The well-known jury paradox – the more demanding the hurdle for conviction is, the more
likely it is that a jury will convict an innocent defendant – heavily relies on Bayesian updating.
However, with ambiguous information (e.g., a forensic test with accuracy of 60%, or more),
standard Bayesian updating becomes invalid, challenging the existence of this paradox. By
developing novel theoretical models and by testing their predictions in laboratory settings,
this thesis advances our understanding of how individuals process more realistically imprecise
measures of information reliability and how this impacts on information aggregation for the
group decision-making. Hence, our findings inform the institutional design of collective
deliberation, from small to large group decision-making.