DAOs Were Supposed to Run Themselves. Two Papers, One Unfinished Argument.

DAOs were built on a simple promise. Decisions made by the crowd, executed by code, with no boss in the middle. Two recent papers take that promise and pull in opposite directions - and neither one closes the case.

One paper, from three researchers writing in Frontiers in Blockchain (August 2025), builds a framework to help founders design good community governance. The other, from a team at Imperial and INSEAD, runs 2,380 DAOs through DeepDAO and finds that most governance doesn't work as designed.

Put them together and you don't get an answer. You get an open argument about whether the original idea can hold up in practice.

The theory keeps missing reality

Both papers agree on what goes wrong. Power concentrates. Voters don't show up. Whales decide.

The Frontiers team treats these as risks you can manage. The Imperial-INSEAD team measures them and finds the gap is huge. Over 80% of voting rights sit with under 1% of addresses. Fewer than 10% of tokens ever vote. Only 8% of DAOs run governance fully on-chain, the way the code-runs-itself dream requires.

That's the recurring theme across the whole field. The theory says decentralized and automatic. The practice keeps coming back centralized and hands-on.

The fix that quietly breaks

Here's where the two papers actively disagree.

The Frontiers framework recommends delegated voting and liquid democracy - let busy holders hand their vote to someone reliable. Sensible on paper.

But the empirical work shows delegation doing the opposite of what it promises. Big delegates almost never break from the pack. So delegation doesn't spread power. It just pools it in fewer, quieter hands. One paper's solution is the other paper's problem.

But it's not settled

This is the part worth sitting with. The "it's structural, you can't fix it" view isn't the last word.

Newer studies find designs that do better. A 2025 study of Internet Computer DAOs measured around 64% voter turnout - roughly double what Ethereum DAOs manage - using a different delegation model. Other 2025 work argues reputation-based and share-based systems can soften the concentration that plagues token voting.

So the honest state of play is: token-weighted DAOs keep failing the theory, but nobody has proven the whole idea is doomed. The design space isn't exhausted. It's barely mapped.

Some authors to follow if you want to track this

The argument is being carried by a small, active set of researchers:

• Robert Fritsch and Roger Wattenhofer (ETH Zürich) - the voting-power measurements everyone cites.

• Tanusree Sharma - large-scale studies of how DAOs actually operate.

• Okutan and colleagues - the liquid-democracy counterexamples that keep the debate open.

• Ying-Ying Hsieh and Phanish Puranam - the organizational-theory read on what DAO failure means for how we think about organizations at all.

• Philipp Reineke, Riitta Katila and Kathleen Eisenhardt - the wider question of whether decentralization in organizations is real or a mirage.

• Vitalik Buterin - still the sharpest voice on where decentralization is supposed to matter, and where it isn't.

The unresolved question is the one none of them has answered yet. Is DAO centralization a bug you can engineer out - or the price of the whole design? Until someone settles that, everything built on top of it is a bet.

Sources: Viaro Bridi, Andrlić & Gonçalves, Frontiers in Blockchain (Aug 2025); Hsieh, Puranam & Gelman-Hofstadter, forthcoming in Industrial and Corporate Change; DeepDAO dataset; Okutan et al. (2025) and Fritsch et al. on voting power and turnout.h