Deductive rationality in human reasoning: Speed, validity and the assumption of truth in conditional reasoning

We proffer the thesis that, in the process of defeating an inference on the basis of a factual truth that falsifies it, people move from a hypothetical truth-value to a factual truth-value of the conclusion. We will present evidence that shows (a) that some people spontaneously make a truth assumption and constrain their inferences to logically valid inferences, (b) that people tend to abandon the truth-assumption when they have factual evidence to the contrary, (c) that people, however, can and do in fact reason logically when they are informed about the rules of the language game (i.e., the truth-assumption) and (d) that adhering to the truthassumption in the face of conflicting evidence to the contrary requires an investment of time and effort. The findings are discussed in relation to contemporary theories of human reasoning. General Introduction We all reason: We draw inferences from the multiple sources of information we are confronted with and make decisions based on them. This allows us to move around in a changing world where the capability to comprehend the contingent nature of our environment determines for a large part our successes as an individual, as well as a species. The study of human reasoning is therefore important to advance our understanding of the general mechanisms of thought. The turn of the century has provided the stage of a paradigm shift in human reasoning research. The nineties provided the scene for polemical debates as regards basic human reasoning competence. This basic reasoning competence (i.e., the basic machinery that allows us to draw inferences) was mostly studied by means of abstract knowledge-lean inference problem. By using arbitrary relations (e.g., ‘if the letter is an A, then the number is a 2") no content-specific background knowledge would be triggered to influence the reasoning process towards accepting or rejecting the conclusion. Abstraction was made of the specific content of that about which people were reasoning. It is within this research milieu that theories became specified as regards human deduction. In the study of human deduction one studies necessary inferences derived from certain premises. One asks people to draw logically valid inferences, and these are defined as inferences that must be necessarily true if the premises are true. Presently there is an increasingly prominent body of evidence that shows the pervasive influence of content and belief (Cummins et al., 1991). Our beliefs are uncertain (i.e., they are true to a certain degree: e.g., even Newton’s mechanics are not universally applicable). This observation induced a shift towards the study of the subjective probabilistic properties of that about which we are reasoning as well as commonsense reasoning or reasoning under uncertainty. The present research is situated within this timely clash between experimental paradigms and associated theoretical approaches. Theorists sometimes like to boost the polemics between dichotomized opposites (it does make for simpler, and hence more easily publishable reading). For instance, it is claimed that theories that have focused on reasoning under certainty (i.e., deductive reasoning) are incapable of being extended to reasoning under uncertainty (i.e., probabilistic reasoning). The ‘core argument’ (Oaksford & Chater, 1998) is that common-sense reasoning is nonmonotonic, whereas logic systems are monotonic: valid inferences cannot be invalidated; they remain valid. The validity of everyday inferences however would be revisable. For instance, when being given the argument: ‘If it is a bird, then it flies; Tweety is a bird who, thus, can fly” almost everybody will accept it. At the same time, when subsequently being told that Tweety is an ostrich, almost everybody will reject the original inference and will state that Tweety cannot fly. The rationality debate in the cognitive science of human reasoning is partly muddled by a failure to distinguish the defeasibility of a conclusion from the non-monotonicity of an inference. For instance, Oaksford and Chater’s (1998) core argument is subverted when taking count of the distinction between truth and validity. Monotonicity concerns the validity of inferences; defeasibility concerns the truth of conclusions and this “distinction between validity and truth ... is basic to deductive logic [and] many people find the distinction difficult to grasp” (Glass & Holyoak, 1986, p. 338). The abovementioned definition of logical validity use the notion of truth but the truth of a valid conclusion is always hypothetical (if the premises are true, then the conclusion must also be true). The truth-value of a defeated inference however is not hypothetical. It is factual: it hinges on a factual truth (i.e., our belief, at a particular moment in time and space that something is true in the ‘real’ world). The present study intends to show the importance of the truth-assumption and by consequence the hypothetical nature of the truth of logically valid inferences. We proffer the thesis that in the process of defeating an inference people move from a hypothetical to a factual truth-value of this conclusion. I present evidence showing (a) that at least some people make the truth-assumption and spontaneously constrain their inferences to logically valid inferences, (b) that people abandon a truth-assumption when they have factual evidence to the contrary, (c) that people, however, can and do in fact reason logically when they are informed about the rules of the language game (i.e., the truthassumption) and (d) that adhering to the truth-assumption in the face of conflicting evidence to the contrary requires an investment of time and effort. In the general discussion we will then return to the theoretical and conceptual issues that are touched by the evidence for people’s propensity to exhibit deductive rationality in reasoning hypothetically on the basis of a truth-assumption.