Otis Kim: Iris, hey — quick question before we start: do you think the food you were told was healthy when you were a kid was actually healthy?
Iris Holm: No. Definitively, no.
Otis Kim: Same. And that's actually not a random question — that's the episode. In 1967, the sugar industry paid researchers to publish findings blaming fat for heart disease. Not sugar. Fat. And those findings shaped dietary guidelines for a generation. A policy-level intervention funded by the industry it was supposed to be studying.
Iris Holm: And the tobacco industry did the same thing. Funded studies, bent results toward sponsor interests, suppressed findings on health harms. Two industries, same mechanism.
Otis Kim: Right — and here's what I actually find more disturbing than the scandal itself: we know about it. It's documented. The food industry's systematic influence on nutrition research design and conclusions — that's not a secret. And the system keeps running. So today I want to understand why.
Iris Holm: The question assumes the system is broken. I'd push on that. The system is producing exactly what it's designed to produce — research that serves whoever is funding it. Funding bias isn't a bug they failed to patch.
Otis Kim: Designed by whom?
Iris Holm: That's the load-bearing question. Research agenda-setting — the whole process of defining what counts as fundable, what questions get asked at scale — that power sits with funders. NIH, NSF on the government side. The food industry, the tobacco industry on the private side. They all shape the agenda. They just have different interests.
Otis Kim: I want to push back gently — because NIH being shaped by congressional priorities feels categorically different from the sugar industry literally paying for results.
Iris Holm: Maybe. Or maybe the distinction is about intent, not mechanism. Sponsorship bias — the documented tendency for studies to support their financial sponsors — operates at a population level. Across biomedical, nutrition, nanotechnology fields. It's not about individual corruption. The structure produces the output.
Otis Kim: Which means even government funders like NSF — their topic priorities, the billions they allocate through grant competitions — that's still agenda-setting. Still a kind of bias, just pointed at different targets.
Iris Holm: Exactly. So the question hanging is whether all funders are equally guilty — or whether some biases are actually worse. And I'm not sure the answer is obvious.
Otis Kim: Okay but that framing — 'all biases, different targets' — I want to actually test that. Think of it like a restaurant where three different people can pick the menu. Government orders what's politically popular, industry orders what sells, a billionaire orders what they personally find interesting. The food is real. But the menu was never yours.
Iris Holm: That's the mechanism, yeah. And the menu problem started the moment we institutionalized the funding. NSF was created in 1950, NIH scaled up post-WWII — and the second you put competitive grants at the center, scientists become structurally dependent on external validation just to work.
Otis Kim: Dependent from day one.
Iris Holm: From the design. And it's not just the U.S. KAKENHI in Japan, UKRI in the UK — cross-national comparisons show every national system embeds structurally different priority frameworks. The bias isn't a flaw in one broken country's setup. It's a feature of how competitive grant systems work, full stop.
Otis Kim: Wait — UKRI is interesting because that's where the impact agenda actually got codified, right? The Research Excellence Framework in 2014 formally required researchers to prove non-academic impact. That's not a nudge. That's a structural requirement.
Iris Holm: Right. And researchers pivoted. Not because they were corrupt — because the grant-writing incentives changed. The REF didn't break science. It redirected it. The system worked exactly as designed.
Otis Kim: Which is somehow more unsettling.
Iris Holm: Now here's the number that reframes all of it — federal R&D funding was roughly 70% of total U.S. research spending in the 1960s. Today it's approximately 21%. That gap didn't close. It filled. Philanthropy stepped in.
Otis Kim: Okay, 70 to 21 — that's not a trend, that's an inversion. And philanthropy filling that gap... I mean, actually, is the gap even the same shape? Because filling a gap assumes what replaced it fits what was there. But a foundation's agenda answers to one person's read on what matters. That's not the same shape at all.
Iris Holm: It's not. And the irony that lands hardest — blue-skies research, curiosity-driven work, the antidote to the impact agenda — it's become a fashionable philanthropic cause. So even the cure is subject to philanthropic fashion.
Otis Kim: Wait, so we traded one biased funder for three different biases that don't even coordinate with each other.
Iris Holm: Three biases minimum. And the NIH menu is still partly political — the Cancer Moonshot is a concrete example of a presidential initiative overriding purely scientific priority-setting. So even the government slice isn't neutral. It's just accountable to a different principal.
Otis Kim: So the question isn't how do we remove bias from the menu. It's — which biases do we actually prefer, and who gets to pick.
Iris Holm: Which biases we prefer — but that framing lets everyone off the hook. And I want to push on that. Because the uncomfortable version is: the bias doesn't require anyone to choose. It runs automatically.
Otis Kim: Okay, but that's exactly the weak spot I want to press. You're saying architecture, not people. Fine. But every reform we actually implement — conflict-of-interest disclosure forms, integrity training, researcher ethics codes — those are all aimed at people. If the structural diagnosis is right, we've been aiming at the wrong target for decades.
Iris Holm: Yes. That's precisely the problem.
Otis Kim: Wait — so you're saying the policymakers are wrong, not the diagnosis.
Iris Holm: The PLOS One nanotechnology analysis is the evidence. Large-scale review of funding acknowledgments across an entire scientific field — sponsorship source predicts publication impact at population scale. Not one lab, not one bad actor. Across the whole field. No individual researcher had to do anything wrong. The incentive architecture produced the pattern.
Otis Kim: Conflict-of-interest rules — even well-enforced ones — don't touch that, because it's not about hidden misconduct. It's baked into what gets built.
Iris Holm: And here's the part that actually reverses how I thought about this — industry-funded nanotech papers didn't get cited less. They got cited more. Bias as amplification, not suppression. Disclosure doesn't help when the problem isn't hidden. It's just louder.
Otis Kim: Hold on. Cited more?
Iris Holm: More. Which means the suppression narrative — the one that motivates disclosure requirements — is actually wrong in that field. The industry-backed work got amplified. It got attention, infrastructure built around it. That's a different problem than a hidden finding buried in a drawer.
Otis Kim: Yeah and — wait, that actually makes disclosure worse than useless, right? Because you disclose the conflict, the paper still gets the citations, the field still warps around it. You've just put a label on the thing you're not stopping.
Iris Holm: That's the gap. And it's why the fix has to be institutional — not individual. Though what that actually looks like at scale, the peer review piece, the concentration problem — those are the hard parts, and we should get into it.
Otis Kim: I mean, actually, is there a working model? Because DARPA tolerates risk and longer time horizons in a way NSF simply doesn't — but that's one agency, defense-adjacent, not a template anyone's scaled.
Iris Holm: DARPA is the exception that makes the rule visible. And the REF 2014 is the proof-of-concept in the other direction — a policy mechanism that measurably redirected researcher behavior within one evaluation cycle. Both cases say the same thing: the lever is institutional design. Pull it differently and you get different science.
Otis Kim: Which means the Johns Hopkins postdoc scenario — that's not a moral failing. That's just the system running.
Iris Holm: Put it precisely: she has two project ideas. One is the question she thinks matters most scientifically. The other is the one she thinks will get funded. The grant portal is open. Which one does she write?
Otis Kim: She writes the one that gets funded. Obviously.
Iris Holm: Every single time. Not because she's compromised — because it's rational. The REF impact requirement, the national priority alignment, the measurable outcomes box she has to fill — those aren't suggestions. They're the gate. Blue-skies work, curiosity-driven questions with unpredictable downstream value? You literally cannot pre-specify impact for those. The form rejects them structurally.
Otis Kim: Wait — not even hard to get funded, actually impossible to frame correctly on the form?
Iris Holm: The REF 2014 requirement was non-academic societal impact. You have to name it in the proposal. A genuinely open-ended curiosity project — what's the impact statement? 'I don't know yet, it might matter in forty years'? That fails the rubric before a reviewer reads a word.
Otis Kim: Okay but — actually, no — is that the NSF astronomy problem showing up in a different coat? Because the topic investment finding wasn't about bad proposals getting rejected. It was that funding concentration flowed to already-funded topics, independent of scientific output quality. The postdoc isn't just choosing what to study. She's choosing which pile to join.
Iris Holm: That's the compounding effect. Peer review was supposed to correct for that — expert panels, scientific gold standard. But peer review reinforces concentration. Reviewers are experts in funded topics. The well-funded areas have more papers, more citations, more infrastructure. It looks like merit. It's sedimented priority.
Otis Kim: So the fix — the democratic allocation model, equal base grants, anonymous peer donation — does that actually break the cycle or does it just... I mean, you're still asking researchers to vote, and researchers already live inside the concentration.
Iris Holm: It decentralizes the gatekeeping. That's real. But social bias doesn't evaporate when you remove the central committee. It reproduces through the network. Who you know, whose work you've read — the donation goes to the familiar. Different form, similar shape.
Otis Kim: So the REF is the clearest case we have of a government system mechanically producing the distortion — not industry corruption, not a billionaire's hobby. A government made scientists optimize for the wrong thing and the science moved.
Iris Holm: That's exactly it. And DARPA proves the lever works in the other direction — longer horizons, higher risk tolerance, the model exists. It just hasn't scaled because scaling it means NIH and NSF voluntarily redistributing agenda power. That's the mechanism nobody has.
Otis Kim: Yeah — and the postdoc at Johns Hopkins is still at her desk, writing the fundable proposal. The architecture doesn't need anyone to be villainous. It just needs the form to stay open.
Iris Holm: And the form stays open because nobody has the incentive to close it differently. That's what I come back to. Federal R&D at 21% — that number is the whole argument. Because when government held 70%, there was at least a nominal public interest mandate. Flawed, political, yes. But accountable to something broader than one donor's conviction.
Otis Kim: Yeah and — I mean, that's the part I can't quite get past. The large-scale research linking publicly funded science to downstream public uses — patents, policy documents, Wikipedia, news coverage — that analysis exists. We can measure it. Publicly funded science gets consumed broadly. And funding allocation still doesn't correspond to that diffuse public value. The government system was already failing its own stated rationale. So the question is: if philanthropy fills the rest of the gap, and the questions that go unfunded are just the ones with no wealthy sponsor, no national-priority frame, no fashionable foundation narrative... is that actually worse than before? Or are we mourning a public interest mandate that was mostly rhetorical anyway?
Iris Holm: I don't have a clean answer to that.
Otis Kim: Neither do I. And I think that's actually — I mean, that might be the honest place to stop. Not 'the system is broken and here's the fix.' Just: somebody is always the curator. We used to know roughly who. Now it's diffuse, and quieter, and the questions that don't fit anyone's agenda just don't get asked. Who decides that's acceptable? I genuinely don't know who that question even goes to.
Iris Holm: It doesn't go anywhere. That's the problem. Appreciate you pressing on it.
Otis Kim: Good conversation. Genuinely.
Iris Holm: The postdoc is still at her desk.
