Abstract
The article introduces the concept of the imaginary of behavioral governing to capture the view on the role of behavioral research in governing behavior that is widely shared in the academic and public discussions about behavioral policy (nudging), including the recent debates about reliance on big data and algorithms to influence people's behavior. It is believed that behavioral science provides knowledge of stable regularities of behavior and of the cognitive processes that lead to them, and that policymakers/governments act upon this knowledge to change behavior of individuals. I argue that this set of claims about the knowledge provided by the behavioral sciences is not substantiated in behavioral research. The formal theoretical frameworks of behavioral science come to be interpreted — via the imaginary of behavioral governing — as relating to human agents that power can act upon. I reflect on the possible point of entry for critique of this imaginary and its effects.
The US public health crisis in Flint, Michigan, that started in 2016 after the city's water supply was contaminated with lead has become an example not only of the incompetency and irresponsibility of public authorities, problems with the country's aging infrastructure, and environmental racism, but also of effective civic activism. Flint is less known as a site where the Social and Behavioral Sciences Team under the Obama administration tried to bring in insights from behavioral science to address some aspects of the crisis. The team directed by Maya Shankar planned to use subtle forms of peer pressure to encourage handwashing and to remind people to change water filters distributed by the local community. They also sought to deploy new strategies of framing information to deal with misinformation about the quality of water (Stillman 2017). These simple interventions were inspired by scientific knowledge of how people process information based on cognitive and social psychology. The Social and Behavioral Sciences Team was dismantled after President Obama left the White House in early 2017 and its behavioral science interventions were not introduced in Flint. Nonetheless, they remain a good illustration of how behavioral experts think about addressing societal challenges.
The Rise of Behavioral Policy and Nudging
The idea of bringing behavioral science to policy gained momentum in 2008 after the publication of the book Nudge by the Nobel Prize–winning behavioral economist Richard Thaler and the Harvard University professor of law Cass Sunstein (Thaler and Sunstein 2008). Behavioral policymaking aims to approach economic and societal issues with the help of policy solutions justified by behavioral science (e.g., Shafir 2012; Oliver 2013). It relies on the behavioral sciences (mostly cognitive psychology and behavioral economics) to diagnose and address societal challenges such as poverty, unsustainable consumption patterns, social interactions during pandemics, obesity, and decreasing retirement savings. Behavioral policy is aligned with the so-called evidence-based approach to policymaking.1 Cognitive psychology and behavioral economics are supposed to provide evidence about cognitive processes to diagnose causes of societal challenges. Research on people's limited attention, reactions to framing of information, and intuitive judgments about probability, which lead to systematic behavioral patterns often called biases (such as loss aversion or status quo bias), is also used to design policy interventions, sometimes called nudges. Nudges are a public policy technology that allows people's behavior to be guided by small changes in the contexts in which they make choices—changes in the “choice architecture.” Nudging is defended, normatively and politically, from a position called libertarian paternalism. The claim is that the shaping of people's choices by governments is justifiable when it preserves freedom of choice (that is, if one can easily abstain from nudging) and when it brings welfare-promoting outcomes. This approach to policymaking has been implemented across the globe by state and public-private organizations: from the United Kingdom's Behavioral Insights Team and the Social and Behavioral Sciences Team in the United States, via nongovernmental and nonprofit initiative Nudge Lebanon to the Government of Singapore's Behavioral Insights Unit.
Behavioral policy has been already extensively studied and discussed. Scholars view the rise of nudging, with its focus on changing individual behaviors as the target of public and social policies, as part of the neoliberalization of the state (Leggett 2014; Jones, Pykett, and Whitehead 2013; Wirth 2020). Others suggest it is a technocratic approach advanced by “nudge” experts invoking the authority of behavioral research to gain influence in the policy context (Strassheim, Jung, and Korinek 2020). Foucauldian scholars analyze behavioral policy as the biopolitical project par excellence, with its declared interest in increasing the “wealth, health, and happiness” of populations (Jones, Pykett, and Whitehead 2013; McMahon 2015). When applied in the Global South, along with evidence-based policies, it promotes a version of modernity along with material power relations extending from the colonial era (Berndt 2019; Greenhalgh and Russell 2009). We also know that nudging is being employed by several types of organizations (states, charities, universities, private companies) and that recently the architects of these technologies of behavior are increasingly utilizing big data and AI tools to infer information about patterns in people's behavior (Yeung 2017; Whitman 2020). Accordingly, behavioral policy and nudge have been criticized for perpetuating neoliberalization of societies and polities; for their incompatibility with democratic legal institutions (Lepenies and Małecka 2015); for manipulation of citizens’ choices (Bovens 2009) and infringement of their autonomy (Cohen 2013). Acknowledging many of these criticisms, still other authors argue that behavioral policy is not necessarily an ally of technocracy, neoliberalism, or new managerialism, but could also be aligned with democratic projects of self-governing (John, Smith, and Stoker 2009).
Defining the Imaginary of Behavioral Governing
This scholarly reflection allows us to grasp and critically assess diverse aspects of behavioral policy and nudge. However, the debates take for granted claims about behavioral science made by proponents of nudging: it is presumed that cognitive psychology and behavioral economics identify robust behavioral patterns caused by cognitive processes. It is also widely accepted in the discussion that behavioral policy, based on behavioral science, can produce specific and targeted outcomes. Neither of these presumptions is warranted. In what follows, I argue that the knowledge provided by behavioral science does not justify the claims that are made about applicability of behavioral research to policymaking. I show that behavioral policy and a large part of the debate operate within what I call the imaginary of behavioral governing. This imaginary obfuscates understanding of what can and cannot be known on the basis of behavioral research. I elaborate on this concept, suggesting that criticism of the behavioral sciences, which are used to justify behavioral policies, may prove a fruitful critical strategy for challenging proposals of behavioral public policymaking and nudge.
The concept of the imaginary of behavioral governing that I introduce in this essay is inspired partly by the work of Michel Foucault on governmentality and partly by the analysis of science and technology (STS) scholars who rely on the notion of sociotechnical imaginary to study mobilization of science in political projects. Let me start by explaining the Foucauldian inspiration. By governmentality Foucault meant the emergence of new practices of governing that marked the shift from feudalism “into the administrative state during the fifteenth and sixteenth centuries” in Western Europe (Foucault 2009: 144) and have continued throughout Western modernity. Governmentality refers not only to this historical shift but also to the form of governing constituted by the relationship between power and knowledge. Governing or government is understood by Foucault as the “conduct of conduct”: shaping, guiding, affecting the conduct of some person or persons, based on knowledge of who and what are to be governed (Dean 2010). The modern technique of power is characterized by forming a specific type of subjectivity through “the development of knowledge of man around two roles: one, globalizing and quantitative, concerning the population, the other, analytical, concerning the individual” (Foucault 1982: 784). This technique was born within liberalism and has been carried out by neoliberalism (Foucault 2008).2 The originality of Foucault's analysis lies in understanding liberalism as a way of governing and not as an economic theory or political ideology. In the so-called governmentality lecture Foucault made the following remark:3
In contrast to sovereignty, government has as its purpose not the act of government itself, but the welfare of the population, the improvement of its condition, the increase of its wealth, longevity, health, etc.; and the means that the government uses to attain these ends are themselves all in some sense immanent to the population; it is the population itself on which government will act either directly through large-scale campaigns, or indirectly through techniques that will make possible, without the full awareness of the people, the stimulation of birth rates, the directing of the flow of population into certain regions or activities, etc. (Foucault [1978] 1991: 100)
Note that the indirect techniques “directing . . . the flow of population” “without the full awareness of the people” are exactly how nudging techniques are defined, though in the depoliticized terminology of policy's behavioral influence.
Nudges and behavioral policies influence, direct, and govern people in a particular way. Thus, they are a way of exercising power. The aim of nudging is to conduct behaviors of people in a socially desirable direction backed by a policymaker, but also in a direction people would wish for themselves. Hence, nudging is a form of governing for the well-being and health of the population based on knowledge about both individualizing processes that concern individual preferences and their heterogeneity, and totalizing processes that concern the entire population. Behavioral science purports to provide knowledge about behavioral regularities at the level of population (such as loss aversion, framing effect, status quo bias) and explanations of these regularities by the cognitive processes at the individual level, the “psychological principles that underlie human behavior” (Thaler and Sunstein 2008: 112).
This view of behavioral research's role in influencing or governing behavior is widely shared by proponents of behavioral policy and its critics. It is believed that behavioral science provides knowledge of stable regularities of behavior and the cognitive processes that produce them, and that policymakers or governments act on this knowledge to change or govern the behavior of individuals. I call this idea the imaginary of behavioral governing. The concept of the imaginary is often used by STS scholars to analyze how science and technology are invoked to define and design political and social projects. For instance, they have introduced the notion of sociotechnical imaginaries: “Collectively held, institutionally stabilized, and publicly performed visions of desirable futures, animated by shared understandings of forms of social life and social order attainable through, and supportive of, advances in science and technology” (Jasanoff 2015: 4). One of the seminal examples of the sociotechnical imaginaries is the redefinition of nationhood in the United States and South Korea through the relations of these states to nuclear power (Jasanoff and Kim 2013).
Behavioral Policy Seen through the Lens of the Imaginary of Behavioral Governing
Clearly, nudging as part of behavioral policy is a political project defined by references to the current state of the art in behavioral research. It is imagined that behavioral science offers knowledge about population (behavioral patterns) and individuals (cognitive mechanisms), which can be used to influence (govern) behavior to achieve policy and societal goals, such as welfare, well-being, or health. As we have seen, this vision of public policy, its goal, and how it “works” is almost a paraphrase of Foucault's work on governmentality, in which he identified an imaginary of governing individuals through knowledge that appeared in modern Western Europe.4 Behavioral policy, as well as the academic and public debates about it, operate within this imaginary. Even though most discussions about nudging are not formulated in the Foucauldian terminology, they all presume that behavioral science provides knowledge to influence (govern) the behaviors of individuals. However, it has not been recognized yet in the discussions, including in STS and Foucauldian scholarship, that this imaginary—the view of how policy (power) and behavioral science (knowledge) relate to each other—simplifies the understanding of what can and cannot be known on the basis of behavioral research.
The behavioral sciences of nudging policies have roots in the cognitive revolution in psychology of the 1970s, particularly in the interdisciplinary research settings that merged computer science, cybernetics, and cognitive science and were funded by US military research agencies during the Cold War era (e.g., Edwards 1996; Erickson et al. 2013). An offspring of the cognitive revolution, cognitive psychology presumes that the human mind is analogous to the computer. Its theories model cognitive processes as processes of information processing (Miller 2002), but they are not causal theories about how cognitive factors influence behavior (Eronen 2020). This is contrary to the way these findings, in particular the work of Tversky and Kahneman (Kahneman and Tversky 1979; Tversky and Kahneman 1974), are being interpreted in the debates on behavioral policy and nudging. Furthermore, the behavioral patterns studied are less robust than typically claimed. For instance, the phenomenon of loss aversion, defined as “the aggravation that one experiences in losing a sum of money appears to be greater than pleasure associated with gaining the same amount” (Kahneman and Tversky 1979: 273), has been called particularly robust by proponents of behavioral economics, but the robustness has been questioned by researchers who point out that the phenomenon is highly dependent on experimental design (Małecka 2021).
What does this mean? The project of behavioral policy and nudge rests on a set of widely accepted claims about the knowledge provided by the behavioral sciences that is not substantiated in behavioral research (Małecka 2021). Behavioral policies are promoted, discussed, and criticized as if we really had the knowledge that the proponents of behavioral policy claim to have. Relying on behavioral science to design nudges is justified by the ability of policymakers to influence behaviors by mobilizing the cognitive processes. Critics of nudging argue that behavioral policy may manipulate behaviors by influencing cognitive mechanisms of individuals’ minds or brains. However, behavioral research does not offer such knowledge: it is much less robust than typically claimed, and importantly, it does not identify causal psychological mechanisms. It may still be the case that people are manipulated by nudge policies, but a mechanism through which this happens must be different from the one imagined in the debates about nudge policy.
The imaginary of behavioral governing influences the understanding and interpretation of scientific knowledge invoked for the sake of governing (nudging) behaviors by states and other organizations. The formal theoretical frameworks of behavioral science that provide procedures of decision-making come to be interpreted as relating to human agents which power can act on. For instance, the research of Tversky and Kahneman is interpreted as studying cognitive processes underlying behavior and as providing a “realistic” account of human cognition and decision-making. As noted above, however, this research assumes an analogy between a human mind and a computer. Thus, it comes up with abstract, highly formal, algorithmic-like procedures of decision-making. There is nothing straightforwardly “realistic” nor “human” about the research. It produces formal, abstract architectures of decision-making that are very difficult to interpret as causal mechanisms of the human mind (Marr 1982).5 When we look closely at the details of the formal frameworks of this research, we notice procedures of decision-making that can be performed by the computer rather than agency of biased human subjects. The imaginary of behavioral governing provides the lenses through which knowledge produced by behavioral science is interpreted as applicable to human agents. It also obfuscates the Cold War military legacy of behavioral economics and cognitive psychology (Sent 2000; Mirowski 2002).
Recently, the proponents of nudges, such as Cass Sunstein, have begun to elaborate on algorithms as tools of governing (their term!) by an administrative state (Sunstein 2022). One could think that Sunstein's interest in this topic is just jumping on the bandwagon of the hype of AI, algorithms, big data—and their policy relevance. This may well be the case. Yet once we understand the links between cognitive psychology, behavioral economics, cybernetics, computer science, and research on AI, such a move may seem less surprising. All these sciences have Cold War origins and were influenced by information sciences and information technology. In light of this, the recent merging of nudging with behavioral data science and algorithmic tools is a development that has a common basis in the history of behavioral science, as well as in the interest of controlling, steering, and influencing behaviors that have facilitated the emergence of behavioral research since the beginning. Note that the contemporary discussion about the role of algorithms in generating knowledge about stable behavioral patterns and the possibility of big tech companies or states to use this knowledge to influence people's behavior operates also within the imaginary of behavioral governing: knowledge about patterns in behavior is supposed to help steer the behavior of individuals, this time with the mediation of AI technology. Such use of big data and algorithms to influence behavior is often called automated influence (Benn and Lazar 2022). There are many ways in which these proposals and developments are criticized (e.g., Eubanks 2018; Barocas and Selbst 2016; Benn and Lazar 2022). I believe it is worth considering whether an understanding of the knowledge about behavior obtained by algorithms is not simplified and obfuscated as in the cases of cognitive psychology and behavioral economics, and for the same reason: it may get interpreted via the imaginary of behavioral governing.
What Next? Critique of the Imaginary and Critique of Behavioral Science
What is the possible point of entry for critique of this imaginary and its effects? One way to start is by analyzing the kind of knowledge actually produced by the behavioral sciences. Once we understand what we know on the basis of behavioral research, then we can juxtapose this understanding with the imaginary of behavioral governing which imagines knowledge that power could frictionlessly act on. In my previous work I have initiated such an analysis by following analytical strategies of feminist philosophers of science (Małecka 2021). We learn from feminist philosophers of science that to understand the knowledge produced in science, it is not enough to look at end results of experimental work or theoretical claims made by scientists. We have to understand how evidence is produced; in light of which background assumptions data is treated as evidence for a hypothesis; and how concepts are defined, operationalized, and measured (e.g., Longino 1990, 2013; Anderson 2004; Haraway 1981). We still need such a detailed engagement with scientific research produced by behavioral scientists.
Let me come back to the Flint water crisis and the attempts to use behavioral science to address aspects of it. The Social and Behavioral Sciences Team under the Obama administration was aware of the social and political complexity of this public health crisis and “its racial history, its socioeconomic circumstances, all of it,” as the director of the team stated (Stillman 2017). Yet it is difficult not to feel dissatisfaction with “behavioral tweaks” that the team wanted to introduce. They seemed banal, neutralizing the politics of the crisis. However, pointing out that depoliticization is achieved via reliance on behavioral science often also presumes that this science provides a value-free account of “mechanisms of human cognition underlying behavior.” As I have argued, such a presumption is also widely shared in the debates about behavioral policy and is a result of operating within the imaginary of behavioral governing by most scholars. Research in cognitive psychology and behavioral economics does not offer robust knowledge about cognitive causes of human behavior. Furthermore, this research is not free, as no science is, from value-laden preconceptions entering it at diverse stages of scientific investigation. Behavioral science is embedded in preconceptions about social reality, but this point remains unscrutinized in the debates about its use in policy (Małecka 2021). These preconceptions may not allow for the generalization of the results of behavioral research to all humankind (if this research concerns human agents at all), as moral, social, or political values permeating background assumptions of behavioral science—behavioral economics and cognitive psychology—may make it difficult to apply its results to contexts that differ from conditions assumed in the background of behavioral research. The feeling of uneasiness about behavioral interventions proposed in the context of the Flint crisis may be related to the fact that the attempts to employ them in this particular context revealed and made visible, at least partly, values embedded in behavioral research. Analysis of the knowledge produced within the behavioral sciences allows us to challenge claims about the ways in which policies informed by behavioral research impact individuals, which are a manifestation of the imaginary of behavioral governing. Such analysis can also be helpful in detecting the politics of this research, which, it seems, may carry on since the Cold War era in the United States.
I would like to thank Alondra Nelson and Charis Thompson for curating the excellent seminar Science and the State at the Institute for Advanced Study in Princeton during the academic year 2020–21, and the colleagues who participated in the seminar for their inspiring conversations. I am grateful to Joy Rohde and Joshua Barkan for insightful and generous feedback on earlier versions of this essay and to Sonja van Wichelen for her patience and understanding during the process of editing the special issue. The work on this essay has been supported by the funds of the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement no. 754513 and the Aarhus University Research Foundation, as well as by the funds from the Sloan Foundation granted to the School of Social Science at the Institute for Advanced Study in Princeton, NJ.
Notes
Evidence-based policy is an approach to policymaking that advocates designing policy solutions on the basis of the best available scientific evidence. Its proponents suggest that scientific research should be consulted to diagnose a problem that policymakers address and to test policy proposals in experimental settings, during randomized controlled trials. The idea of experimental tests of policies is closely related to the view of policymaking as finding efficient and effective interventions (so-called what works policymaking), which was promoted by the new public management approach (Head 2013). Adherents of such an approach believe that it allows us to limit or eliminate “ideologically-based decision making” (Banks 2009; see also Cartwright and Hardie 2012).
One of the important differences between liberalism and neoliberalism, according to Foucault, lies in the understanding of the subject. In classical liberalism homo economicus is understood as an agent whose activity should remain untouchable by government, whereas in neoliberalism a subject is redefined as someone who is responsive to modifications in the environment through which power is exercised; this “is someone who is eminently governable” (Foucault 2008: 270).
The governmentality lecture was a portion of the lecture series delivered by Foucault at the Collège de France in 1977–78 and later published under the title “Security, Territory, Population.”
Foucault looked at how the relationship between governing and knowledge was thought of, imagined, and postulated to work, but not how it actually worked. For instance, his interpretation of liberalism and neoliberalism is based on his reading of treatises and works of political and economic thinkers (John Stuart Mill, Friedrich Hayek, Gary Becker, Walter Eucken). However, the way in which Foucauldian scholars analyze behavioral public policy and nudging (Leggett 2014; McMahon 2015; Jones, Pykett, and Whitehead 2013) does not suggest that they interpret Foucault's work as identifying an imaginary of how behavioral science influences behavior via policy. Like most other scholars participating in the debate on nudges, they seem to presume that behavioral policy influences (governs) behavior of individuals on the basis of insights from behavioral science.
The imaginary of behavioral governing has also played an important role in producing behavioral science. The idea of producing scientific research to control, influence, or change behavior facilitated development of the behavioral sciences. This ranges from the behavioral social science funded by the Ford Foundation in the 1950s, the aim of which was to transform the social sciences into policy-relevant research based on general “laws of behavior” (Crowther-Heyck 2006), through behaviorism in US psychology, particularly in B. F. Skinner's research, that came up with technologies of behavior aimed at changing people's habits or routines (Rutherford 2009), to cognitive psychology that had affinities with the military's research communication, command, control, information approach (Edwards 1996). However, it does not mean that this research indeed “delivered” the knowledge that its patrons, donors, and proponents hoped, and imagined, to gather.
