How do we reason?

Kostas Katsanevas


To begin a discussion on human decision making and rationality one must first unearth the conceptual roots of reason – those found in the science of philosophy. This concept has been the theme of philosophical discussion since classical Greece, taking on many different interpretations throughout time and by various thinkers. The ancient Greeks revered reason, setting its stature to the highest degree; a connection of the mind with the greater cosmos.


In his treatise The Republic, Plato delineates reason to be one aspect of the tripartite soul, further arguing that reason must rule over the psyche if the man is to be just (Republic, IV). In Plato is where logic, as we understand it today, derives one of its main principles – the law of non-contradiction – a fundamental axiom upon which the rationality discussion depends (Republic, IV).


Other interpretations of reason vary from such thinkers as Kant, who also believed we should be governed by reason (Kant’s Critique of Pure Reason, Kitcher, 1998); Herder the romantic, who argued that language must come before thought which in turn forms reason (Herder: Philosophical Writings, Forster, 2002); and Rousseau, who used reason to escape the state of nature (Rousseau The Social Contract, Tozer, 1998); amongst many others who viewed this notion as the tool for greater understanding of human, natural and metaphysical processes.


Interestingly enough, one could argue that there is no such thing as reason and that the only certainty one can be sure of, is one’s own mind – aptly named solipsism. Consequently, reason can be a symbol, a representation of the individual’s own schema – a sub-category of the society in which it breathes. By and large reason can represent both an individual and his surroundings, the fabric of a polis and its innate structure through the looking glass that bares its soul.


Science & Reason

The extent to which this can extend can be exhaustive and as the rationality debate has evolved, several different sciences have taken on the responsibility of attempting to formulate an absolute standard from which reason is based. Psychology for instance has led the discussion in relation to rationality and human decision making. One way it has done this is by quantifying particular sets of problems through the use of key assumptions in an attempt to determine various views of rationality.


The most prominent example of testing whether humans use logic to reason is the famous Wason Selection Task (Wason, 1966). This paper will attempt to illustrate how different notions of reason can be viewed as correct based on their structured models as illustrated through the WST.


Decision theory is at the center of this explanatory endeavor. Most of what decision theory entails are prescriptive frameworks, that is, how people ought to reason. On the other hand, the practical application of prescription is methodological, a tool on helping people actually make decisions. Theories about cognitive heuristics and biases are descriptive in nature as they can be tweaked post hoc, this being a major difference. That is to say, normative theories must be consistent with human behavior, whereas descriptive theories can be modified so that they can fit any irregular behavior observed in experiments.


Nevertheless, if human reasoning stays consistent with a particular normative framework then it can be said to be rational. Thus, it can be inferred that in order to have a ‘correct’ notion of a specific decision one must be consistent with the particular normative framework that he or she subscribes to. In essence, more than one answer can be the ‘correct’ one since there is more than one normative theory to explain human decision making.


The classical normative approach in decision theory is namely, logic. This method follows certain axioms that vary based on context, but operates under clear universal principals. These can be understood as the laws of thought in most cases and are defined as “rules that apply without exception to any subject matter of thought, etc.; sometimes they are said to be the object of logic” (Audi, 1999). One example of these rules is the idea that something simply is or isn’t, it cannot rain and not rain at the same time, and hence such paradigms are assumed to be self-evident truths.


These valuable underpinnings are understood as axioms and are used to derive conclusions that all fit neatly under the umbrella of the logical framework.  Thus, reasoning based on logic means to make decisions in a way that is consistent to undisputable truths. If one agrees with this approach and consistently uses logic to reason, then they can be said to be rational or ‘correct’ in their decision making.


Having said this, an extension of logic, or its descriptive counterpart is called logicism. It is the attempt to answer how exactly human beings reason consistent with logic. Logicism’s foundations were first set by Frege, who grounded the idea on mathematics as the optimal manner in which to represent logical argumentation. His work cleared the way for Bertrand Russell to author Principia Mathematica – an attempt to create a logical basis for mathematics (Bertrand Russel, 1927).


Following the underlining premise of logicism, the Wason Selection Task (WST) was fashioned, as mentioned earlier, to asses whether or not human reasoning is consistent with the principles of logic. The WST “has often been claimed to be the single most investigated experimental paradigm in the psychology of reasoning” (Manktelow, 1999). The premise of the test follows a basic deductive reasoning process where subjects are given a conditional rule, something not unlike ‘if p then q’.

A conditional rendition of Aristotle’s syllogisms resembling ‘A’ and ‘B’. Subjects were then asked to consider four cards as presented under the claim: If a card has a vowel on one side, then there is an even number on the other side. Following this, participants were asked to flip over the card or cards necessary to tell whether the rule is true of false. Interestingly enough, the results stood in stark contrast to the premise of whether human beings use logic to reason. Very few participants in the experiment correctly performed the flipping of the cards.


In order to solve this test, one needs to use reason that stays consistent with logic, choosing the cards that will give an absolute conclusion, not just support for the rule but more importantly, to choose the card that will potentially reject the claim. This type of problem solving refers to persons using modus tollens (a way to reject the rule)rather than modus ponens (a way to support the rule).


In so far as ‘correctness’ lies under the normative framework of logic, it was concluded that not even 10% of the subjects had the correct solution as the majority only used modus ponens alone (Johnson-Laird & Wason, 1977). In essence, this test gave grounds to reject logic as the ‘correct’ normative theory, or for our purposes notion, from which human beings derive reason.


Consequently, to explain this behavior descriptive approaches came in for the rescue. In light of the results Wason argued that there is a sort confirmation bias occurring, in short meaning “an irrational tendency to search for, interpret or remember information in a way that confirms preconceptions or working hypotheses. It is a type of cognitive bias and a systematic error of inductive reasoning” (Miller, Vandome, & McBrewster, 2009). Moreover, Evans & Lynch argued shortly thereafter that regarding the WST, “subjects tend to match rather than alter named values when constructing verifying and falsifying cases of conditional rules” (Evans & Lynch, 1973).


Ultimately, this was an attempt to use descriptive theories so as to patch up any holes in classical logic as the archetypical normative approach to human reasoning. Besides, at their core these are heuristics that give a descriptive answer detached from any real normative theory of how humans should reason. Yet there is an alternative notion that once again can give a complementary way to back up logic in the WST, namely Evan’s analytic heuristic theory. The theory assumes that reasoning is facilitated by certain mental schemas that “maximize relevance to current goals” which inevitably might lead to logically relevant information being omitted (Evans, 2006). This Evans claimed, can account for any deviances from logic observed in the WST.


However, there is intuition behind such descriptive approaches. Heuristics are cognitive short-cuts that give slack to the effortful task of applying logic to reasoning. But then again having said this, it is important to point out that descriptive theories are weak in that they attempt to explain results post-hoc and can thus be altered once new information comes to light. This is why Cheng & Holyoak argued to loosen this rule, setting the stage for context-specific rules they dubbed pragmatic reasoning schemas. This idea carried with it another notion for ‘correcting’ the deviance in logic by WST.


In short, “people reason using knowledge structures […] which are generalized sets of rules defined in relation to classes of goals” which are important experiences, permissions, obligations and causations in our lives (Cheng & Holyoak, 1985). Using context as reference points Cheng and Holyoak were able to reexamine and adjust for the results seen in the WST, and hence added to the collection of ‘correct’ and “monotonic” (Chater & Oaksford, 1991) notions of classical logic in terms of card choices in the selection task.


Well, it seems as though all the required holes have been filled, as we have come to understand classical logic as the normative framework and applied particular heuristics as the lens through which subjects reason in the WST. Yet psychologists are relentless in their search for notions of correctness, leading to a reawakening of the rationality debate. Rationality as it was understood began to be related to optimality, as Anderson stated, “To be rational is to be optimally adapted to environment” (1991).


In order to understand rationality as an adaptation to the environment one must use probability to reduce the uncertainty of truth or entropy, to the rule associated with selecting cards in WST (Oaksford & Chater, 1994). This theory illustrated that reasoning using a Bayesian model of optimal data may actually be the normative framework that accounts for WST results.


Needless to say this is currently a widely accepted view arguing that human reasoning is consistent with the principles of classical probability theory. This in turn is a new direction for the rationality debate and consequently an alternative intuition into how the WST results can be seen as ‘correct’. Proponents of classical probability theory point towards odds in gambling, particularly to the Dutch Book Theorem. The idea is based on ship trading insurance companies in the 19th century who would set up odds so that no matter what would happen to their insured objects, they would make a profit. Being Dutch Booked in this sense translates to being set up for a loss regardless of the outcome. This is due to odds not being coherent under the axioms of classical probability theory.


One must come to a conclusion so far as to which normative framework and/or descriptive theory is the ‘correct’ one in relation to the reasoning behind the WST subjects’ irregular behaviour. We can derive some clarity and peace through context. In the context of testing whether the WST claim is correct or not, logic can be considered the ‘correct’ prescription to reason. Having said this, in a situation where one is describing human behaviour, probability theory is the more probable candidate. In the overall scheme of things, the big-picture if you will, well, it really depends on which prescription you subscribe to.

Get in touch...