More actions
Main article: Ratings system
In our last meeting (7/3/2024) we discussed methods for delegating trust weights. Dan’s preferred system is to give everyone 100 points to dole out as they see fit. You can spend points on other people and accumulate points spent on you from other people. One question that came up is whether you could spend the points you were given by others. The consensus seemed to be that yes, you could. If you accumulated say, 500 points, you could proxy it all out to others if you wanted. Another point was that if you delegate your weight to others, that knowledge should generally be made public in the CRS.
I mentioned that we might want to give out an initial set of weights based on some pre-existing credential, say, a college degree. Some people might therefore start out with more weight. Dan objected to that and said that people themselves using the ratings system should decide that. But what if the community votes and agrees to start everyone out using some heuristic metric? Obviously, it’s their decision. But we can recommend, and make it a default, that the ratings system start from a neutral point and allow everyone to gain experience with their peers before resorting to shortcuts. The ratings system is itself, after all, designed for this sort of problem.
This brings up two related problems: how to handle the past (before the ratings system) and how to handle heuristics more generally in what is certain to be a situation of information overload. Let’s say Alice joins a community where she knows some of the people but not everyone. She’s given 100 points to delegate, as is everyone else. She’s interested in foreign policy so her first task is finding out who in the community knows about this subject. She queries the newly formed CRS and asks everyone to provide their credentials in foreign policy. Some people answer and she evaluates their background and provides some of her weight to each of them. She does this for many policy issues, so many that the evaluation is overloading her. So she writes an AI-based algorithm to do it for her. It reviews everyone’s statement and reduces it to relevance and years of experience. Maybe college degree is included in that or maybe it isn’t. Her algorithm scores each respondent and she then assigns her weight to them based on that automatically. It’s like a rough pre-rating. Everyone in the community decides to use Alice’s method because they’re all feeling the pain of initially weighting people they don’t really know. As Alice comes to know them, and sees how they are being rated, she will modify her weight accordingly. For now she needs something simple to go on.
The community could of course choose to wait before assigning any weights and just gain experience. But it might be awhile. And while they’re waiting they could have some very unqualified people making foreign policy. It would take some time for the community to know who the truly qualified candidates are. Suffice it to say that in the end, the community would be doing its own weight assignments without relying on heuristics, whether or not they started with them or not.
There is a tradeoff here: The heuristic approach is inherently less accurate (and fair). But everyone-equal-at-first seems like it could lead to some fairly bad policymaking at first and require significant corrections later. And I would guess that the weights eventually agreed to, regardless of how we got there, will roughly correlate with an initially heuristic process. In my view, if we’re going to end up at the same place anyway, we should choose the system that yields the best governance, not the one that seems more democratic.
The objection to this admittedly utilitarian attitude is obviously the highly qualified candidate who does not have the degree or years of experience or any other credential to prove himself (initially). But he is, in fact, really good at climate science. The heuristic system discriminates against this person. But the ratings system will begin to include this person and shift weight to him as his opinions become voiced in the community. It’s certainly wrong that he has to start on the backfoot but he can rectify this, unlike current systems which will never allow him entry until he presents official credentials.
Whatever method is decided, we should discuss heuristics more generally, and keep repeating this point: The system requires us to digest a lot of information and rate on a lot of things. People are just not going to pay attention to all of it. We need to allow them to develop shortcuts. Indeed they will do so whether we allow it or not. Our system should be prepared for it and offer guidance and tools for this task.
Getting back to the scheme itself, we have a system where everyone gets 100 points at the beginning. Once the distribution is made, presumably it can be unmade. You should be able to take back your points if the person you delegated to loses your confidence. If this is the case, we might have to revisit the ability to spend your delegated points (as we said we should allow). Also, to account for growth, everyone should be given more points with the passage of time. That way, as new people enter the system, or as more experts rise in trust, there will be points available for them.