Probabilities and polarity biases in conditional inference.

Area of Science:

• Cognitive Psychology
• Computational Modeling
• Human Reasoning

Background:

• Conditional inference tasks often reveal polarity biases.
• Existing models struggle to fully explain these biases, particularly those observed with negation.

Purpose of the Study:

• To propose a probabilistic computational level model of conditional inference.
• To explain polarity biases using a novel assumption about negation probability.

Main Methods:

• Developed a probabilistic model where negations represent higher probability categories.
• Conducted three experiments manipulating category probabilities to test the model's predictions.

Main Results:

• Confirmed that high-probability categories, like negations, elicit a high-probability conclusion effect.
• The findings support the probabilistic account of polarity biases.

Conclusions:

• Polarity biases in conditional inference are a rational consequence of category probabilities.
• The proposed model offers a parsimonious explanation for these biases in human reasoning.