New normative standards of conditional reasoning and the dual-source model

Area of Science:

• Cognitive Psychology
• Decision Science
• Philosophy of Mind

Background:

• Human reasoning research is increasingly adopting Bayesian and probabilistic frameworks, viewing inference under uncertainty as central.
• This paradigm shift emphasizes reasoning from uncertain beliefs rather than arbitrary assumptions.

Purpose of the Study:

• To empirically test normative standards of reasoning within the probabilistic paradigm.
• To investigate how people make inferences in conditional reasoning tasks with uncertain premises.

Main Methods:

• A novel probabilized conditional reasoning task was developed to empirically test normative standards.
• Participants' responses were analyzed for confidence preservation (p-validity) and coherence with mental probability logic.
• A descriptive dual-source model was compared against normative accounts.

Main Results:

• For everyday conditionals with a causal link, only the conditional probability of the consequent given the antecedent predicted the probability of the conditional.
• Participants' inferences were confidence preserving for Modus Ponens (MP) but not for Modus Tollens (MT).
• The dual-source model, integrating background knowledge and logical form, provided a strong account of the data, comparable to models using only task-specific data.

Conclusions:

• The findings provide empirical support for probabilistic approaches to human reasoning, particularly for MP inferences.
• Normative standards partially explain reasoning, but descriptive models like the dual-source model offer a more comprehensive account by including subjective factors.
• The study highlights the importance of both probabilistic and subjective factors in understanding human conditional reasoning.