Learning through the Grapevine

          @misc{ scivideos_23034,
            doi = {},
            url = {https://old.simons.berkeley.edu/talks/learning-through-grapevine},
            author = {},
            keywords = {},
            language = {en},
            title = {Learning through the Grapevine},
            publisher = {The Simons Institute for the Theory of Computing},
            year = {2022},
            month = {nov},
            note = {23034 see, \url{https://scivideos.org/simons-institute/23034}}
          }
          

Abstract

We examine how well someone learns when information from original sources only reaches them after repeated person-to-person noisy relay. We characterize how many independent chains a learner needs to access in order to accurately learn, as these chains grow long. In the presence of random mutation of message content and trans- mission failures, there is a sharp threshold such that a receiver fully learns if they have access to more chains than the threshold number, and learn nothing if they have fewer. Moreover, we show that as the distance to primary sources grows, all learning comes from either the frequency or content of received messages, so learning only from the more informative dimension is equivalent to full Bayesian learning. However, even slight uncertainty over the relative rates of mutations makes learning from long chains impossible, no matter how many distinct sources information trickles down from. This suggests that forces which lengthen chains of communication can severely disrupt social learning, even if they increase the frequency of communication.