Five maps so far. Is someone doing this by hand, the hard way? I figured it was an AI someone programmed, but if it's an individual or small team, big respect. Very neat project.
this post was submitted on 26 Oct 2025
223 points (99.6% liked)
science
22258 readers
187 users here now
A community to post scientific articles, news, and civil discussion.
rule #1: be kind
founded 2 years ago
MODERATORS
They provide links to their github that explains their whole methodology. This is a scientific effort and is as transparent and well-documented as a project can be. They provide the code so you can understand the exact mechanics at play or just fork the project if you want to take the work in a different direction.
It’s a great project and long overdue. I personally think scientific journals are incredibly outdated and haven’t been necessary for a couple of decades. Just put your work on a stable web site and cut out the parasites at the journals.
I hope they are able to grow this without compromising the quality or vision, because I'm sure they'll have lots of people willing to get involved, but maybe not all for the most ethical reasons.
github links show 2x contributers. cool project
AI would probably be pretty useful for this. You'd have to assume most of the "answers" are in the abstract, so you could just build one to scrape academic texts. Use an RAG so it doesn't hallucinate, maybe. Idk if that violates some T&C nonsense that doing it by hand doesn't though.
This is a bad idea. It's extremely likely to hallucinate at one point or another no matter how many tools you equip it with, and humans will eventually miss some fully made up citation or completely misrepresented conclusion.
Google RAG
There are tons of AIs that are not auto regressive LLMs
I'm a professional software engineer and I've used RAG. It doesn't prevent all hallucinations. Nothing can. The "hallucinations" are a fundamental part of the LLM architecture.
Are the down votes because people genuinely think this is an incorrect answer, or because they dislike anything remotely pro-AI?
Both probably. Thought terminating cliches and all that. The most useful tool maybe ever. Wild.
I use LLMs daily as a professional software engineer. I didn't downvote you and I'm not disengaging my thinking here. RAGs don't solve everything, and it's better not to sacrifice scientific credibility to the altar of convenience.
It's always been easier to lie quickly than to dig for the truth. AIs are not consistent, regardless of the additional appendages you give them. They have no internal consistency by their very nature.
What would the failure rate on this be? What would the rate have to be to actually matter? Literally it would just poll the abstract and spit out yes no undecided. That is in the abstract. There is very little chance of there being any hallucinations that are meaningful at a degree large enough to vary literally anything.
Have you never had it organize things or analyze sentiments? I understand if that's not your use case but this is pretty fundamentally an easy application of AI.
And this isn’t even really a great application for RAG. Papermaps just goes off of references and citations. Perhaps a sentiment analysis would be marginally useful, but since you need a human to verify all LLM outputs it would be a dubious time savings.
The system scores review papers very favorably and the “yes/no/maybe” conclusion is right in the abstract, usually the last sentence or two of it. This is not a prime candidate for any LLM, it’s simple database operations on srtuctured data that already exists. There’s no use case here.
Perhaps a sentiment analysis would be marginally useful, but since you need a human to verify all LLM outputs it would be a dubious time savings.
Thank you, yes. That's exactly my point. You'd need a human to verify all of the outputs anyways, and these are literally machines that exclusively make text that humans find believable, so you're likely adding to the problem of humans messing stuff up moreso than speeding anything up. Being wrong fast has always been easy, so it's no help here.
It's cool that shows all the papers and not just some abstract metric or yes or no answer.
it's still only five topics and you really just have to trust the devs that info is accurate and not biased.
They provide direct quotes from the papers that support their scoring and also direct links to the full papers.
It’s super easy to just check their conclusions. I followed up on several papers yes and no on the vax question. There was no skullduggery as every paper I looked at was represented fairly in the scoring.
As in other scientific efforts, this is not just a ‘trust me, bro’ situation. They provide references.
Not what I really meant. I was after that one has to trust them to actually provide a suitable and representative coverage on all the papers released on the subject.
I see, thanks for clarifying.
I think that concern is partly covered by their scoring. If a bad-faith actor put together a distorted gathering of papers that favored their conclusions but weren’t cited widely, those papers would have very small circles.
So it would be visually apparent that either: they were being dishonest in their research gathering, or the question has not yet been studied widely enough for this tool to be useful.
The more I think about this the more I love this project and their way of displaying the state of consensus on a question.
Something I've seen on some PubMed meta-analyses is the inclusion of the various search terms and inclusion/exclusion criteria used; something along those lines maybe?
Please add a section about nature! Global warming, deforestation, and other human effects on nature.
You can suggest new maps. They ask for links to papers, so if this is a thing you are passionate about and have some recent papers, especially review papers. Reviews seem to get more points in their schemes.
I love this project too and have a personal passion in neurobiology studies related to benefits of yoga. When I have a couple of hours, I will submit a map suggestion for that topic.
What does the circle size indicate?
From the docs in GitHub: "The size of the dots corresponds to the number of reviewed papers for literature reviews (non-reviews have the smallest size)..."
I wondered the same. It doesn't seem to correlate to P-Size, citations, or participants. Maybe a combined factor of each that is calculated? I'm really not sure.
Couldn't quite work that out either. I initially thought it might have been to do with the number of citations but that didn't pan out
I think it might be the number of papers used to answer the question. Not totally certain though