It’s important to note that Silvio did not say that participants in decentralized consensus protocols have no incentives. Instead he said that it was simply too difficult to reason about these incentives. I personally find this a bit of a defeatist outlook, which is why my focus is on analyzing and developing cryptoeconomic mechanisms in which we analyze incentives to the best of our ability.
First I think we should try to define what incentives are. According to our overload Google, incentives are “a thing that motivates or encourages one to do something”. From this definition, claiming that there is no incentive for people to share files is clearly false. Let’s look at the incentives for a user downloading a movie:
- User wants to watch movie, motivating them to find the torrent hash.
- User wants to easily download movie, motivating them to download a popular torrenting client.
- Torrenting client developer wants a healthy torrent network, motivating them to automatically seed movies as users download them.
Result: User gets to watch movie, user shares movies to other users.
In Bittorrent the stakes are low. We don’t have millions of dollars on the line held up in a single smart contract. This means while there are incentives (which I just outlined), they aren’t on the order of millions or billions of dollars. However, in blockchains this is not the case.
Recently Bitcoin Gold suffered a 51% attack which stole something like 18 million dollars. That’s high stakes. When we are building systems which need to be trusted with millions of dollars, analyzing financial incentives is a necessity. I would not feel comfortable condoning any software which is entrusted with large sums of money without strong justifications for why that money is safe.
That said, even Algorand provides justifications for why their money is safe. They make a particular honesty assumption which includes honest users running full nodes without direct financial compensation. I think this property is what you meant when you asked “can cryptocurrency work without incentives?”. From their honesty assumptions & protocol specification, their security properties can be derived and proved correct. That’s great but the harder question is: do you personally believe their world view is realistic? Next question: does their analysis convince everyone else? If the answer to both is yes, then they’ve done their job!
If you were to ask me if I think their world view is realistic, I would probably say no. However, that’s not so important–I could be wrong! What is more important is improving how we analyze what security properties our internet protocols actually provide, so that we have compelling justifications for why everyone can feel safe when using them. That means I strongly disagree with Silvio’s defeatist outlook and instead believe we need to develop our methods for analyzing incentives. Right now the tools we use to analyze incentives within our protocols are quite limited. However, thankfully there’s a lot of research being done in this area under the heading “cryptoeconomics” and that’s clearly what I’m interested in.
So specifically regarding Algorand, I tend to agree with the sentiments expressed by Vitalik, Vlad, and Emin in the article you linked. Still, even if I am wrong what is important is the fact that we are talking about incentives in the first place. If we are to build complex systems which many humans & bots rely on, we are going to need to develop the tools for reasoning about these complex systems. We need cryptoeconomics (dot study) !