Any details of that AIRA does exactly?

This tendency to confuse credibility of claims with “were the authors doing their best” is strong in psych. Eg defensiveness about Cross Lagged Panel Models despite now knowing they have false positives rates. “But they didn’t know that at the time” doesn’t increase evidential value now.

We do this in Bern! There is even cake for attendees

Chaotic evil: N(0,10) applied linear probability model.

@solomonkurz.bsky.social schooled me on this a year ago, mind was blown.

How are you out only-child-ing me

This slide unfortunately generalizes well 🥲

"Schrödinger's causal inference" (n):

The practice of making causal claims or interpretations within a scientific article - typically in the title, abstract, implications, or conclusion - while simultaneously warning that the study design is unsuitable for causal inference.

It would be a step backwards rather than forwards to have a vast and expensive integrity assessment system that was both ineffective and provided false trust.

But here too the metaphor breaks down, because the evidence that the TSA is effective is very scant. DHS red-team exercises show very high failure rates, published risk analyses questions its assumed efficacy, many critiques of 'security theatre' over actual efficacy etc.

These are all demand side arguments about what is needed, but it’s a supply side problem. Who will do these reviews, who has the skills, time and interest to spend 3x the amount of time they’re already not being paid or rewarded to do in their career? There are fewer levers to pull on here.

My current interest is in making post publication more feasible, rapid, and effective. Concerns that would kill a manuscript during peer review are currently ignored after publication, when the burden of proof switches to critics needing to prove beyond a doubt there are issues, for no good reason.

This is where the analogy breaks down for me, apart from my discomfort with comparing it with serious crime: police don’t make specific efforts to prevent murder, they are investigated after the fact, which is at odds with the TSA analogy.

This is the primary tension. Do we actually endorse such changes as feasible and worth pursing? I don’t (currently) advocate for universal pre publication integrity checks because I consider them unfeasible to implement.

Maybe pre-publication peer review can be strengthened, but the review system is already under strain and this has huge scalability issues. Strengthening post publication review, eg through targeted citation-triggered reviews, is another option. Eg: mmmdata.io/posts/2025/0...

I think you make a very useful and honest point when you say “we rely on the good faith of authors and their institutions”. Trust requires clarity about what is or isn’t checked, and journals by and large currently have no immune system to protect against fraud.

Now, I patiently wait for @briannosek.bsky.social to do a wellness check on me for uncharacteristic optimism.