Should we be rational?


One of the fundamental methodological issues in research on human behaviour is whether we should take rationality as a norm. This is not the same as studying rationality, such as in game theory. It is about whether rationality should be a tool of measurement to identify causal mechanisms that in fact guide human behaviour. A foremost example is the work of the Nobel laureate Daniel Kahneman. For most of Kahneman’s work, the science of statistics serves as a litmus test to identify human failures of decision making. Indeed, in many circumstances we are very poor statisticians, and we rely on wrong arguments in justifying our actions, even defending them against contrary evidence. In this respect, Kahneman’s work also obtains critical value and merit in our modern societies governed by experts who might claim epistemic privilege where there is none.

However, I think that this approach is fundamentally misguided in relying on rationality as a norm structuring empirical evidence. Kahneman does not identify ‘mechanisms’, in the sense of our project. Even worse, he creates misplaced essentialist re-descriptions of his evidence in terms of two ‘systems’, the system 1 that builds on intuition, and the system 2 that builds on deliberation and reflection. System 1 does not know statistics, system 2 can learn and apply it (but is ‘lazy’ in doing that…).

To be fair, in his bestselling book ‘Thinking, Fast and Slow’ Kahneman emphasizes again and again, that he does not think of ‘systems’ as being more than mere metaphors, and he does not make any empirical claims that they might have material embodiments in the brain. But at the same time, he continues with what appears to be a mere stylistic shorthand, namely treating the systems as causal agents in his descriptions of human behaviour. But this shorthand turns real, as the reader, and probably he himself, are always pulled into the orbit of thinking of ‘systems’ as causes. This shows what is fatally missing in this book, namely a genuine causal explanation of the phenomena in question.

If we take rationality as a norm, we end up with implicitly partitioning human individuals into a ‘rational’ and a ‘non-rational’ part, and we even would adopt the conclusion that the rational part should take control, if necessary, externally imposed, because human welfare would increase (as measured, of course, by rational criteria). This is the idea of ‘nudging’, and it raises important ethical and political concerns. I think that these debates start too late, because the real issue is the methodological one, whether rationality can serve as a litmus test for identifying real properties that humans have. This is different from the question whether, when designing institutions and policies, we should follow norms of rationality that are established by some philosophical or scientific thinking.

The notion of the two ‘systems’ is methodologically flawed because it falsely suggests the existence of a ‘mechanism’ type construct that can be causally relevant without actually identifying and describing one. If we take the notion of ‘mechanism’ seriously, it is straightforward to see that mechanisms are not ‘good’ or ‘bad’, but that they are our causal constructs to explain observed behaviour. Whether this behaviour is ‘rational’ in the sense of meeting standards of optimality in certain contexts of decision making, does only matter beyond the domain of research, and might be a topic of moral or political philosophy. To put it bluntly, I think that ‘rationality’ is not a scientific concept at all.

Let me give you one example taken from the experience of Witten/Herdecke University, one of the project partners of INSOSCI. The university applies a complicated system of interviews admitting students and claims that this is a pillar of its excellence as an academic institution. In Kahneman’s view, and he has important related research on that, this is a self-conceit, and he warns vehemently against such procedures, pointing to a lot of evidence that shows that such procedures cannot be better than a simple checklist approach based on statistically relevant categories. Indeed, I remember that twenty years ago, when I was member of the faculty, a colleague applied some statistical test on the results of the admission process, and showed that the process, more or less, generates similar results as a checklist approach that would just look at high school performance and some other indicators. So, should the university just scrap the procedure?

The problem is that Kahneman’s argument takes a certain goal of the procedure for granted, which is the selection of the best students. However, the procedure has also other aspects, which might not even be explicit. For example, it is the first step in the socialization process of the university, imbuing successful applicants with the feeling of being ‘the chosen ones’. They receive appreciation from mixed teams of experts, often including important figures from business and society. For the rejected students, it can be a source of inspiration, since people learn from the interviews also about themselves. So, the procedure is not one-dimensional, just focusing on the selection issue in narrow terms.

This fact of multi-dimensionality seriously limits the comparison between experiments and real-world settings. One danger is that important causal factors may be seen in a biased way. In previous blogs, I pointed to the role of narratives. In Kahneman’s work, narratives are only seen as misguiding ‘system 1’ via obstructive impact on ‘system 2’. The Witten/Herdecke university admission process is a part of a series of narratives about the educational process at that university, and contributes to the construction of personal narratives by which the applicants make sense of their careers. Clearly, this needs to be distinguished from the question whether these applicants will be successful in these careers, according to some societal standard. Kahneman is right in emphasizing that these careers are shaped in highly complex environments manifesting many random forces. For predicting this, statistics is the most powerful tool. But the question is, whether what is ‘rational’ is reasonable, in terms of the meanings of life that institutions such as a university should shape.

Funded by: