Modeling and critiquing the confirmation bias in human reasoning

In trying to solve even small problems, humans often form wrong hypotheses by looking only at the most salient factors and by ignoring contrary evidence. Once they form their initial beliefs, they exhibit a strong tendency to seek only additional confirming evidence. This type of behavior is the confirmation bias. Empirical evidence is presented suggesting that computers can interact with humans by qualitatively modeling and detecting their confirmation bias and by offering criticism strategies that will effectively mitigate the unwanted effects. The degree of improvement is sensitive to the strategy and content of the critic. A number of strategies are constructed and evaluated, and it is shown which ones are most likely to lead to success for varying circumstances and user types. Additional fruitful research directions are pinpointed