On June 8, 2017, I was part of a Roundtable at the Conference #NudgeInFrance held in Paris by Nudge France on 'The Limits of Nudge: Ethics & Manipulation'. Together with Cass Sunstein, Anne-Lise Sibony and Mariam Chammat we had a lively discussion led by Elsa Savourey from NudgeFrance. Prior to the panel we were send some questions for reflections, my notes on which I thought was worth sharing for those who couldn't attend, or was too busy listening to make their own notes. The notes will be shared over the next couple of days.
Elsa: As an introductory question: is nudge always “for good”? Are there nudges "for bad"? Are such considerations in relation to good or bad nudges integrated in the definition of what is a nudge?
Pelle: A nudge is defined as a psychological function of any attempt at influencing people’s judgment, choice or behaviour in a predictable way, that is (1) made possible because of cognitive limitations, biases, routines, and habits in individual and social decision-making posing barriers for people to perform rationally in their own self-declared interests, and which (2) works by making use of those limitations, biases, routines, and habits as integral parts of such attempts, see (Hansen 2016). In particular, this implies that a nudge is intended to cause a behavioural effect independently of (i) forbidding or adding any rationally relevant choice options; (ii) changing incentives, whether regarded in terms of time, trouble, social sanctions, economic and so forth, or (iii) providing hitherto unkown factual information and rational argumentation.
From this it is seen that the definition does not imply any normative criteria for an aspect of an intervention to qualify as a nudge. It includes interventions that may usually be evaluated as “nudges for good”, e.g. decreasing the unit size of cakes served at a buffet to reduce intake of sugar and fat; as well as interventions that may usually be evaluated as “for bad”, e.g. increasing the unit size of cakes served at a buffet to increase intake of sugar and fat. Both interventions utilize behaviour changing functions that qualify as nudges.
Elsa: Nudge plays to mostly unconscious factors. Aren't people who are nudged just manipulated into a given choice?
Pelle: First, I have to say that I’m not sure that I agree with the premise of this question. I think it is a widespread misunderstanding that Nudges mostly play on unconscious factors. It’s not a coincidence that many people tend to think this. First of all, many academics talk straightforward about cognitive bias and heuristics as unconsciously influencing our behaviour. Second, some of the most memorable behavioural insights from behavioural economics and cognitive- and social psychology are those that really surprise and fascinate us – and those insights that really surprise and fascinate us are typically those that influences our behaviour in a way we may label “unconscious”. Hence, due to availability bias, they are the examples that come to our mind when we happen to think about what characterizes nudges.
However, from a scientific point of view we should be more careful when we describe how nudges work. As Cass described it in his keynote, dual process theories often explain their findings in terms of system 1 and system 2 cognitive processes. System 2 processes are what we call ‘reflective’, and as such they are consciously accessible to our inner experience by definition. This leads many people to believe that System 1 processes must be unconscious. But if you listened carefully to Cass he did not characterise system 1 processes as unconscious as such, but – quite correctly – characterized system 1 processes as automatic and intuitive, where automatic means non-intentional. And here is the important point: that cognitive processes are automatic and intuitive doesn’t imply that they are unconscious in the sense that they are invisible to our ‘inner eye’.
The substitution heuristic provides an example of this. We often answer a complex question by answering a similar, but simpler one. On an everyday basis this usually works quite fine for our purposes, but it always comes at the cost of precision and correctness. Now, while the substitution heuristic is automatic, we can actually recognise it from our experience; and knowing of the heuristic and attending to our thinking we may even experience it as it happens and decide to reject the result of our own thinking. It’s not that it’s possible for us to do this all the time. It requires way too much effort and attention. But it is possible, which shows us, that we can actually experience, monitor and sometimes even override some of our automatic processes. They are not necessarily ‘unconscious’, but they are often non-conscious.
So that was the premise. What about the question? "Aren't people who are nudged just manipulated into a given choice?"
Now, we just drew a distinction between automatic and reflective processes based on Kahneman’s Dual Systems Theory. Following our earlier paper “Nudge and The Manipulation of Choice” (Hansen & Jespersen 2013) this distinction may also be used to distinguish between ‘choices’ and ‘behaviour’, such that choices are by necessity behaviours resulting from reflective processes, while behaviours caused mainly by automatic processes cannot be described as such (even though this is usually done in the micro-economic lingo that underpins most discussions of nudges and behavioural insights).
Of course, this is a conceptual distinction and thus it runs the danger of simplifying too much. Yet, I believe that it does improve what we can articulate about the ethics of nudge, since this distinction makes it clear that not all nudges are aimed at influencing choices as such.
From an ethical perspective this is an important point. Core concepts of ethics such as ‘responsibility’ presuppose agency and are thus tied to the category of choices. From an ethical point of view then, nudges aimed at influencing choices should be carefully considered as they result in those nudged obtaining responsibilities. In particular, if nudges are about manipulation, then we have a problem since this would mean that we could be manipulating people into making choices for which they, and not us, have the responsibility. Yet, it should be noticed from the outset, that nudges aimed at influencing choices, although most relevant from an ethical point of view, are only a sub-category of nudges aimed at influencing behaviours more generally and as such any conclusions pertaining to the former does not necessarily generalise to the latter.
Still, the main question hinges on whether nudges are manipulative as such. In order to answer that question we need to draw an additional distinction.
This is the distinction between what we have called epistemic transparent and non-transparent nudges; a distinction aimed at distinguishing between the manipulative uses of nudges from other types of uses, see (Hansen & Jespersen 2013). Also, it is coined to capture the general intuition referred to by Thaler and Sunstein’s when they say that an attempt at influencing other people’s behaviour, including choices, may be objectionable, “because it is invisible and thus impossible to monitor” (Thaler & Sunstein 2008, 246).
Revising our 2013 paper a bit, I now define a transparent nudge as any nudge part of an intervention provided in such a way so that those affected can infer three things about the intervention as a result of the intervention:
A non-transparent nudge, on the other hand, will be defined as a nudge working in a way that those nudged cannot reconstruct the architect, intention, or the means by which behavioural change is pursued.
A couple of points for those of you who want to dig further into this definition: The notion of transparency defined here is very close to what Luc Bovens refers to as “token interference transparency” in (Bovens 2008) although the present notion is more specific. Also, for those of you interested in ethics and philosophy, you may notice, that since subjects should be able to infer these three things as a result of the intervention, a transparent nudge is defined in such a way that it qualifies as a basic speech act providing what (Austin 1962) called ‘uptake’. Finally, for those of you more interested in ethics and the applied aspects of behavioural science, you should notice that epistemic transparency may vary according to the epistemic competencies of those nudged, i.e. epistemic transparency is defined as an empirical, and not a normative concept.
Returning to the argument: defined as such, a transparent nudge is by necessity not a case of manipulation, when we understand by “manipulation” the psychological sense of the concept that is relevant for a discussion of ethics; that is, manipulation in the sense of intending to change the perception, choices or behaviour of others through underhanded deceptive, or even abusive tactics.
Thus, to the extent that a nudge is transparent people are not being manipulated into a behaviour, let alone a choice.
Elsa: How then do we determine whether a nudge is ethical or not?
The answer to this question will be posted in the near future. Sign up for news, if you don't want to miss it.
Austin, J.L. (1962) How to Do Things With Words. Cambridge (Mass.) 1962, paperback: Harvard University Press, 2nd edition, 2005.
Luc Bovens (2008) 'The Ethics of Nudge', in Till Grüne-Yanoff and Sven O. Hansson (eds) Preference Change: Approaches from Philosophy, Economics and Psychology (Berlin and New York: Springer, Theo- ry and Decision Library A, 2008).
Richard Thaler & Cass Sunstein (2008) Nudge – Improving Decisions about Health, Wealth and Happiness (New Haven, CT: Yale University Press 2008)