Reflections on the reflective system


After returning from Kasnäs, I delved into the literature recommended by Professor Wiers (see my previous blog entry). A central point is the distinction between the two processes or systems (unfortunately, the literature often tends to use the two terms interchangeably, although I think there are fundamental differences between ‘process’ and ‘system’ in ontological terms). In this post, I present some principled reservations about the notion of ‘reflective system’ that is developed in the most influential model by Strack and Deutsch, which is the standard in psychology, but less well known in behavioural economics, which mostly follows Kahneman. Yet, for our purposes the models are sufficiently similar to validate my criticism for both. In a nutshell, the model assumes that there is a parallel processing by stimuli in the impulsive and the reflective system, and that they have different process characteristics, such as the former mainly building on associations in eliciting behaviour, and the latter on propositional judgments. The Strack and Deutsch model has some important specific features, such as not requiring consciousness for the reflective system.

In this post, I just summarize what I perceive as major flaws of the model.

  1. What is ‘reflection’? There is no clear definition of what ‘reflective’ means, and in fact there is a blurred relationship with ‘consciousness’. Reflection seems to refer to a peculiar kind of knowledge, namely which is propositional and has a truth value (which appears to be conscious, after all). I think that this is far too general. If I take Strack and Deutsch’s example of the recognition of a person as ‘elderly’, this can be both active in the two systems, and in the reflective system it is a proposition. I do not think that this is ‘reflective’, but just represents a state of the world. In my view (and many of the philosophical literature), ‘reflexivity’ would refer to more specific processes, in particular, firstly, self-reference of a subject, and secondly, the construction of meta-level propositions and judgements. For example, ‘this is an elderly person’ would not be reflective, but ‘I feel doubts that this is an elderly person’ is reflective. There are good arguments to assume that this specific kind of reflexivity is unique for humans and evolved for the evolutionary reason of enabling complex cooperation, such as mutual taking of roles and understanding intentions of others. This is Mead’s ‘me’, and Mead suggest just this understanding of reflexivity, namely the reflexivity that emerges when we view ourselves through the eyes of others.
  2. Language seems to be the medium in which propositions are formed. But even if we extend beyond language and include broader types of symbolic representation, it seems highly questionable to relate reflective processes only with propositions that have a truth value. This ignores crucial insights of the philosophy of ordinary language that introduced speech act theory and other approaches half a century ago. Reflexivity is also involved in all kinds of utterances that do not and cannot have truth value, such as giving promises or declaring intentions. In fact, I think that for understanding human action propositions with truth value are just tools. One extremely important issue is how reflexivity relates with identity. If a reflective agent takes the decision to start exercising in order to become a healthy person, the inner speech that represents that intention does not have a truth value, but aims at making something true (mind-to-world direction of fit, in Searle’s parlance). Trickier, we could consider the truth value of that representation of an intention (“Do I really want to become a healthy person?”). So, the dual process model, in being based on the truth criterion, would not be able to deal with essential aspects of reflexivity, such as self-doubt or self-deception. In fact, it generalizes a very special kind of reflexivity that emerged in modern scientific thinking, while at the same time also over-extending its non-reflective contents.
  3. The issue of identity has an even more fundamental significance. The Strack and Deutsch model approaches reflection as being based on inferential knowledge. Although the authors often refer to complex aspects of reference being constituted by processes in the impulsive system (such as meaning related to sensorimotor schemata), separating the reflective system in this way seems to create a far too narrow systems boundary. There are many approaches to cognition (such as ‘conceptual blending’ or ‘metaphor’) that would put non-inferential modes of thinking in the first place. This is especially important in the context of establishing identities of objects in dealing with the world, and, of course, the social world, in particular. For example, in Professor Wiers’ field of empirical research, addiction, for creating a specific motivational pattern, it is essential how people perceive themselves. I claim that if we put at least equal emphasis on non-inferential modes of thinking, the clear-cut distinction between the impulsive and the reflective system collapses.

Let me stop here. I think, the direction of my argument is obvious. Even if we consider a dual process model, we need a much more precise idea of what ‘reflective’ means, and in which modes it operates. This can be clarified in setting up more detailed mechanistic models of reflective processes. That might include, as I argued in my previous post, ‘fast reflective processes’ that closely interact with ‘impulsive processes’. In fact, I expect that once a fully-fledged mechanistic model of that interaction is in place, we can discard the duality of systems as a mere heuristic device.

Strack, Fritz, and Roland Deutsch. “Reflective and Impulsive Determinants of Social Behavior.” Personality and Social Psychology Review 8, no. 3 (August 2004): 220–47. doi:10.1207/s15327957pspr0803_1.


Funded by: