Facilitation of Bayesian decision making

Bayesian reasoning requires people to calculate a posterior probability by revising prior information with diagnostic information. Results from previous studies have shown Bayesian reasoning to be a challenging task for humans, revealing systematic biases and errors. This dissertation examines means by which Bayesian reasoning and Bayesian decision making (the actual decision making that relies on Bayesian reasoning) may be facilitated. The first three experiments involved manipulating the presentation of information. In the first experiment, subjects rendered a verdict based on DNA evidence. Some subjects were presented with prior information in natural frequency format, whereas others were presented with prior information in probability format. There was no effect of format on decision making or posterior probability estimation. However, there was an effect of explicitly presenting the posterior probability on decision making. In the second experiment, there was no facilitating effect of identifying information on Bayesian reasoning or Bayesian decision making. There also was no facilitating effect of first giving an easier problem on reasoning or decision making for the more difficult problem. A subset of the sample received an instructional intervention aimed at teaching subjects to construct frequency trees. This instructional intervention improved Bayesian reasoning and decision making. The results of the third experiment showed no effect of giving visual aids on facilitation of Bayesian reasoning and decision making. Results from these three experiments suggest that, other than explicit presentation of the posterior probability or explicit instruction in Bayesian reasoning, none of the manipulations that were examined facilitated performance on Bayesian reasoning or decision-making tasks. The fourth and last experiment began an examination of individual differences---namely, numeracy---on Bayesian reasoning and on susceptibility to framing effects due to changes in presentation. Numeracy was predictive of ability to make better Bayesian decisions and correct calculation of the posterior probability. In addition, highly numerate subjects were less susceptible to a framing effect of format.