addressing longstanding challenges in experimental science.