Care needs to be taken when using methods and specific assumptions should be explicit. The crowd is heterogeneous and the responses ofindividuals could be weighted and calibrated based on previousexperiments.  More research is requiredto develop techniques for calibrating probabilistic forecasts (P Pinson, PMcSharry, H Madsen (2010))

Quarterly Journal of the Royal Meteorological Society136 (646), 77-90; Reliability diagrams for non‐parametric density forecasts ofcontinuous variables: Accounting for serial correlation). It is also well knownthat some individuals are over-confident in their ability to forecast orassess.  Rather than accepting a pointforecast from an individual, it would be good to ask for a range of values.Techniques for combining probabilistic responses are available and this couldlead to a more accurate risk assessment.