unSILOed with Greg LaBlanc

The episode contains substantive material on quantitative disinformation, research integrity failures, and the mechanics of how false claims propagate through institutions. However, significant portions consist of conversational throat-clearing, personal anecdotes about gambling, and repetitive explanations of concepts (e.g., correlation vs. causation explained multiple times). The core insights - on tribalism in research, diffuse accountability, and the gap between primary research and policy - are valuable but not densely packed.

“it's not conspiracy, it's not Incompetence. It's something else... It's some kind of tribalism. It's something that you want something to be true. You have to say it's true. To be a member of the tribe”

“there is this compression to the abstract of the article and then this expansion in the journalistic press to the point where what actually gets out there to form policy, to decide legal cases for people to vote on has no relation to what the researchers know”

Brown presents a useful framework - tribalism rather than conspiracy or incompetence - for understanding research failures, which is somewhat fresher than standard critiques of 'publish or perish.' However, the core diagnosis (citation gaming, p-hacking, conflicts between researchers' incentives and truth) is well-established in meta-science discourse. The specific examples (Chinatown bus study, eviction moratorium) are concrete but not novel observations. The connection to broader institutional dynamics is insightful but lacks counterintuitive depth.

“It's something that you want something to be true. You have to say it's true. To be a member of the tribe, the tribe needs it to be true”

“You got to be able to come in and say, okay, everybody else looked at this. Everybody else did a regression. But regression is just the wrong tool”

Aaron Brown is a genuinely credentialed practitioner: former Chief Risk Officer at AQR (a major hedge fund), published author across multiple domains (poker, finance, quantitative methods), and long-time professor at top institutions (Columbia, NYU, now New Mexico State). He brings real operational experience in quantitative finance and has done the hard work of attempting to replicate flawed studies. This is not a career podcast guest; he has authentic depth in risk management and empirical rigor.

“chief risk officer at aqr, the well known hedge fund... also a teacher of finance and quantitative methods at a bunch of different schools including Columbia and New York”

“I spend a lot of time trying to replicate studies, and I often have to get in touch with the authors”

Brown provides several named case studies (NTSB Chinatown bus study, Duke eviction moratorium research, Lancet USAID study, Channel Tunnel fire study, marijuana-cardiovascular study) with concrete details about the errors (data exclusions, statistical impossibilities, timeline mismatches). He cites specific numbers (40% of COVID deaths claimed to be from evictions, fire occurring every 2-3 years vs. predicted 1 in 800 years). However, he often describes problems at a high level without exhaustive methodological detail, and some claims rest on logical inference rather than published corrections.

“they take a whole bunch of Greyhound accidents, they call them curbside accidents, and then they say, hey, curbside accidents cause seven times as many deaths as traditional carriers”

“40% of COVID deaths were caused by evictions. Which you start doing the numbers, and that says, okay, every single evicted person must be dying of COVID”

LeBlanc asks solid opening questions and occasionally pushes back (e.g., 'why do people do this?', 'is it just a free rider problem?'), demonstrating he's thought about the material. However, much of the conversation follows a predictable 'guest explains problem, host nods and asks next question' pattern. LeBlanc rarely challenges Brown's claims directly, seldom probes for nuance or limitations, and allows several tangential discussions (gambling history, venture capital) to consume time without sharpening the central argument. The discussion of Bayesianism at the end is particularly underdeveloped.

“Well look, I mean it's clear there's a marketplace for information, a marketplace for studies, right, and reports”

“Is this just a free rider problem?”