6 ms·

Suppose the app was smart enough to do some clever analysis. For example, if all the people who leave bad reviews of Bob are all friends with each other, it sh

Suppose the app was smart enough to do some clever analysis. For example, if all the people who leave bad reviews of Bob are all friends with each other, it should count as a single bad review (so as to not double-count the evidence), and if Alice left bad reviews of several of your friends, it should know that you will probably find Alice untrustworthy, because she dislikes the people you like. Not that we can trust Peeple or any other app to get that right, but it sounds like it should be possible.

klodolph 11y ago
> Suppose the app was smart enough to do some clever analysis. When Alice gives a bad review of Bob, I think to myself, "Why is Alice giving a bad review of Bob?" I'm putting myself in Alice's shoes. There is absolutely no hope that a computer program would be sophisticated enough to do that, not yet, not for at least a decade (translation: possibly not in my lifetime). Computers can't even solve the "simple" problem of matching people in romantic relationships. When we throw computers at the more general problem of human interactions and trust, we should naturally expect them to completely botch it.
mdpopescu 11y ago
This is something I was thinking about as a solution for traffic violations: have drivers rate each other; increase the "trustworthiness factor" of anyone who thumbs down few people and those people are also negatively rated by others and/or have accidents; decrease the multiplier for those who just negatively rate everybody and so on. Unfortunately, I think you can hit some mines related to https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theore... but who knows? Maybe it won't happen given that you don't want a transitive relation, you want a numeric score for each driver. (Such a score could be used to influence insurance premiums; as an anarchist, I'm generally thinking about solutions for problems like "suppose we don't have patrol cops".)

klodolph 11y ago

> Suppose the app was smart enough to do some clever analysis. When Alice gives a bad review of Bob, I think to myself, "Why is Alice giving a bad review of Bob?" I'm putting myself in Alice's shoes. There is absolutely no hope that a computer program would be sophisticated enough to do that, not yet, not for at least a decade (translation: possibly not in my lifetime). Computers can't even solve the "simple" problem of matching people in romantic relationships. When we throw computers at the more general problem of human interactions and trust, we should naturally expect them to completely botch it.

mdpopescu 11y ago

This is something I was thinking about as a solution for traffic violations: have drivers rate each other; increase the "trustworthiness factor" of anyone who thumbs down few people and those people are also negatively rated by others and/or have accidents; decrease the multiplier for those who just negatively rate everybody and so on. Unfortunately, I think you can hit some mines related to https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theore... but who knows? Maybe it won't happen given that you don't want a transitive relation, you want a numeric score for each driver. (Such a score could be used to influence insurance premiums; as an anarchist, I'm generally thinking about solutions for problems like "suppose we don't have patrol cops".)