Negative results are an essential and unavoidable part of the research process. When something does not work as expected, taking the time to understand why leads to the creation of new knowledge that informs the research process, and enables other promising research directions moving forward. However, negative results are often undervalued in the peer review process of ML/DL venues.
At the ICBINB initiative, we have put together a repository of (unexpected) negative results. The goal of such a repository is to give more visibility and promote research focused on unexpected negative results, and to facilitate spread of information (as a wiki or shared database) for other researchers to benefit from previous attempts of others.
Researchers can benefit two fold: either by saving time not reproducing something that did not work previously, or by building on top of those failed attempts that might inspire novel pieces of knowledge or methodologies.
Check the Repo prototype within the Notion platform (below a screenshot):
We would like to collect some early feedback on such a repo of negative results, and would appreciate it very much if you could spend 1-2 min answering this form with your thoughts/some first reactions.
1. What do you think of this project (any first reactions)?
2. Would you use such a repo and how? What features would make you want to use such repo?
3. Any thoughts/ideas on how to make this initiative impactful? Any questions or thoughts we should consider?