Trustless: episode 6 -- trustless science

In the previous episode we discussed politics, where trustlessness has been put to very good use. Another area that is moving into this direction is science.

You may have heard that science has a replicability crisis. This means that many scientific results can't be replicated by others, due to lack of sufficient detail, access to necessary data, or flaws in the authors' study.

Many fields are now moving towards a more "trustless" model, wherein each study must publish its full input data along with the study. Ideally, it should also publish the full code that they used to analyze their data, basically every step from raw data to figures, tables and statistical tests.

This does not fully avoid academic fraud, in the sense that people could still fake the input data (unless its integrity could be checked by a blockchain?), but it could remove a large amount of unserious/shoddy work.

What do you think, is trustless science a good idea? What are the hidden costs?