1 month, 4 weeks ago
star star star star_half star_border
I read a few papers from Dražen Prelec and coauthors. The common thread in these papers is solving the "crowd wisdom problem". They present various tricks, most importantly the "surprisingly popular" algorithm.
The algorithm asks participants some question like "Is X true?", then asks a follow-on, "Will most respondents say yes to the first question?". The crowd answer under SP is the answer with the largest difference in number of responses between the first and second question.
SP is more robust against certain empirically-observed failures in decision making and gives more accurate results than majority-voting when a small number of participants have a disproportionate amount of information.
doi.org/10.1038/nature21054 doi.org/10.1126/science.1102081 doi.org/10.1509/jmr.09.0039