You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$355,000 left to reach the funding goal.
SaneCite is an open project, and the goal is to spot and hopefully correct places where retracted papers are still being cited in scientific papers today. So far I found that about 40.7% of the citations we could read still use the retracted paper as support.
I created this project because I saw a CBS News Chicago report on researchers at Northwestern University who study this problem, and I knew I had the coding skills to do something about it. I was really surprised by how widespread the problem really is. I think the public deserves to know, and there needs to be a mechanism in place to fix this stuff.
What I want to accomplish this year is to hire two checkers who can manually review my work: people in the biomedical field, or experts in whatever field we need. They can verify that the work is sound. I've gotten pretty good responses so far from the scientists themselves, but I think having human checkers in the loop is going to be really important for credibility.
Once the human checkers verify that our process is sound, the floodgates really open, especially with funding, and we can check the entire scientific record. I think that would be really revolutionary.
I've been spending my own money and time on this, and I can't afford to hire the human checkers too. So this is what the money would go toward. The good news is that as AI improves, the cost of the AI part goes down, somewhat exponentially.
Me, full time: $100,000, plus $20,000 for health insurance and self-employment tax
Two checkers: $145,000 ($72,500 each)
Computing: $60,000
Legal: $20,000 for media liability insurance and a lawyer's review before anyone is named
Data access and hosting: $10,000
Total: $355,000
Well, I have a lot of different projects. You can see them at saneapps.com, and they've been pretty popular, some open source and some transparent code, depending on the project. So I have a background in making well-functioning apps, and I designed SaneCite from scratch, the whole program and the methodology. I wrote to every scientist on the Lists to get their feedback and thoughts before anything was published. So far the feedback has been pretty good, and I think I'm on the right track.
I want to see integrity restored to the scientific record and public trust rebuilt, in a nonpartisan way.
The human check could show the AI is a lot worse than it looks. Then the headline number goes down and I say so publicly. That still answers the question, and the check can be reused. I doubt the AIs are all hallucinating since they check each other's work but that needs to be proven via human checks.
I might not find good checkers at this pay. Then I'd hire part time or start with one person and go slower. So far it's just been me so anything more is a win in my mind.
Worst case someone named on the Lists could sue. I don't think it's likely since I stick to facts from public records, never guess at motive, and give everyone a chance to reply. I'm really careful not to do anything without giving people plenty of time to review and respond. I have had some upset people but no one has threatened to sue because of course it's all public record! I'm just organizing it clearly. I've also had some good responses as well. So far about 18% of scientists I messaged with findings (10 of 55) engage seriously with what I found. That's way better than zero. I don't see any way this is a waste of time. I've been putting my own money and time on the line with no funding so far.
No grants so far. I've paid the costs myself. I've also applied to Lightcone Commons and for compute credits from Lambda and Anthropic. Nothing has come in yet. I had Cloudflare startup credits for the AI reading, but the AI part is used up. So right now I'm paying out of pocket to do this. Need help! :) Join me in bringing integrity and trust back to science. I read somewhere that trust is at an all time low and I think public accountability and transparency using tools that were never available in the past is the way to fix things. Many scientists didn't even know the things they cited got retracted after publication. No one tells them!