It is a well-established observation of behavioral psychology that choices – all sorts of choices, from relatively minor ones like which menu item to order at a restaurant to truly significant decisions about the management of money or health care – are subject to significant influence by external factors outside the control, or even the conscious awareness, of the people who make them. This field of study is relatively young, yet its insights have had a profound influence on business, policy, and economics, where one of the earliest pioneers in the field, Daniel Kahneman, won a Nobel Prize for his work usurping the predictive power of the long-held “rational maximizer of self-interest” model of human economic decision-making.

In response to the emergence of behavioral psychology, a new philosophy of government and private management is emerging, emphasizing the manipulation of choice architecture – that is, the context surrounding the presentation of a choice – in order to prompt people, without forcing them, to make choices that will enrich their lives or promote some other pro-social goal. The object is to find areas in which peoples’ everyday intuitions, default choices, or heuristics tend to conspire to lead people to make decisions that they would not if they were giving a full application of their reason, and change some or many aspects of the choice architecture to subtly push people toward the “better” choice. These subtle pushes, called nudges by Cass Sunstein, may take many forms: placing office amenities like bathrooms far away from offices, thus increasing the odds that people will run into one another the halls, in workplaces that want to encourage collaboration and collegiality; designing cafeteria lines to promote the selection of healthy options; providing sugary drinks to judges to combat the effects of cognitive fatigue on their decision-making. The result is to demonstrably promote welfare and improve behavior while preserving freedom of choice.

There are numerous attractive features to this model: nudges are typically cheap to implement, they feel natural, and their preservation of choice avoids problems related to coercion. But because they rely on exploiting errors in decision-making, nudges must rely on a deep knowledge of the human subconscious to be effective. After all, the judge who denies a higher and higher proportion of parole applications as lunchtime draws near is not doing so because she thinks this is a right and good way to make decisions about peoples’ lives, she does so because it takes more cognitive effort to deviate from a default action than to not, and the glucose she needs to make that cognitive effort has been depleted by a long morning of hard intellectual labor. Only by understanding this subconscious process can a nudge to correct the flaw influencing her decisions can be designed.

Thus, we come to a limitation of this method and the nexus to our class. Because nudging requires deep knowledge of subconscious processes, it can only take advantage of general subconscious errors, and moreover, only those general subconscious errors that are discoverable by the methods available to researchers: small–scale controlled experiments, limited real world observation. But with the advent of online tracking software, as more of our behavior moves into the digital realm, it has become not only plausible but probable if not certain that the idiosyncrasies of individual subconscious processes may become discoverable by observers who know what they’re looking for. It stands to reason that individually-targeted nudges might be even more effective than others, and therefore, peoples lives might be improved more completely and more quickly, which surely is a good goal of governments, producers of self-help products, or both.

Many objections to this concept spring to mind. Corporations and governments are not necessarily benevolent institutions and these tools in the hands of a bad actor would likely be at least as destructive as they would be positive in the hands of a good actor. Even assuming good faith, it is overly presumptuous to believe that a third party could know what’s best for another individual and so nudge them in positive directions. It is overly invasive to seek this sort of knowledge. All these and other criticisms are meritorious and worthy of careful discussion and consideration. However, there is one objection that I think is most interesting. That is: individually-targeted nudges are a violation of autonomy.

By altering the choice architecture around each individual, based on specific knowledge of the idiosyncrasies of that individual’s subconscious processes, we move from nudgers to puppet masters who have completely shattered the illusion of choice or free will from the targets of the nudge. It is one thing to know that generally, Chicago-area public school students will purchase 25% more produce with one cafeteria configuration than with the original; it is quite another to know that Jimmy Eckhart of Downer’s Grove Elementary wasn’t in a good mood this weekend so perhaps there should be a special treat – just one, the last one – for sale at a reduced price at eye level at the end of the cafeteria line when he goes through so that he’ll be 45% more likely to buy it, and therefore 32% less likely to throw a fit during the math class for which he probably did not do his homework on account of his mood.

This is a fascinating argument to me because it has a very strong intuitive appeal despite the fact that nothing has changed on the surface of the decision. If there is a cafeteria line, it must be laid in out some way, and it would be absurd to say that this fact destroys autonomy simply because the choice of how it is laid out will have some effect relative to all other possible layouts; to call it a destruction of autonomy would serve as a de facto denial that there can ever be autonomy, and thereby negate the point. But once the choice architecture begins changing to specifically influence each individual person, there is an inevitable sense of transgression.

I arrive at the conclusion of this essay with nothing firm to conclude: I very much like the potential of nudges to do good in the world; I think the objections to individually-targeted nudges have a good deal of merit, but I don’t believe any of them are knock-down arguments. However, I suspect that there is something important in the tension identified in the last paragraph; I hope to tease it out either in the revision to this essay or perhaps in a new essay next semester.

-- TomLawrence - 05 Jan 2015

2.0

A growing body of economic and political science literature focuses on the use of nudges, that is, the manipulation of choice architecture according to known tendencies of human behavior for the purpose of priming the subconscious processes of people to make them more likely to make certain choices in certain contexts. As this field is applied to government and regulation, it is called “libertarian paternalism;” the private-sector corollary has no unified name that I know of but examples pervade our lives, from urban planning, architecture and interior design to the massive fortunes being made by Google, Facebook, and others in the realm of targeted advertising. As the preceding sentence implies, this is a technique that may be used equally by those with good intentions – a government trying to make its citizens lives go better, a company trying match consumers with products they will genuinely like – or for less altruistic purposes.

Until recently, nudges could only be based on behavioral traits common to most people: the insights of behavioral science that allow the design of nudges were limited to those to researchers: small–scale controlled experiments, limited real world observation. But with the advent of large scale behavioral data collection as more of our behavior moves into the digital realm, it has become not only plausible but probable if not certain that the idiosyncrasies of individual subconscious processes may become discoverable by observers who know what they’re looking for. It stands to reason that individually-targeted nudges might be even more effective than others and therefore invite serious questions about our traditional conceptions of liberty and autonomy. If an entity can influence our behavior with a high degree of efficacy because it knows more about us than we know about ourselves, what do we mean by choice, what do we mean by coercion, and what intervention from government, if any, is appropriate?

Black’s Law Dictionary gives the following definition of “coercion”: “Compulsion; force; duress. It may be … implied, where the relation of the parties is such that one is under subjection to the other, and is thereby constrained to do what his free will would refuse.” Is this applicable to nudges? Take a classic demonstration of nudging, the study of how the layout of cafeteria lines affected the purchasing behavior of Chicago-area public school students. In this study, cafeteria officials found that they could affect the purchasing of certain foods by up to 25% in either direction. So, for simplicity, let us assume that cafeteria officials are benevolent and arrange their cafeteria lines to promote apples at the expense of Doritos chips. If apple sales rise by 25% and Doritos sales fall by a corresponding 25%, it stands to reason that there must be students, perhaps individually identifiable, who are now purchasing apples where their will would previously have led them to buy Doritos. Have those students been coerced? Certainly the students are under subjection to the cafeteria officials, in that the officials make decisions about how the cafeteria is laid out and the students had no say. The promotion of buying apples also corresponds to the will of the officials (that the students should eat healthily) rather than the will of the students (who might reasonably be expected to choose Doritos given a direct choice). That a student could, if she really wanted to and expended the effort to find them wherever they were situated in the line, still choose Doritos seems not to remove the case from the realm of coercion; many other examples of coercion allow the coerced to go with their will if they are willing to accept, e.g., a financial penalty that they could in principle afford to pay. Is the question thus settled? It would seem so, until we consider the following.

What if the nudge, by showing how choices can be changed by the context in which they are made, is telling us something deeper about how we choose? What if we think of the student not as having will to buy Doritos that may be violated by cafeteria line design, but rather think of the student as having the will to move through the cafeteria line quickly and efficiently, to stay within his allotted budget, and to meet these goals by buying any foods belonging to a larger reference class of foods the student has some baseline appreciation for, with the actual selections made according to what draws the student’s attention? In this view, both apples and Doritos are members of the reference class, and so the student is actually roughly indifferent between them. This view is challenging because it does not correspond well to subjective experience. Yet it does seem to be correct: it not only explains why students’ choices can be influenced among apples and Doritos, but also why we would not expect, for instance, to learn the nudge could be used to make students purchase plates of bubbling green-brown mystery sludge.

When we take this view – that nudges do not violate our will to eat Doritos by making us eat apples, but rather shift probability mass among our existing will to eat Doritos, apples, and any number of other things – we can answer “are nudges coercive” with a confident “no.” Even when a nudge is individually targeted and crafted according to deep knowledge about our behaviors unknown even to us, the preservation of choice means our will cannot be violated completely. This is positive because it would be impossible to ban the practice: even if data mining were forever stopped, choice architecture is inescapable. But still, we may be unsettled: what to do about those who would nudge us to eat Doritos rather than apples?

Just as allowing more speech is taken to be corrective of bad speech, here too the answer seems to be “allow more nudging.” We should work to empower more creators of choice architecture, so that people may make meta-choices about which architecture they will use and therefore how their first-order choices will be influenced.

-- TomLawrence - 01 Apr 2015