Solal Pirelli
@solalpirelli
Scala compiler engineer @ EPFL. Scientific integrity, and random rants. Previously: PhD in automated software verification.
Today's dinner, crushed unicorns. Or is it cold beetroot soup? You'll never know.
Today in beautiful figures appearing in definitely totally scientific papers from entirely legitimate and unproblematic ACM conferences.
Sometimes it also leads to nominative determinism. Who better to support your posterior?
Will Trump doom the Swiss far right's latest initiative? We might have to thank Donald like so many other countries did... ("Isolating ourselves from Europe? Definitely not right now!")
Reaction from @reeserichardson.bsky.social, me, and @abalkina.bsky.social to IEEE's handling of conferences. (Pic by @elisabethbik.bsky.social)
Next, connecting evidence-informed policy and research integrity in France. Created a directed acyclic graph between events. Very flat: direct determinants, chains of causation are hard to find. But usually they would be more actionable (causes, not symptoms)
Next talk is on AI-assisted integrity checks. Existing checks already flag a nontrivial number of trials, some retracted ones are influential enough to change pooled conclusions. Checklist at inspect.sr
Session 6.D of #wcri2026 starts with a talk on paper mills in China. Formal businesses with contracts, clients are often clinicians who must publish but have no realistic way to do this.
Finally, a talk that tries to flip the script: open science as a requirement for research security.
Now for a talk about knowledge security in Germany. I couldn't make up "a German uses a world map that centers Russia in 2026" if I tried.
If you needed a Venn diagram specifically for your bingo card, don't worry.
As was agreed yesterday in the session about sleuths and publishers, what we really need is more policy recommendations. (Not.)
Everything could be a security issue. (That does not seem like a very productive way to look at it)
How to fix this in one slide, she jokes. Reward accuracy, not speed. It'll slow down science but make it better.
"the rat penis, always good to see" (comment from the audience, name redacted to protect the innocent :p)
10 years later, over 40% of the papers she flagged in 2015 are still not corrected or retracted...
Big Pharma has strong incentives to reproduce research, but even they often can't do it. Easy to ignore outliers in hopes to get a Nature paper that changes your career compared to hitting the brakes and looking at what's going on.
And now, PubPeer! I wonder what % of comments are represented in the first couple of rows in this audience, many recognizable commenters around 👀
Openness, eg open data, open code, etc can help fix the system and its broken incentives.