Anomalies Wanted: How to Do Continuous Science

Academic research is under serious fire right now. The suspects fueling a replication crisis include the peer-review system, academic journals, and the system of evaluating faculty for tenure. The questions are also not new. The challenges are structural, baked into the underlying incentives. There are no easy answers it seems to the challenges.My guest for this episode is Mathïs Fédérico, founder of the startup company Bycelium, which aims to rethink science with Bayesianism. Mathïs shared his personal journey through the traditional research pipeline and explained how the emphasis on publication count and narrative crafting distorts scientific progress. Our conversation explored Bycelium’s approach to measuring the credibility and impact of scientific hypotheses in real time by incentivizing the sharing of data and negative results and encouraging honest debate rather than just novel publications.As Mathis told me, “Science is never perfect. Science will never tell you that something is true or false. Science will just nudge the credibility of things thanks to evidence.”Will Bycellium work? It’s too early to say, but I find the ideas behind it intriguing and illuminating.Challenges in academic research trendsMichael HornAlrighty. Welcome to the Future of Education. I’m delighted because several months ago Mathïs Fédérico reached out to me somewhat on a whim, I think because he had seen something I had posted about the research challenges in academia and higher education. And my hypothesizing that actually, you know, a lot of my research and writing is about how we need to reinvent the teaching and learning model itself is broken. And I said, you know, there’s this whole other thing that’s also broken, which is the research model itself. And then since then he reached out and we’ll talk about why he did in a moment. But one of the things that’s happened since then is a lot more people are very dialed into the challenges that the research process has. We’ve had this Nature article coming out that said, you know, 3,900 studies published in 62 journals and half of them could not be reproduced.We’ve seen a lot of people realize, hey, actually a lot of the Nobel Prize winners are not coming from traditional higher education pathways. As of late. We have DeepMind, Google, pharma companies, a lot of researchers that are the most impressive breakthroughs aren’t coming from the universities we expect to produce the research. We have a lot of claims of not just reproducibility challenges or replication, but outright falsifying of research and the like. And then you have this other backdrop, which we know well, is that people, when they publish in research journals, increasingly to get published, not only do you have to have something that looks statistically interesting, you also have to have something that is unique. So by definition almost not replication. And that has caused many faculty to look at narrower and narrower and narrower questions that have less and less relevance to other people. And as a result I don’t have the stats in front of me.But very few peer reviewed journals, articles that get published are ever, ever read. Very few get more than say two or three citations ever out there. And so you have a bunch of challenges hitting all at once. And so my guest Mathïs Fédérico, he has started an entity called Biselium, which we’re going to talk more about to actually solve the root causes of this challenge. But first, Mathïs, welcome. Thanks for reaching out. I’ve been like, we had this conversation on the phone and I’ve been buzzing ever since and saying, I think we have someone who might have an answer actually. So first, welcome and thanks for being here.Mathïs FédéricoWell, thanks for having me. I’m very glad to be here, I guess. Yeah. All that you’ve said, which seems very diverse in a way, and seems like a lot of things are happening independently, but all linked to the same source, which is, we are not pushing for the right thing. The thing that academia is optimizing for is not the good objective. And so whenever you try to maximize an objective like this, well, what’s happening is you maximize the wrong thing. You have a lot of the wrong things. And for us the wrong thing is the number of publications.That’s what we maximize today. The number of publications is what we use to evaluate. It’s what we use to quantify breakthroughs, to quantify how good a researcher is. But it’s not what matters. It’s not what matters at all. What does matter though, is how much did you change the minds of others, what