Naive Learning Through Probability Overmatching: Along Any Vanishing-Influence Network Sequence, Extremization DeGroot Processes Eventually Agree on the True State with Probability Tending to OneResearch Paper
Models of naive social learning ask whether agents who update by simple averaging rules, rather than by Bayes' rule, can still aggregate dispersed information. In the classic DeGroot model every agent repeatedly replaces her belief by a weighted average of her neighbours' beliefs. Golub and Jackson (2010) showed that when agents communicate real-valued…