We need a Journal of Disappointing Results

I recently had the privilege of hearing about an unpublished study that would save (academic, corporate or nonprofit) scientists in the field a lot of misery if they knew about it. But the study “failed” in the sense that it did not produce the hoped-for results, the intervention failed to show any effect. There are countless obstacles to the publication of scientific studies, such as the need to maintain secrecy, to avoid signaling directional strategies to competitors. When the studies fail to meet their “expected endpoints”, there is a whole range of additional obstacles, the biggest of which is cultural: one simply does not publish disappointed results, for fear of ridicule or simply because it is the norm. And yet, they are as valuable as their successful cousins. In the era of AI-everything, not having “negative examples” is lethal. We need a repository of p>0.1 studies, a Journal of Disappointing Results.

Another, probably more pragmatic, approach (as I understand it) is the one taken by trailblazers @rene wegrzyn and @anna marie wagner with their new @transfyr.bio venture, using AI to help surface knowledge buried in research notes and in implicit know-how. They are tackling one of the most difficult and important problems in science.