I want to raise a concern about the upcoming futarchy experiment.
It may well be that there’s something I haven’t entirely understood about it; in that case, I would be happy to learn!
I just received an email and an invitation to sign up as a forecaster.
In the mail, it was explained that forecasters will vote on which protocols will be the most successful (if given a grant) in terms of TVL. The 5 protocols with the most votes will then receive a grant. An additional 5 protocols will receive similar grants based on the Grants Council’s votes.
Forecasters who make accurate predictions will be rewarded.
At the end of the email it says:
“By participating as a Forecaster, you’re not just making predictions – you’re helping shape the future of decentralized governance.”
I have a few concerns, both of them have to do with my impression that forecasters may be shaping more than the future of governance.
Are forecasters not incentivized to try to make their own predictions come true? If so, they would be best off voting for whatever they believe to be the most sensitive to manipulation/hype.
Won’t this produce a “race to the bottom of the brainstem” where not the best but the most manipulative protocols will receive many votes and get grants - because forecasters are rewarded not for picking out the protocols that they would like to see thrive, but the ones that they believe will be most viral/greedy/aggressive?
I understand that TVL is the core of this season’s intent, but I worry about creating the impression that we are simply being neutral in supporting what is expected to do well, when surely we are also shaping what will do well.
And I am concerned that this experiment seems to directly (through a grant) and indirectly (through the incentivized behavior of voters) work towards what people think is the most likely phoenix - not a desirable one.
Hi @joanbp, thanks for the comment. I hear your concerns about potential for manipulation in this experiment. The fact that collective beliefs lead to actual decisions (i.e., grants allocations) rather than merely representing collective expectations is a pretty interesting difference from most other implementations.
Realize this, and other constraints like forecasters depositing play tokens rather than real money in this experimental pilot, does change incentives and dynamics. This is why we’re intentionally starting small to first observe, learn, and iterate about additional decision-making tools for the Collective.
Couple things to consider:
While forecasters might try to manipulate votes, the idea is that the further predictions deviate from reality, the stronger the incentive is for others to correct them (“manipulation paradox”)
The concern about forecasters voting for the most likely rather than most desirable phoenix is also valid – this resembles a “Keynesian beauty contest” where instead of voting for what they think will be most impactful, forecasters would vote based on what they think others will predict will succeed
A key difference here is that top forecaster rewards are based on actual TVL generated after 3 months, not just collective beliefs
You’re right that the impact metric here is based on the Collective’s Season 7 intent – raises interesting questions about finding a metric that is difficult to game in a Futarchy-inspired setting
Right. Let me see if I understand what you’re saying:
It is a slightly paradoxical dynamic where a vote will
impact the allocation of a grant (high impact if you vote like others on the protocols with high expected TVL, the reward will go to the protocol) and
be rewarded based on the actual outcomes (highest impact if you voted unlike most others and you were right, the reward will go to yourself)
Did I get that right?
That is an odd dynamic to look at as a potential voter.
Also, A follow-up question. In the email I received, it says:
“Given a 100K OP grant budget, how much TVL will this protocol bring to the Superchain in three months?”
Does this mean that only bets on the protocols that actually end up getting a grant will count towards the voter’s accuracy score? (We will never know how well other protocols would have done, if given a grant)
It seems to me that what we really need is metrics that we would want people to game. Metrics that capture what is clearly good, and what will not turn bad in the case of unexpected, overwhelming success. Metrics that ensure balance and sustainability, not blind pursuit of one goal at the expense of everything else.
For sure – by “difficult to game” we mean can’t be manipulated in ways that will backfire. So in this sense we do want people to “game” or optimize for something that, if they do more of, it will benefit them while simultaneously supporting a positive outcome. One simple illustration of this is a points program that rewards ETH staking – someone can “game” this all they want and it’s still beneficial to the system. Ofc this gets messier in governance where things like voting or delegating or allocating grants have more dimensions (harder to disentangle quality versus quantity of governance participation)…