Hacker News Reader

Comment by natejenkins | Hacker News Reader

The problem with getting scientists to publish their data is that the incentives are not aligned properly. Clearly it is better for science in general, but the benefit to the scientist publishing the data is less clear. For what it's worth, I am working on reshaping the academic article where the raw data and analysis can be stored alongside the text and figures, and furthermore that the reader will be able to play with the data and analysis directly inside the article. The idea being that this will be a more interesting article from the reader's perspective, and thus result in more visibility and citations for the author.

This doesn't solve the problem for research where the datasets are in the many terabytes, but then again there are many papers where the datasets are well under a gigabyte.

View on HN

Incentives really are the problem - there's absolutely no reason for me, the researcher, to put much work into maintaining my data. If I do, it's because of some personal standing, or the good of the field or what have you.

No one got tenure for a well curated data set.

There may be a contrived way out of this conundrum, but I think really the most sensible way is simply regulation.

We've had a ton of scientists sacrificing their own individual reputation and relationship with publishers (and of their groups) in name of something everyone agrees but few really stand up for, because the personal gains are almost exclusively negative. That's textbook use case of regulation.

I don't know specifically what should be done but perhaps a requiring publishers certain obligations (e.g. responsibility of maintaining papers for a long date and turning them public afterwards); or maybe simply a universal obligation to open publications after say 5 years.

People fear this will compromise quality or sustainability of publishers. But the community need publishers. It's a tag of credibility. So if publishers are into trouble (and they're really needed) they'll find a way by e.g. demanding payments from publications from the most wealthy labs.

The NIH, Howard Hughes Medical Institute, and a number of other major funders of science both in the U.S. and U.K. already require papers funded with their money to be open access after 12 months.

The problem is that whether or not a paper is Open Access, "Data is available on request from the author" may be an undocumented bit of spaghetti code, may be stored on a Zip disk around here somewhere I'm sure, or may just be lost.

Regulating "You must make your data accessible, and maintain it well" is much harder to implement, and much harder to check. Some grants now have sections describing what will happen to the data etc., but right now there really is very little reason beyond their own personal desire for researchers to maintain good quality software and data repositories.

Precisely. Publications require data. Data costs time/money/brains. Grants/promotions/tenure/survival cost publications. It's not hard to see that data serves a role as currency, and there is little incentive to give it away as long as publications are what puts bread on the table.

Additionally, the discussion over shared data is usually a cry to improve reproducibility. In this publication-centric world, one can't publish a paper that is just "we reproduced an existing paper", so outsiders look to "poach" new applications or findings from the data.

Who's going to maintain it, even at the small size?

The data from my dissertation is on a Zip disk. I didn't mean for it to get lost to the world, but, you know...

I'm on a project now where provision of the raw data to the granting agency for public availability is a requirement. But the metadata, database structuring, and answering questions from people who take my data and have a question about it has added noticeably to my workload.

"..answering questions from people who take my data and have a question about it has added noticeably to my workload."

I know this is an imposition under current incentive structures but it seems to me on a macro level that this is exactly what we the public would want to see occurring, people actively reading other people's research and data and unrelated scientists asking questions of each other and reviewing each others findings.

Reply on news.ycombinator.com