Learning Well from Being Wrong | Niveditha Iyer

nivedithasi.github.io · 1,211 words · saved by 2 readers

Four years ago, as a summer research intern at a drug discovery lab, I had three months to build a pharmacokinetic prediction pipeline for small molecules. My task was to model the absorption, distribution, metabolism, excretion and toxicity (collectively "ADMET") characteristics of candidate drugs. ADMET is an important screening criterion for candidate compounds towards the long and expensive process of drug development. As a beginner scientist, I had many ideas and an acute shortage of time and resources to test them. If I had a magic wand that summer, I'd seek to quickly shrink my experiment list by scouring research archives for all relevant ideas that had been tested, not just the relatively few true positives. The reasons for sharing research failures are well understood. There is value in reporting errors that turn out to be about a system or tool (Nature). Sharing unsuccessful work at scale saves resources by eliminating unproductive approaches for the community (Reddit thread

This essay was a winner of the Astera Institute Metascience Competition. Based on it and other submissions' arguments that "null results ... often hold real scientific value but don’t fully see the light of day", Astera launched the Hidden Science Competition. Read other winning entries in What Scientists Said. ✦ · ✦ The unavailability of negative and null results in science leads to wasted time and resources in experimentation. It also skews scientific theory to fit positive results and ignore what didn't work. Scientific progress will accelerate if we incentivise sharing negative results…

Learning Well from Being Wrong | Niveditha Iyer

saved by

related reading