A rational analysis of the selection task as optimal data selection by Mike Oaksford and Nick Chater challenges the traditional view that human reasoning in hypothesis-testing tasks, such as Wason's selection task, is biased. Instead, the authors propose a Bayesian model of optimal data selection in inductive hypothesis testing, which aligns well with human performance on both abstract and thematic versions of the task. This model suggests that reasoning in these tasks may be rational rather than subject to systematic bias. The selection task involves four cards, each with a letter on one side and a number on the other. A rule is presented, such as "if there is a vowel on one side, then there is an even number on the other." Subjects must decide which cards to turn over to test the rule. Logically, only the p (vowel) and not-q (odd number) cards should be turned over. However, in practice, only 4% of subjects select these cards, indicating a discrepancy between normative theory and behavior. The authors argue that the selection task is a laboratory version of the problem of choosing the best experiments to test scientific laws. Popper's falsificationism, which posits that experiments can only falsify general laws, has been criticized for not aligning with the history of science. Instead, a Bayesian approach to optimal data selection is proposed, which emphasizes the expected information gain from each card. This approach suggests that subjects act as Bayesian optimal data selectors, optimizing the expected amount of information gained. The authors apply this Bayesian model to the selection task, showing that subjects' behavior can be explained by the rarity assumption, which posits that properties involved in causal relations are rare in the environment. This assumption allows the model to capture a wide range of experimental results, including the standard abstract results, nonindependence of card selections, negations paradigm, and thematic tasks. The model's behavior is illustrated in Figure 2, showing the expected information gain for each card based on the probabilities of p, q, and the independence model. The model predicts that the p card is informative when q is rare, the q card is informative when both p and q are rare, the not-q card is informative when p is common, and the not-p card is not informative. These predictions align with empirical data, suggesting that subjects base their card selections on the expected information gain of each card. The authors also discuss the nonindependence of card selections, showing that similarly valenced cards are positively associated, while dissimilarly valenced cards are negatively associated. This pattern is consistent with the model's predictions. The model is further extended to account for data from the negations paradigm selection task, where the antecedent and consequent of a rule can contain negated constituents. The model explains the observed behavior in these tasks, including the matching bias and the effects of thematic content on card selection. Overall, the Bayesian model of optimal data selection provides a rational analysis of the selectionA rational analysis of the selection task as optimal data selection by Mike Oaksford and Nick Chater challenges the traditional view that human reasoning in hypothesis-testing tasks, such as Wason's selection task, is biased. Instead, the authors propose a Bayesian model of optimal data selection in inductive hypothesis testing, which aligns well with human performance on both abstract and thematic versions of the task. This model suggests that reasoning in these tasks may be rational rather than subject to systematic bias. The selection task involves four cards, each with a letter on one side and a number on the other. A rule is presented, such as "if there is a vowel on one side, then there is an even number on the other." Subjects must decide which cards to turn over to test the rule. Logically, only the p (vowel) and not-q (odd number) cards should be turned over. However, in practice, only 4% of subjects select these cards, indicating a discrepancy between normative theory and behavior. The authors argue that the selection task is a laboratory version of the problem of choosing the best experiments to test scientific laws. Popper's falsificationism, which posits that experiments can only falsify general laws, has been criticized for not aligning with the history of science. Instead, a Bayesian approach to optimal data selection is proposed, which emphasizes the expected information gain from each card. This approach suggests that subjects act as Bayesian optimal data selectors, optimizing the expected amount of information gained. The authors apply this Bayesian model to the selection task, showing that subjects' behavior can be explained by the rarity assumption, which posits that properties involved in causal relations are rare in the environment. This assumption allows the model to capture a wide range of experimental results, including the standard abstract results, nonindependence of card selections, negations paradigm, and thematic tasks. The model's behavior is illustrated in Figure 2, showing the expected information gain for each card based on the probabilities of p, q, and the independence model. The model predicts that the p card is informative when q is rare, the q card is informative when both p and q are rare, the not-q card is informative when p is common, and the not-p card is not informative. These predictions align with empirical data, suggesting that subjects base their card selections on the expected information gain of each card. The authors also discuss the nonindependence of card selections, showing that similarly valenced cards are positively associated, while dissimilarly valenced cards are negatively associated. This pattern is consistent with the model's predictions. The model is further extended to account for data from the negations paradigm selection task, where the antecedent and consequent of a rule can contain negated constituents. The model explains the observed behavior in these tasks, including the matching bias and the effects of thematic content on card selection. Overall, the Bayesian model of optimal data selection provides a rational analysis of the selection