In his book Explaining Social Behavior. More Nuts and Bolts for the Social Sciences Jon Elster argues that the interpretation and explanation of a work of art are closely related. He makes the further claim that "a successful work of art is one that can be given a rational-choice explanation" (p. 246). This is a bold claim, even when applied to (neo-) realist literature, to which Elster confines himself.
According to Elster to understand the actions of a fictional character is to construe a rational-choice account that explains them. Elster concedes that characters may exhibit weakness of will and other deviations from strict rationality. But as such their actions are nonetheless intelligible. Therefore "intelligibility rather than rationality is the most useful idea for the task of interpretation" (p. 247).
Elster goes on to claim that the author writing a classical drama or novel faces two tasks: to develop the plot through the actions of the characters and to maximize aesthetic value while doing so. The first task requires that the characters are intelligible, the second that the author is rational. Elster lists four demands that rationality imposes on the author: that the actions of the characters are intelligible, that everything that happens matters in some way, that there is a minimal appeal to accidents and coincidences and that the novel or drama as a whole offers a psychologically gratifying pattern of the buildup and resolution of tension.
Against this view one might argue that, obviously, as Elster himself admits, it only applies to (neo-) realist fiction and that the author’s intentions do not matter for the work’s interpretation. This is old wine. My own comment is that Elster does not extend the argument in the rest of his book to literature.
Rational choice theory assumes that individuals choose an action based on their preferences, the actions they could take and their expectations about the outcomes of those actions. It is further assumed that an individual’s set of preferences is complete, meaning that either A is preferred to B or B is preferred to A and if not they are equivalent and that preferences are transitive, meaning that if A is preferred to B and B is preferred to C then A is preferred to C. Given these assumptions a choice is rational if no other alternative is preferred over the chosen option. This makes sense. If someone were to prefer action A, but still do B, we might assume that something withheld him or her from doing A, but then A would not have been the preferred action given these constraints. In other words we would construe a situation which would make the chosen action appear rational.
While this sounds reasonable in theory, in practice people’s preferences need not be complete. One may prefer A over B, C and D but have no ranking of B, C and D. It is also possible to construct examples in which transitivity fails, for example when ranking preferences according to two of three different aspects. What’s more not all possible alternatives may be known and their outcome may be uncertain and contingent upon another individual’s actions.
Given these constraints it is unsurprising that in reality people frequently violate the principles of rationality. For example people evaluate potential losses and gains differently, they tend to be loss averse, are inconsistent in their time preferences and so on.
Now, if and insofar as literature is a representation of reality one would expect to find the same kind of deviations from rationality in fiction. If authors are prone to the same kind of deviations from rationality as the average person one would expect them to endow their characters with the same characteristics, for the simple reason that they themselves may not understand say conditional probability.
It could also be argued, by contrast, that on average fictional characters are more rational than one would expect based on empirical findings of how people actually behave. Fictional characters are purposefully crafted by their creator. Authors may use "intelligibility" as an implicit yardstick when judging the narrative and in so doing impose a greater sense of rationality on their characters. This is a hypothesis and I must say that I have never read a novel looking for examples of "predictably irrational" behaviour, to use the term coined by Dan Ariely.
Life itself is to a large extent governed by chance. Events that in real life strike us as a coincidence would seem forced in a novel. Reality is a complex system of interacting parts and feedback mechanisms. Most novels cut through these layers of complexity with a simple narrative, focusing on one causal chain while omitting other factors.
Thus my hypothesis would be that precisely because it endeavours to be rational and satisfy the conditions set out by Elster the realist novel fails to be realist. But of course there is still some deconstructionist blood running through my veins.