In practice, many things can affect the verdict in a trial, including the testimony of eyewitnesses. Eyewitnesses are generally regarded as questionable sources of information in a trial setting: cases that turn on the testimony of a single eyewitness almost never result in a guilty verdict. Multiple eyewitnesses can, under some circumstances, collectively exhibit more robust behavior than any witness individually does. But how reliable, exactly, are multiple eyewitnesses? The legal literature on the subject tends to be qualitative. Here I describe a highly idealized Bayesian network model of the relation between eyewitness behavior and trial verdict. In a companion paper, I describe a more refined Bayesian model of the same setting. It turns out that the highly idealized model provides nearly as much information as the more refined one does.
展开▼