Will Manifold natively support "committing to actions" within 6 months?
31%
chance
Reasons to do this:
We've actually seen this particular use case picking up steam among users:
- https://manifold.markets/misha/will-i-finish-a-triathlon-4ea28f3c991e
- https://manifold.markets/guzey/will-i-create-at-least-one-more-pre
- https://manifold.markets/misha/will-i-stick-with-assigning-probabi
Also, one of our most ardent supporters is @dreev, the founder of Beeminder. I think Beeminder is counterfactually responsible for about 1 more blog post and 2 exercise sessions since I started ~2 months ago; I've also paid them ~$100 in fees, which seemed easily worth it.
Also, commitment devices are a great way of managing acausal trade: https://forum.effectivealtruism.org/posts/ndvguMbcdAMXzTHJG/what-we-owe-the-past
Also, @Sinclair has been musing about managing reputation/flakiness/commitment publicly via markets on eg "Will I actually show up to the party I said I was for sure going to go to?"
Reasons not to do this:
- More complexity is bad (very good reason!!)
- Not enough dev time to get to
- ???
Restricting whoever's making the commitment from selling seems right - I don't think you want people to be able to back out of commitments like that! Allowing others to sell seems okay.
Would you also restrict sales then?
Additional cheeky sidenote to the buying-restriction thing: I'm no lawyer, but it seems like that kind of market might not count as gambling, so you could maybe use real money? 🤷♂️
Another (purely aesthetic UI) change would be renaming YES and NO - maybe SUCCESS and FAILURE? If you restrict who can buy what, then since any one person can only buy one kind of payout anyway you'd just need to show how much was at stake, and how much was from the committer vs everyone else.
One change I would make to a commitment market is to restrict who can buy what.
If I make a commitment, it makes sense to restrict me to buying YES, since buying NO would be tantamount to going back on my word. And it makes sense to stop others from buying YES - I should be the one reaping the rewards for doing what I said I would and people wanted to incentivise me to do, not someone else who predicted I was capable. We're not betting on beliefs here, we're incentivising the reliability of a specific person or group.
So that person or group should only be able to buy YES, while everyone else should only buy NO.
It occurs to me that with those restrictions in place, a DPM market creates better incentives than CFMM on both sides - if I keep my word I get all the money people wanted to pay to incentivise me, while if I fail people get their money back plus an amount more-or-less proportional to how much they invested, with earlier investments being favoured.
And if multiple people are party to the commitment, their payout is proportional to how much they were actually committed to making the thing happen, again with earlier investments favoured.
Sinclair Chen is betting YES at 31% 17 days ago
@undox why did you stop?
were there too many ughs from going from thinking "I want to commit to X" to having a market that does that?
Seems like a good idea. I used MM as a commitment device early on and it worked well. Reminds me to do this again.
Sinclair Chen bought M$10 of YES17 days ago
not promising to do this myself, but I think this is a great idea!
Agreements to cooperate for greater gain are an important part of how civilizations get wealthy. The technology exists to do this more or less correctly between corporations, but not between friends. I think a lot of possible fun is lost because friends don't trust each other enough in our modern atomized society. Knowing that friends who turn out to be evil can just run away to another social group - the sadness of this is not mainly in the occasional betrayal and hurt feelings, the main sadness is missing out on all the lovely arrangements that could exist if only people trusted each other more
probably not all that different from a prediction market, just with a better UI for this specific use case
How would this differ from just making a prediction market and putting a bunch of M$ on whatever outcome you want to commit to?
Oh nice - I'd seen you link to it but I hadn't realized you also built it! Looks very pretty and awesome (I wish Beeminder itself had this level of visual polish 😂)
Yay! It's amazing to hear this. Thank you!
(Side note: I can't decide if I should bet YES because I think it would be amazing, which I do, or bet NO to contribute to the implicit bounty on this!)
As for @Sinclair's idea, we started building something like that as a side project called commits.to and I'm pretty enamored with some of the ideas in the spec: https://doc.beeminder.com/willspec