Folks who make garbage can regressions tend to be very concerned about p-hacking.

This is a fantastic article about James Watson but mostly it’s master class in dispatching of ghouls in science. Often the problem isn’t just that they’re amoral but that their being a(im)moral makes them dangerously wrong in ways that have cascading externalities on our understanding.

"What James Watson got wrong about DNA" By the great Sohini Ramachandran (@sramach.bsky.social) and your boy for The Boston Globe (@bostonglobe.com). www.bostonglobe.com/2025/11/14/o...

Post Malört ergo propter malört

What usually gets lost in the focus on what an evil racist James Watson was is just how colossal of a dumbass he was and just how far he set the field back. But @sramach.bsky.social and @cbo.bsky.social kept their eye on the ball.

A corollary of this is that something can be normally distributed in the parameters but not necessarily in the data. Using the right link functions and model, you’d be surprised how much the normal distribution can do.

The Hitler genetic work is like an over the top hypothetical for the ethics of using onymous dna to diagnose post genetic conditions of deceased individuals.

Daily dose of Fred

If only she’d left a diary detailing her thoughts and experiences.

Little guy got a clean bill of health.

WOW a loggerhead shrike without intense heat distortion, what afind!

And what will Mexico feel like, shithead? www.gatesnotes.com/home/home-pa...

The Biden administration worked hard to accomplish this new free IRS filing system. It’s astounding that one should have to pay fees in order to pay their taxes. But this rent seeking is entirely consistent with the economic policies of the current administration that benefit the most wealthy.

It’s ok, he’ll just fund folks to say the election was p-hacked.

sorry but mamdani and cuomo should have worn blue and yellow hats, respectively

Giving at seminar at Princeton CITP today at 12:15 est. It’s live-streamed to feel free to tune in! spia.princeton.edu/events/citp-...

This paper claims to provide evidence offline networks matter more than online for voting preferences, but winds up mistaking noise for signal and seems to forget that some folks online are quite influential. academic.oup.com/pnasnexus/ar...

Best World Series ever. This is insane.

Meet Fred! You’ll be seeing more of him….

It’s more pernicious than this because you’re also testing if your model is a perfectly unbiased representation of the data generating process.

A helpful heuristic for me is to think of science a bit like data chain of custody. An llm synthesizing data, annotating it or summarizing it breaks the chain from, say, humans to your final statistical inference or graph.

Y’all we gotta get better at regression. www.nature.com/articles/s41...

Assessing impact in this way and allocating funding is pretty: “yo dog I heard you like cumulative advantage in science so I added cumulative advantage to cumulative advantage”

the tech industry is a threat to scientific knowledge, academic freedom and evidence-driven industry regulation. the industry systematically captures and/or interferes with and undermines independent knowledge production processes that challenge it arxiv.org/pdf/2510.19894

Similar themes to a paper I gave at UW in April depts.washington.edu/societytech/...

I was pleased to see @olivia.science and @irisvanrooij.bsky.social mentioned in this NYT article (gift link below). But I wish there was more analysis. Boasts and critiques re: #GenAI were strewn throughout, requiring readers to sort the concerns from the hyperbole. For example… (1/2)

The response to this has been fascinating... plenty of public engagement from outside of the field and lots of private messages from people within it. It's somewhat telling that there appears to be a "chilling effect" for pointing out concerns over industry funding in the field.

This recent preprint on tech industry influence over external research is compelling. Beyond documenting the low rate of industry-funded researchers actually disclosing that fact, the paper points out that journals have routinely waived IRB review for industry work. arxiv.org/abs/2510.19894

Let’s not be naive: We need to be very aware of industry influence in computational social science! Through control of data (and experiment) access the situation is even more problematic than in other industries, which makes the data access rights under DSA article 40 particular important!

Who today is influencing science a la Big Tobacco or Oil and Gas? In our new preprint we show how tech companies like Meta are capturing research on their product, using mechanisms that subtly (or not so subtly) shape what science gets produced

1. We ( @jbakcoleman.bsky.social, @cailinmeister.bsky.social, @jevinwest.bsky.social, and I) have a new preprint up on the arXiv. There we explore how social media companies and other online information technology firms are able to manipulate scientific research about the effects of their products.

👀

seems like this should be a bigger deal than it currently is: "Currently, industry-led experiments are publishable without [IRB] approval [39]. There is little reason to exempt industry from standard journal policies..."

Going to go ahead and put this at the top of my reading list.

"Despite the academy’s best efforts, it is clear that most of the research findings about harmful impacts from tech remain hidden as trade secrets. To protect the public, policies that bring this information to the light of day are needed."

We should start having this conversation explicitly in criminology. These distortion mechanisms and perverse incentives are absolutely relevant for crime data. inquest.org/our-evidence...

I couldn't be more on board with this. Focusing on corruption/fraud/misconduct at the level of individual scientist appears to be misguided and counterproductive when systemic forces and selection biases are at play. This appears to be true even beyond the context of industry manipulation.

Important work

I'm excited to finally have a preprint of this paper up, a few years in the making. In it we argue that industry-driven manipulation of social media research is well underway and that norms and institutions in the field are ill-prepared to resist tech's influence. arxiv.org/abs/2510.19894

When there is a random way to do something, there is a less random way that is better but requires more thought. In this case, regression models that make no sense don't belong in a multiverse analysis. An inferential regression without a causal justification is like an opinion without reasons.

I’m drinking the gorgeous coffee and it occurred to me that my grinder is a decent metaphor for multiverse analyses and theory. It has a bunch of letters and numbers, which we’ll pretend correspond to predictors we can include in a model. Coffee are the data. Hypothesis: the coffee is good.

This is, against all odds, the worst #metascience take I've ever seen.

There is much wisdom and vision here ... "multiverse analyses require thorough, theory-based model selection. Otherwise, they become a “dangerous tool” that drowns valid models in misspecified ones, needlessly eroding trust in science." #TheoryShouldAlsoGuideModelSelection 🧪

Great piece on the absurdity of brute force multiverse analyses. www.pnas.org/doi/10.1073/...

Can AI simulations of human research participants advance cognitive science? In @cp-trendscognsci.bsky.social, @lmesseri.bsky.social & I analyze this vision. We show how “AI Surrogates” entrench practices that limit the generalizability of cognitive science while aspiring to do the opposite. 1/

When I bring up the fact that most replications are uninformative and unjustifiable by design, hence we should instead do better research, I always get the same argument back and I'd like to think through this reasoning carefully. It goes like this:

Multiverse studies seem doomed to perpetuate a cycle of "look it varies!" "but not that much" "does too!". Here's another exchange that I had missed. "Robustness is better assessed with a few thoughtful models than with billions of regressions" doi.org/10.1073/pnas...

This is why data access and independent tech research is so important. When you're reliant on the platforms they can bias the results by shifting what you have access to. techbullion.com/independent-...