Is boost the new nudge?

Here, I summarize the discussion on the normative differences between nudges and boosts presented in Sims and Müller, 2019 ¹.

Behavioral Economics studies the systematic biases in economic decisions that occur because our cognitive processes are constrained and, thus, context-neutral optimization is impracticable. This area of research started with Tversky and Kahneman (1974) ², and has influenced the proposal of nudge interventions, which seek to make the behavior as close to economically rational as possible. For instance, consider the choice between using energy conservation methods and not using them. The methods require taking a costly action today that will be more than paid for in the long run. The way in which this information is presented has important consequences: if the information framed as a loss (if you do not use energy conservation methods, you will lose €350 per year) people tend to be more responsive than if framed as a gain (if you use energy conservation methods, you will save €350 per year). In the second case, people tend to think that the initial costs are more relevant than the future savings. A nudge intervention will be to offer the information framed as a loss, understanding that the rational individual prefers using the energy conservation methods.

More recently, some authors have proposed a different type of intervention, based on simple heuristics (rather than on heuristics and biases, as nudges), that emphasizes the importance of an adaptive fit between the reasoner and her environment (Gigerenzer et al. 1999 ³). According to this approach, the intervention should not seek to condition the behavior of the individual, but to present the case in a way that better suits the individual heuristics or to educate the individual with new heuristics that are better suited to make the decision. This kind of intervention is called “boost”. E.g., medical doctors are more likely to make better inferences and provide better information to their patients if a premise containing statistical information is put in the form of “Ten out of every 1,000 women have breast cancer” rather than “The probability that a woman has breast cancer is 1%.” A boost will consist on either changing information about risk into a frequency format rather than a percentage or probability format which is less readily understood, or alternatively, on training people in conversion between either representation.

The nudge alters the choice architecture and conditions the behavior to be more rational, while the boost allows individuals to reflect better on the decision and come out with the rational behavior by themselves. Thus, it seems that boosts have better normative properties. Nudges appear to bypass rational deliberation in a way that is morally ambiguous and manipulative. Boosts, however, just restructure the environment so that it is better suited to the capacities of the individual or just add to the heuristic repertoire of the individual. Boosts require the individual’s active cooperation and, therefore, need to be explicit, visible, and transparent (Grüne-Yanoff and Hertwig, 2016, 2017 ⁴⁵).

Sims and Müller, 2019 [1] have a different opinion. To begin with, the association of transparency and autonomy-preserving with boosts may be a circular definition. An intervention that involves altering the environment of the decision-making so that it is better suited to the individual’s heuristics, like expressing the probability in terms of frequency in the example above, does not necessarily involve any need for motivation or cooperation on the part of the individual. If the involvement of the individual is included in the definition of boost, then the example will no longer be a boost. To avoid this problem, Sims and Müller consider two strategies for making the distinction between nudge and boost that contains normative aspects: (i) to define the category in an intuitive manner through the use of paradigmatic examples and distinctive properties, and (ii) to draw the distinction by means of the causal mechanisms by which each operates.

According to the first strategy, one should first assemble a collection of examples and extract a list of properties that set nudges and boosts apart. Then one should show that some of these properties make boosts immune to the relevant normative criticism. However, none of the paradigmatic traits attributed to boosts are able to entail the required respect for autonomy and transparency. For instance, consider the case of abstinence-only sexual education. This intervention assumes the desired goal to avoid STIs or unwanted pregnancy. The targeted individual is free to engage and cooperate, with the intervention. According to one argument, this cooperation and motivation should be enough to conclude that the intervention is transparent. However, if abstinence-only programs do not adequately teach about alternative methods of preventing unwanted consequences cannot be called transparent. This example shows that cooperation does not imply transparency. Finally, there are many doubts that interventions can clearly be classified as nudges or boosts. Take the example in which the individual is given the statistical information in terms of frequencies. This can be arguably seen as a nudge, just the same as when information is given in terms of losses rather than gains.

Following the second strategy, different authors have proposed different causal mechanism that distinguish nudges from boosts: (i) dual-process architecture vs. cognitive malleability, (ii) intervening on behaviors vs. intervening on competences, and (iii) the reversibility criterion.

The dual-process architecture proposes two cognitive systems, where one of these types, System 1, is fast, intuitive, and stereotyped in its operations; and where the other, System 2 is slow, deliberate, rule-governed, and places a heavy load on working memory. Allegedly, nudging operates on System 1 processes to deliver the decisions System 2 would make if it just had the time and ability. However, this view is at odds with the assumption that cognitive architecture is malleable, a view that proponents of boosting necessarily agree on, because the heuristics which are acted on by nudges are assumed to be stereotyped biases incapable of change. Thus, a classification based on this difference will assume a contradiction: boosters will be committed to saying that nudges are just a sort of ‘short-term boost’. Likewise, nudgers will be committed to saying that boosts are nothing more than a sort of ‘educative nudge’. Both will say the other is mistaken with respect to causal processes.

The other causal mechanisms have their own problems. There is no clear way to distinguish competence-change form behavior-change, and for the reversibility criterion, empirical research has found that nudging interventions may also have lasting effects.

Having discussed and rejected the two strategies, the authors do not argue that they reject a distinction between nudges and boosts, rather that the distinction cannot be the basis to answering normative questions in the study of BPP. Normative issues should be answered on a case-by-case basis, and not on whether the BPP is categorized as a nudge or a boost. To advance in this program, the authors end the article with two points for normative evaluations. One requires to argue for the special relevance of a particular theory of individual autonomy or, alternatively, to produce a normative evaluation of the heuristics on which BPP interventions rely. The other one must consider that the social context of the intervention may play a larger role in its evaluation.