DAOs in Healthcare

Yeah, absolutely. Decentralised science is an extension of open science. Open science is the idea that the outputs of a research project should be available to everybody — the whole paper should be readable by everybody, and you should have access to the primary data. That's open science.

Decentralised science is one layer built on top of that, which says: okay, now not only can you read, but you can also write and execute. You can raise funding for your scientific project online, and everybody can do that. And you can take that funding and deploy it online as well — so you can pay a colleague to help you on a research project, or pay a laboratory to run a particular experiment for you and report the data back. So it's really an extension of open science.

A DAO stands for decentralised autonomous organisation. It's just the native form in which people can organise on-chain, on the internet. We all know WhatsApp group chats, we all know Facebook group chats — a DAO is literally just that, with everybody in the group chat having control over a shared crypto bank account.

Think about that. If you had a Facebook group chat and it had an actual bank account, it would be really hard to make sure the interests of all the different parties in that chat were reflected in how the money was spent. Now, with these primitives, it's relatively achievable to make sure the community can make decisions and the funds are actually spent in a way that reflects the interest of the group. So a DAO is like a WhatsApp group chat with a bank account, plus some mechanism by which the interests of the whole group can be put forward — not just the interests of a few leaders.

One of the first ones we have to solve is basically building small organisational units for scientists to collaborate. What we're seeing right now — and this is happening within our community — is a scientist with a paper that's almost finished, but there's just one more analysis they want to do, and they don't have the capacity within their direct ecosystem. So they post about it. And suddenly another scientist comes along and says, "Hey, I can analyse your immunofluorescent images," or "I can run an in-silico docking to check whether your small molecule is binding a protein." They get into the computer and they work together.

And yes, they need a shared bank account, because when a lot of scientists come together, the time they put in costs money — someone needs to pay for it. So ideally, over time, we'll be able to raise nonprofit grant funding into these communities, so they can have these virtual laboratories where they work together. The IP that's generated is captured within those labs, and everybody's contributions are tracked through token-based mechanisms. So instead of an author list that's just one array, you have a whole network graph of token balances — who has how much ownership in the project, and how they interacted with each other.

That's currently what I'm focused on — making scientists operate really well together. Other DAOs emerging in this space are funding DAOs. You have a group of people who really care about research into a particular disease. They come together online, pool their funding, maybe raise some more, and then use that funding to decide which scientists to sponsor and how to allocate capital. And I think that's just scratching the surface.

Another example — we're not there yet, but I think we'll get there soon — is patient-interest DAOs, where patients with a rare disease come together and say, "We want to sponsor the research into our own disease, or into the disease our loved ones have." There are companies supporting the emergence of these groups right now.

And then the last step in that evolutionary tree could be data commons. Every patient has a record, has data they have access to. There have been projects like Count Me In, run by the Broad Institute, where patients would ask their local pathologist to send a sample to the Broad, and the Broad would sequence the DNA and do the analysis. That project was quite a success — a whole active online community, one institute doing the analysis, and the data shared globally. You could have something similar, but instead of one institute, it's an online group of patients. They all ask their practitioners for their healthcare data, pool it together, govern that pool collectively, and then work with scientists to analyse it — potentially with some ownership in the results, potentially giving access to for-profit entities in return for capital, and distributing that among the patients who brought their data together. I think that's going to be quite a common pattern in the coming years: patients pool their data so it's valuable, find ways to control it, and also get some financial return from it.

Yes. That's something you learn the very hard way when you explore this space: it is not all free-flowing and structureless. In fact, it only works if there's a certain amount of structure, and if you have something like a leadership structure. A lot of what we've learned about how humans interact over the last thousands of years is still valid in these online groups. You want stewards who have more investment in the group's success, more subject-matter expertise, who can guide the community and its decision-making. Groups that don't have that very quickly figure out they need to set up some structure — or they become very inactive, very quickly.

Right. These decision-making tools are still being actively developed and are still far from perfect. The most common tool people use for any type of voting is token-based voting: everybody who owns a governance token for that DAO has a vote. But that also means a person with a lot of tokens has a lot of votes, and someone who just joined with one token has only one. So it's not very democratic — it's one token, one vote — and a lot of people in the field have criticised that for the right reasons, because you get these imbalances of power.

Other methods being explored are one person, one vote, which strikes us as the most democratic — but it's surprisingly hard to implement. When you're online, what's stopping you from creating 20 accounts and having 20 votes? That attack is called a Sybil attack: you create multiple accounts and impersonate multiple participants. How do you make sure there's really just one person, and not one person controlling 20? More and more tools are being developed to solve that, and I'm pretty optimistic we'll get to a point where we can have these more democratic systems.

First, I have to say a lot of what I'm about to describe is theory that hasn't yet been pressure-tested. At VitaDAO we've done a set of these agreements, and we've minted a set of these IP-NFTs, but we haven't yet seen a situation where, for example, a biotech company emerged around that IP and we had to really think through how you create a returning cash flow out of it. So that's the first thing to put out there in all fairness.

The way it works in theory: you have a group of researchers doing work, either within a traditional organisation or outside one. When they eventually make a discovery — let's take the traditional-institution case — you would previously have done a sponsored research agreement with the university's tech transfer office when you funded them, which lets you capture some of the IP and get an exclusive licence. That exclusive licence sits with a company, a service-provider company that made the agreement with the university. Universities don't usually do legal agreements with online group chats — we're not there yet, and that's not happening any time soon. But there are companies willing to do agreements with both online group chats and university tech transfer offices. Companies like Molecule.

So Molecule makes that agreement and holds the exclusive licence to that piece of research. Then — and this is part of the agreement — it's a transferable sub-licence. They mint an NFT and say: whoever owns this NFT holds the transferable licence to that exclusive licence. That's basically the whole magic. There's one real-world entity holding the actual contract with the university, and they emit a token that represents that ownership.

It's very similar to what's happening with token-based carbon credits, where one company holds the actual credits issued by traditional organisations and emits a token representing that physical document. With these IP-NFTs you then have something more marketable, that can be purchased by other people. And eventually the hope is that enough information accumulates that there's a strong case to launch a company around the IP. At that point the token would be burned, the actual licence transferred into that company, and people would work with the company to turn the technology into a marketable product.

Now, how do you give the people who did the research participation? This is where we get into research happening outside institutions — which is what I care a lot about with LabDAO. If we provide the laboratory capacity for scientists to work online, a lot of science can happen outside existing institutions. In that context, one way we're thinking about it is to track contributions within the lab through peer-to-peer feedback — every month, people give each other feedback on how much they contributed to the project's success. And then if there's IP, we capture it using the existing legal infrastructure: you file a patent, the DAO has a legal representation, and that entity files the patent for that group of scientists.

It could be a dedicated entity set up to represent that particular group of scientists. And the people who really want to follow that journey further could be the founding team of a company that emerges out of it. Or if they're not interested in building a company and just want to keep doing science — that's great too, and then we find a way to give them some kind of inventor revenue-share agreement, the same way institutions do with their inventors.

No, that's not how it works — that would be a security, right? The token is purely a governance token. There's no dividend that's ever going to be paid out. If you contribute to a funding DAO, you're committing capital and in return you get governance rights, and something you could most accurately describe as an endowment. Now there's a pool of funds this online group can allocate to enable science. And yes, that group will capture some IP, but then it'll probably sell that IP into a company that develops out of the pool, or to another funding organisation. That will be a liquidity event for the DAO — but it will not be a liquidity event for you.

You'll likely see one particular set of scenarios, though. There's actually a piece written about this called Hyperstructures — we can share it in the show notes — where people have been thinking about how value accrues if there's no cash flow back to the people who hold governance tokens. The model I tend to use: if you have something that starts as a relatively small community but matures into an institution — really that's what it is, an institution as large as a research university you'd know from the UK or the US — then control over that institution has a monetary value larger than control over a smaller project that just started. So I can't make projections about future value, but if you have functioning research institutions and funding organisations, the value will appreciate. And if you need 51% of the tokens to control the direction of those institutions, the token value will probably correlate with that. But again, I can't make any clear statements there.

Yeah, basically that's something you'd want to do primarily out of the goodness of your heart, because you really care about that research. There might be a secondary motive — you say, "I believed in this science funding organisation relatively early," and if it makes good funding decisions, the value of the organisation increases. But there's no future cash flow coming back to you, because you didn't purchase a security.

Now, there are other cases. Popular examples are The LAO and other investment DAOs, which have a bit more legal clarity. You need to be an accredited investor to participate, and there it's very clear: if you participate, you have future cash flows in case of a liquidity event. But that's a different ballgame — you need to be accredited, you basically do seed investment together, and in that case you probably won't specialise in funding scientists. You'd go into seed and pre-seed investment in biotech companies, because otherwise your risk-return profile probably isn't that stable.

I'm extremely optimistic about the potential of DAOs for rare diseases. It's probably good to bring up two companies in particular. Vibe Bio are really interested in setting up DAOs for rare diseases and patient groups. And another entrepreneur who comes to mind is Ethan Perlstein, who's been committed to rare disease research for quite some time and has been looking into decentralised science as well.

What all these groups have in common is that patients are already coming together online. There are a lot of Facebook group chats where patients share their stories, and the collective intelligence of these groups far exceeds the average training and competence a physician has — even, probably, after finishing residency — in that particular rare disease, because it's so rare. So a lot of understanding is concentrated in these online groups. But an online group chat only gets you so far. The moment it has a bank account it controls, you can actually change something about the physical world: advocating for research, allocating funding, sponsoring research at universities, potentially even capturing some IP that comes out of it. So I'm pretty optimistic about those types of DAOs, especially in rare diseases, to really make a difference.

I need to laugh, because there are multiple ways. Some people have done these setups relatively quickly. And I think there's almost a law emerging, just like the Lindy effect — something that's already been around for some time will also last longer. If you already had an online community for a while and then decide to build infrastructure where you can make funding decisions using open-source payment systems, you're in a far stronger position than if you overnight decide, "I think there should be a DAO for X," and set it up.

The best DAOs have often been around for a while and just didn't know they were DAOs. Facebook groups of patients affected by a disease are de facto patient DAOs already. All that's missing is giving each participant a vote in funding decisions, and a way to participate in the funding. The second part is the legal side, and that's very much a moving target. I spend a lot of time talking to lawyers, and so do a lot of other entrepreneurs in this space, because we want to get it right — we have a lot of responsibility, we're setting up infrastructure for science to be funded, and we want to do that responsibly. A lot of regulators haven't made a clear statement on how they treat certain organisations, so we need to be very careful and follow what we believe is honest and right. And given the current landscape, it wouldn't be wise to have a DAO that was only instantiated on-chain — you probably want a legal entity that gives you real representation, so the liability of all the members is limited, which is one of the more common concerns.

I think the jury is still out, and the history books still have to be written. Because what you described is the wisdom of the crowds — but there's also the madness of the masses, where a lot of people have group behaviour that's somewhat detached from reality.

One anecdote from VitaDAO: what these DAOs do for sure is grow your funnel. Suddenly you have thousands of eyes on the ecosystem, thousands of enthusiasts saying, "Hey, have you looked at this? Here's a scientist doing this project." It's a social network, so a lot more activity is visible, within the consciousness of the group. But taking all that visibility and turning it into good scientific funding decisions — that's the second step, and the jury is still out. We've made a set of funding decisions within VitaDAO; I think a lot of them were good, but it's still not clear.

What I can share is that the decision-making process is actually less democratic than it might sound from the outside. There's a funnel, then a review group of people who volunteer to review the funding proposals. We collect at least three reviews from independent contributors with subject-matter expertise — a grad student, a postdoc, a professor in that area. Then there's a group call every Friday with those scientists, plus people who've worked in venture capital or biotech, and we make a funding decision based on the collective intelligence. Again, we still have to see whether that's a better way to make decisions.

From a theoretical perspective, I think the internet is a tool for collective intelligence, but we need to structure it well. There are examples where building tools for collective intelligence online failed miserably, and examples where it worked. The two that come to mind are two important chess games: Kasparov versus the World, and Karpov versus the World. These were games in the late 90s and early 2000s where a chess master played against an online forum, and the forum had to decide the next move. In both cases the chess masters won, but in one case it was way more difficult than anybody expected.

In the case where it was more difficult, the structure predisposed the group to good decision-making. First, the time from information being available to a decision being made was larger, so the group had more time to think and discuss. Second, they had stewards — subject-matter experts, in this case junior and local chess masters, who moderated and guided the conversation. And third, they had discoverability of micro-expertise. You might have a person who isn't the best chess player in the world, but who knows a lot about that particular opening or that particular move. In that moment in the game, they can come in and perform on par with the chess master. If you have enough of those people and the information is discoverable, you can make extremely good decisions. So setting up the structures for these people to come together in the right way will make or break a lot of these funding DAOs.

I 100% agree. There's this Charlie Munger quote: show me the incentives and I'll show you the outcome. In academic science the incentive is to publish, so you can apply for the next grant, so you can apply for the next faculty position. Redoing someone else's work, testing it for reproducibility, is nothing on that list of financial incentives. So if you can restructure the financial incentives, you can restructure the behaviour. That's extremely powerful and needs to be done with great care. But in this case it's actually a very good idea to say: what if we set up a dedicated fund for people to donate capital towards ensuring reproducibility?

And in some cases we might actually need it. In LabDAO, one core element we're building is the lab exchange. A scientist describes one experiment they want to run but they don't have access to a physical lab — say, expression of a protein in E. coli, a relatively standard operation. They give all the parameters for that laboratory service, and a payment. Then any other lab in the world can claim that offer, do the work, express the protein and test its properties. Now, if you already have an exchange where scientists give instructions and labs take on the offers, it's relatively easy to say, "Can two labs please do this?" It's the same set of instructions — can I do parallel compute? Two labs, please. And maybe I don't only need philanthropic funding for that; maybe it's more of a chore I need to do as a DAO to make sure the quality of the marketplace is there.

And going one step further — maybe it's also something you make laboratories do when they want to join this community. You say, "If you want to join, can you please reproduce this example, so we can see whether you're able to?" You can test the labs already in your community for robustness, and use it as an onboarding test to decide whether you actually want a lab in your community of service providers. So that's a very long-winded way of answering your question, which is: yes, I think these economic systems will have a fair shot at solving some of those reproducibility problems. A, because we have new ways to engineer financial incentives, and B, because some of the economic systems we're building somewhat rely on reproducibility being baked into the way science is done. We have a real financial interest in making sure these things are reproducible.

There are definitely a lot of people thinking through alternative publishing mechanisms, and I think it's about time we rethink the publishing ecosystem. But I'd say this: the whole internet is about content creation and content curation. The content curators may change with their brand name, but there will always be strong brands of content curation, and less strong ones. Right now, scientists submit their manuscript to a publisher before publishing, and content curators with a lot of clout — Nature, Cell, Science — review the work, publish it, and run a very profitable business.

I think that will change, but not as radically as some people think. We'll move to a world where scientists raise their funding online, collaborate online, and then package their work as a bioRxiv preprint — publishing before it's peer-reviewed, so it's discoverable for everybody. And then I think they'll still rely on a curator to discover their manuscript on the preprint server and say, "Wow, this is really interesting, we should highlight it." So I don't think Nature, Cell and Science are going anywhere. They'll just have to adapt to a change where scientists publish their work online without asking permission, and then those journals will still be the curators and highlight it — and, as a function of that, still have a lot of power. But that power is somewhat limited, because science is always shared no matter what.

Right. The gatekeeping function comes from a time when the journal was in print, on paper — there was a limit to how many pages you could print, so you needed to gatekeep. But the internet has completely taken that argument out of the equation. The publishers either wake up to this now, or they'll be painfully woken up in a couple of years when scientists en masse just sidestep the gate. It's already happening with bioRxiv and the arXiv system generally. If you talk to a computer scientist or a mathematician, they don't even care about publishers much anymore — they just put their stuff on arXiv. In biology, life sciences and healthcare, it's gradually happening too. First there was bioRxiv, and now there's medRxiv as well. So even if you're an academic clinician, you can publish your clinical trial, your retrospective analysis, your cohort study on medRxiv today. You write the manuscript before you send it to the Lancet, put it on medRxiv, and everybody in your community can see it. The Lancet might still reject it, but it's already out there — everybody in the world knows who did the work.

That's going to be the new normal. You still want to get published in the Lancet, because the Lancet is still the premier content curator — but it's not the gatekeeper anymore.

Oh wow, thank you for saying that. I think it's about surrounding yourself with people who care a lot about the truth, about figuring things out, rather than giving clear answers. I'm always afraid of people who have very clear, very strong answers. I try to surround myself with people who are comfortable saying "I don't know." In science, it's great — you can find a lot of people comfortable saying that. Even some of the most impressive scientists I've met are extremely comfortable saying "I don't know." So I like to surround myself with those people, and then I tend to say it a lot too, or at least try to. And everywhere I go, I just try to learn something from people, because I don't know a lot of stuff and there's so much to learn.

Yeah. Marc Andreessen said something in an interview a while back that really struck a chord with me. He's not reading the news anymore — he's only reading books, or journal articles, or something posted on Twitter. I don't want to recommend anybody spend more time on Twitter, but this idea of disintermediating the way information about the present is presented to you, and going directly to the source — a scientific publication, or someone with political or economic responsibility live-tweeting about it. And then, on the other end of the distribution, look for the timeless books. I still don't read enough, but I try to spend more time on those two extreme ends of the distribution of content.

I know your audience is probably relatively online, so I don't need to encourage anybody to spend more time on Twitter. But if you're not yet on science Twitter, and you're a scientist or a healthcare provider interested in research, I'd encourage you to spend a bit of time there — you can see some stuff happen before it reaches mainstream coverage.

In terms of people I follow who have a very clear way of thinking: Josh Wolfe from Lux Capital is both a very clear-thinking investor and someone who works a lot with scientists, so he's grounded in reality. I'm active in crypto, so I also lean towards the work Balaji Srinivasan puts out — he's a pretty clear thinker, somewhat provocative, which isn't everybody's cup of tea, but I personally enjoy it.

And on books — going down to the real classics — there's one I recently finished. I had to look up the name so I don't butcher it: it's called Rational Choice in an Uncertain World. It's a university textbook, and you can only get it as a used book on Amazon. We can put the link in the show notes. It seemed like the textbook written before Kahneman made the more popular version, Thinking, Fast and Slow. So if you're into decision-making, take a look at that book.