Today, in prominent economists doing the blockchain we have Al Roth. According to Bloomberg, he has signed on as an advisor to Covee; a blockchain startup creating a market for knowledge workers. Al Roth had recently said to me that his understanding was that the blockchain would be good for establishing trust but that we need to recognise that we already have other ways of establishing trust. So I would consider him a sceptic on this stuff. Therefore, if his name was attached to a venture, it was certainly worth a closer look.
When you go to Covee’s site, you immediately find this White Paper. White papers are typically vague affairs that hardly ever consider the academic literature that might be relevant. Not this one. Authored by six PhD’s, while far from an academic paper, it is very well grounded in the mechanism design literature and isn’t afraid to use or admit it. For the uninitiated, it may seem like a daunting read but, at its heart, it is pretty simple.
For context, here is the problem Covee are trying to solve. When you have teams of knowledge workers who need to come together for a project, we rely on businesses to manage and coordinate those teams. This is because you need to find the right people to perform the team’s tasks and then you need to ensure that everyone does what they are supposed to. Businesses solve that problem by making team members employees and learning about them. They can then use this to review performance and reward or punish them as need be. Covee wants to take the business out of the equation and, as much as possible, replace it with an algorithm on the blockchain. In doing so, it has to automate what the businesses were doing.
This is not a problem that you can find The Solution to. Instead, it is a bunch of problems that you need to find solutions to that won’t cancel each other out. That is a tough challenge. But I have read the white paper and Covee’s proposal actually holds together — at least at a first pass. Will it work or be valuable enough to form a viable alternative remains to be seen. In this space, being not obviously wrong is actually an achievement.
Let me try and explain what they are doing. The starting point is very traditional: if you put a bunch of people in a team and want to rely on them to do a set of tasks (especially if some of those require specialised skills), then how do you make sure they pull their weight? After all, in such teams, if one person does not perform everyone else has a reduced incentive to perform and the whole exercise can unravel.
Covee’s answer to this problem is ‘peer evaluation.’ Now you might think that sounds worrying but it is actually quite clever. Suppose you have a pot of money (say the payment for completion of a project) that you want to divide up amongst people who contributed. If I were to ask you how you thought other people did and to tell you that how much each of you gets is dependent on peer evaluation, you would have an incentive to claim that everyone else was a slacker. You all do that and so you don’t really find out who is slacking, people then slack anyway and the whole thing unravels.
The trick is to make sure that when evaluating your peers, there is no chance it will impact on what you end up getting. The solution Covee implement was contained in a 2008 JET paper by Geoffroy de Clippel, Herve Moulin and Nicolaus Tideman (“Impartial Division of a Dollar”). Each team member reports their opinion of other team members but it doesn’t impact on the share they receive. For instance, you are given a set of points and you have to allocate those points to others in the team. Then your share is a function of the points you receive from others but is, by construction, independent of what you allocate to others. In this situation, unless you have some sort of collusive arrangement with a subset of other members, you have an incentive to report truthfully.
There is, however, one problem with this arrangement. If you are a slacker, it really pays to be part of a team. Even if you get bad evaluations, you still end up with something. Here Covee have an innovation that is often used in mechanism design — staking. If you apply to be in a team, you are asked to put forward a stake (in crypto tokens). If you don’t get hired, you get that back. But if you are hired, then if you are a slacker, there is a good chance you will end up not even getting your stake back. In other words, by throwing your hat in the ring, you risk being worse off than before. What is more, the stakes give the algorithm more money to play with in terms of getting incentives right. So it is a win-win.
Stakes can signal confidence in getting a good evaluation which is useful. However, it might not be attractive. So Covee, like many before them, intends to have peer evaluations go into various reputation mechanisms that are preserved on the blockchain. These will help you get hired for other teams and so potentially create good long-term incentives for everyone involved.
In the White Paper, various other wrinkles are addressed. They all seem fairly reasonable. However, I have a broader worry. Is this all too mechanical? If the project is well defined, as are the tasks and responsibilities, then this might work very well. But what if “something” happens? What if the project or one task proves too difficult? What if it is more costly than expected?
In a business, if this occurs, someone communicates it, and then management, well, manages. They change expectations, responsibilities, provide resources etc, to ensure the project is managed back on track. In a Covee project, because everything is encoded and designed to be ‘mess free’ when a mess arrives, there is no obvious way to remedy it all. To be sure, there are incentives for the team to work out solutions. But at the same time, if the burden falls on some team members more than others, then everyone is going to start thinking how this will impact on their own share of the final reward. That is the very thing, Covee didn’t want anyone thinking about.
That said, they may be on to something here. Certainly, the first application I thought of was applying their solution to group projects in our MBA program. I will watch this with interest to see if that is worth doing.