From Mirror Budget to Weak Anarchism - VECTION
Public.icon
https://gyazo.com/ba50e40bea9280689b12df2f80829f85
Discovered through a tweet by cameturtle.icon
@cameturtle: "What percentage of the other participants do you predict will answer yes to that question?" ↓
By using the fact that the minority who know the correct answer will estimate the percentage lower, collective knowledge can be aggregated. It's interesting to use both predictions for the problem and predictions for other people's predictions when predicting a problem.
The authors of a paper that improved the method of correcting the majority vote called Mirror Budget added a question asking what percentage of the other participants who are predicted to answer yes to the question will answer yes. This is an improvement that increases the likelihood that one's own answer is correct by comparing one's own confidence with that of others. This allows experts The summary is cut off in the middle?
https://gyazo.com/37998380fe23267369612fd56acf876b
As a method of correcting the majority vote, there is also a method of asking about the confidence in one's own answer. If it is assumed that those who have confidence in their answer are more likely to be correct, this seems like a good improvement.
Therefore, the authors of this paper added a question asking "What percentage of the other participants do you predict will answer yes to that question?" This means that the degree of agreement between one's own opinion and that of others is also added.
It's like applying "I know something that others don't understand." It utilizes the Meta-cognition of some experts or knowledgeable people. Previous companies, interviews and appearances related to oneself.icon For example, if someone answers "yes" to the question "Is Philadelphia the capital of Pennsylvania?" (mistakenly believing it to be true), they will naturally assume that others will also answer "yes". Therefore, they will give a high percentage for agreement questions. However, someone who answers "no" to this question (knowing that it is actually more difficult than it appears) will give a low percentage for agreement questions (confident that everyone else will get it wrong).
Using this fact, a strategy can be devised to choose the option that was unexpectedly popular (the surprisingly popular algorithm) as the "correct" answer for the person who showed a lower percentage than expected, or to use prediction-normalized vote with correction that has a similar meaning. In the case of Philadelphia, this strategy would choose the knowledge of the minority. Moreover, in cases where no one knows the answer, it will always surpass the accuracy of simple majority voting or voting with confidence. The author is ysmxysmx.icon?Previous companies, interviews and appearances related to oneself.icon