Sonnemans, J. and van Dijk, F. (2012). Errors in judicial decisions: experimental results Journal of Law, Economics, and Organization, 28(4):687--716.

In criminal cases, the task of the judge is foremost to transform the uncertainty about the facts into the certainty of the verdict. An extensive literature shows that people deviate from rationality when dealing with probability. It seems therefore unavoidable that in difficult criminal cases, miscarriages of justice occur, but this is hard to study in the field. In a laboratory experiment, we examine the relationship between evidence of which the diagnostic value is known, subjective probability of guilt and errors in verdicts for abstract criminal cases. We look at two situations: (1) all evidence is given and (2) evidence can be acquired. In both situations, verdicts are inaccurate. For given evidence, errors are biased toward the most serious type, unfounded conviction. In the situation where evidence can be acquired, participants do not acquire enough which results in many mistakes, evenly divided over unfounded convictions and unfounded acquittals. We suggest ways to reduce error.