A quantitative causal model theory of conditional reasoning.

Abstract

The authors propose and test a causal model theory of reasoning about conditional arguments with causal content. According to the theory, the acceptability of modus ponens (MP) and affirming the consequent (AC) reflect the conditional likelihood of causes and effects based on a probabilistic causal model of the scenario being judged. Acceptability of MP is a judgment of causal power, the probability that the antecedent cause is efficacious in bringing about the consequent effect. Acceptability of AC is a judgment of diagnostic strength, the probability of the antecedent cause given the consequent effect. The model proposes that acceptability judgments are derived from a causal Bayesian network with a common effect structure in which the probability of the consequent effect is a function of the antecedent cause, alternative causes, and disabling conditions. In 2 experiments, the model was tested by collecting judgments of the causal parameters of conditionals and using them to derive predictions for MP and AC acceptability using 0 free parameters. To assess the validity of the model, its predictions were fit to the acceptability ratings and compared to the fits of 3 versions of Mental Models Theory. The fits of the causal model theory were superior. Experiment 3 provides direct evidence that people engage in a causal analysis and not a direct calculation of conditional probability when assessing causal conditionals. The causal model theory represents a synthesis across the disparate literatures on deductive, probabilistic, and causal reasoning. (PsycINFO Database Record (c) 2013 APA, all rights reserved).

Probabilistic single function dual process theory and logic programming as approaches to non-monotonicity in human vs. artificial reasoning

“…Despite Stenning and van Lambalgen (2005) themselves proposing that suppression effects might be dealt with by causal Bayes nets, actual attempts to model these inferences using CBNs have only just begun (Fernbach & Erb, 2013). Sloman and Lagnado (2005) and Ali, Chater, and Oaksford (2011) both looked at conditional inference and CBNs but not explicitly at the classical suppression effects (but see, Oaksford & Chater, 2013).…”

“…In establishing this point we will present some arguments that causal 7 Bayes nets (CBNs) can provide a good framework within which to develop a theory of conditional reasoning and we show that CBNs can naturally account for these patterns of inference (Ali, Chater, & Oaksford, 2011;Fernbach & Erb, 2013;Oaksford & Chater, 2007, 2013.…”

Probabilities, causation, and logic programming in conditional reasoning: reply to Stenning and Van Lambalgen (2016)

“…They found that ratings of the likelihood of disablers explains additional variance in predicting the inferences participants endorse, e.g., MP, over and above the sheer number of disablers generated. This experiment appears to replicate Fernbach and Erb (2013) where participants estimated disabling strength and the base rate of disablers to compute causal power in a noisy-OR gate. For each retrieved disabler, the disabling probability can then be calculated as disabling strength × base rate.…”

The diversity effect in diagnostic reasoning

“…For instance, according to one line of research on conditional reasoning, mental models can be tagged with probability information (e.g., Barrouillet, Gauffroy, & Lecas, 2008;Johnson-Laird, Legrenzi, Girotto, Legrenzi, & Caverni, 1999). Some research suggests that for causal conditionals, those mental models could be mental causal models (Ali, Chater, & Oaksford, 2011;Fernbach & Erb, 2013). Accordingly, causal reasoning with mental causal models of the acquired causal structure that allows for the incorporation of probabilistic evidence (Krynski & Tenenbaum, 2007;Meder et al, 2014;Waldmann & Hagmayer, 2013) could produce the diversity effect.…”