Reading through it and how it's applied (mostly with crypto). I have to say that the biggest flaw I see isn't theoretical, it's practical.
QF assumes that you can know for sure who is an individual. Yet how would you know that with crypto funding?
Let's say I'm malicious and I want to pillage a QF. What stops me from setting up a bogus social project/company, registering it, and then taking my $1000 and splitting it into 1000 wallets with $1 a piece which all contribute to my scam project?
If I know a QF fund is getting setup, it'd be pretty easy to create 1000s of wallets, vary the money in them, and have them all fund my scam. I can even automate some trading between these wallets to make the source of the funds look somewhat organic.
Pillaging these funds seems like it's almost a trivial endeavor assuming you can get your own scam company associate with them. And the more money you have, the easier it'd be to pillage.
> Pillaging these funds seems like it's almost a trivial endeavor assuming
It is, and in fact the authors point this out in the original paper:
"…if the size of this group is greater than 1/α and the group can perfectly coordinate, there is no limit (other than the budget) to how much it can steal."
> I have to say that the biggest flaw I see isn't theoretical, it's practical.
Exactly. The theory is fine -- given all these assumptions hold. In practice, these assumption don't hold.
For example, one of the assumptions is absence of sybil attacks, fraud, or collusion. Obviously, these assumptions may not hold.
You can defend against sybil attacks in various ways. But how do you stop people from colluding (e.g. I $10 to 1000 friends, tell them they can keep $5 if they contribute $5 to my project)? There are collusion-resistant forms of quadratic funding, such as COCM, but these do not have the desirable theoretical properties (such as optimality) that vanilla QF has.
It's funny that a cooky proposal originating from crypto, which is incredibly inefficient precisely because it has to defend against sybil attacks (unlike permissioned systems), assumes the absence of sybil attacks. Hilarious, really.
QF assumes that you can know for sure who is an individual. Yet how would you know that with crypto funding?
Let's say I'm malicious and I want to pillage a QF. What stops me from setting up a bogus social project/company, registering it, and then taking my $1000 and splitting it into 1000 wallets with $1 a piece which all contribute to my scam project?
If I know a QF fund is getting setup, it'd be pretty easy to create 1000s of wallets, vary the money in them, and have them all fund my scam. I can even automate some trading between these wallets to make the source of the funds look somewhat organic.
Pillaging these funds seems like it's almost a trivial endeavor assuming you can get your own scam company associate with them. And the more money you have, the easier it'd be to pillage.