This blog was written thanks to support from the Independent Social Research Foundation’s Mid Career Fellowship programme 2019
‘Just like unlocking the human genome helped identify genetic traits that allow for personalized medical advice, we can think of machine learning as the next step in unlocking a “behavior genome.” By factoring in personality traits, situational features, and timing, we can better persuade people who want to be persuaded’ (Risdon, 2017).
Introducing data, datafication and nudges
For some time now, I have been interested in the actual and potential connections that exist between nudges and smart tech. By nudges I am referring to techniques, inspired by the behavioural sciences, that use subtle prompts to encourage desired behaviours (for example encouraging the payment of those troublesome tax bills by letting people know that others normally pay their tax in full). By smart tech I am referring to that bewildering array of digital infrastructure, algorithms, social media platforms, and wearable technology that is able to tell us more and more about ourselves (and the activates of others) in salient ways, and in real time. There are already myriad examples of how nudges can use social mediavectors to encouraged behavioural goals, and that wearable technologycan capture data that can inform nudge techniques and be a vector for nudging itself.
As I have started to look more carefully at these emerging connections, I have been drawn to a deeper set of alignments that undergird the nudge/smart tech amalgam. These deeper associations relate to the processes of datafication, or more specifically, what van Dijck terms dataism (2014). Dataism is an ideology that has emerged in the wake of the capacity to gather increasingly vast quantities of data on human behaviour and social life. Datafication reflects the technical and sociological processes that have enabled dataism: namely the proliferation of new technological capacities to sense, monitor and record everyday life; and the sociological propensity to plug-ourselves in to such data production systems, and to open up our social life to related dataveillance. Dataism is about more than technical processes though: it propounds a subtle shift in our ontological and epistemological universe (see Beer, 2018). This shift is characterised by a suggestion that data (in all of its inevitably reductionist forms) is ontological: that is to say it not only gives us indicators as to what is going on in the world, but it is a core feature of the nature of reality in itself. The epistemological component of dataism emerges from the analytics industry, which suggests that when properly processed data can not only reveal the shape of reality, it can explain it! What I provisionally consider here is the difference that thinking about connections between nudges and smart tech as an amalgam of behavioural science and dataism can make to critical analyses of the field. I think what is fundamentally at stake is the difference between thinking of smart tech as being an upscaling vehicle for the delivery of nudges and understanding the fusion of dataism and behavioural science as a more fundamental shift in collective understandings of political, economic and social life.
Emerging synergies: From the Age of Data Utility to Augmentation
There is already a body of industry-based reflections on the connections between nudge and dataism. In a fascinating piece in the Behavioural Scientist Chris Risdon (2017) (head of behavioral design at Capital One) succinctly explains the potential and significance of fusions of the behavioural and data sciences. According to Risdon, in an era of algorithmic machine learning, smart tech facilitates enhanced scalingand matchingof behavioural interventions. In other word, smart tech can facilitate the wide-spread up-scaling of nudges (which could now reach large cohorts very quickly). At one and the same time the gathering of big sets of data concerning human behaviour, combined with algorithmic machine learning, means that nudges can increasingly be matched to those who are most in need, or indeed most susceptible, to them. In a 2015 Deloitte Review Report Guszcza outlined a broader set of dialectics that could emerge as behavioural and data science combine:
- The predictive analytics of big data will offer ways of delivering nudges to the most relevant parties at the most salient times. According to Guszcza, this will mean that nudges will no longer have to be delivered in the form of a one-size-fits-all approach at population levels(avoiding the unintended reactance, spill-over effects, and inefficiencies this would involve).
- The use of behavioural insights within data science could facilitate a shift from predictive analytics (what a person is likely to do in the future) to the shaping of behaviour (what Yueng (2016) describes as ‘big data driven decision guidance techniques’). Risdon (2017) has suggested this shift reflects an epochal move in data analytics from the Age of Utility(where data helps to make our lives easier/smarter), to the Age of Augmentation (when data can actually be used to more effectively shape our conducts).
- In addition to behavioural science enabling data science to shape human conduct, data itself could help to produce products of behaviour change: in the form of highly personalised forms of data feedback.
Ultimately the multiple potential fusions of data and behavioural sciences appear to promise a shift from staticnudges (the kinds that are built into the design of buildings, forms, or defaults), to more dynamicsystems (which are highly personalised and liable to machine learning adaption) (see Yeung, 2016).
Deepening the relationship: Performance Enhanced Nudging
There are other interesting features of the emerging relationships between nudge and data science that: 1. suggest that their fusion will extend and deepen in the coming years; and 2. indicate terrains of future critical analysis. The potentially extended and deepened interaction between nudges and data science is based on a series of natural synergies that they appear to have. First, both nudge and the data science movement are in part inspired by an understanding of the human condition in relation to the qualified-self. The qualified self of behavioural science is based on a perception of the human that recognises the cognitive restrictions and limited forms of willpower exhibited by people. Within data science the notion of the qualified-self recognises the limits of knowledge that are associated with embodied senses: limits which can, of course, be overcome within the dispassionate monitoring of biological and social life promised by the quantified-self (Davidson, 2015). Second, there remains significant, if still relatively under-develop, synergies between the behavioural insights on which nudges are based and the technological potentials of data science. For example, the behavioural sciences behind nudge techniques recognise that a key to stimulating desired behaviour is salience(namely that a behavioural prompt is relevant to the target audience/individual). Dataism promises new horizons of salience science, as data is used to provide increasingly personalised prompts and feedback that can be targeted at the most influential times. Furthermore, one of the most powerful behavioural insights associated with nudges is the recognition it gives to the power of social influences (and, in particular, the power of peer-to-peer pressure and herd instincts). More static nudges have, in the past, attempted to change behaviours by informing households how their patterns of energy consumption compare with neighbourhood averages. In an age of algorithmically fine-tuned social media channels, social influence can now be utilised at ever greater scales and with increasing levels of salience. Rather than knowing how your behaviours compare with anonymous local residents, you can see how your conduct relates to named friends and peer networks. A final, if often neglected synergy, between the behavioural and data sciences relates to the process of de-datafication (Dow Schüll, 2016). One of the limits associated with datafication, and the related Quantified Self Movement, is that while it can produce data of increasing levels of intensity, data is not always the best way of promoting behavioural change. It is in this context that future collaborations between nudge technicians and the big data industry will increasingly see behavioural science exploring ways in which data streams can be de-datafied in order to make their meaning more socially relevant. This could pave the way for the proliferation of what Dow Schüll (2016) has described as highly personalised micro-nudges.
Mapping a critical terrain.
If all of this suggests that the behavioural and data sciences are likely to become functionally integrated in the future, this integration suggests significant terrains of future critical analysis. The more cyber utopian interpretations of nudges and dataism have as yet not been matched by a fully formed critical response. There is, of course, much critical work on questions of data-surveillance, privacy, and data mining (see van Dijck,2014; Beer, 2018; Zuboff, 2015). But this work has tended to focus on the ways in which dataism can predict future behaviours (particularly in relation to patterns of consumption), as opposed to shape behaviour. While the potential of behaviour modification is often intimated in analyses of the predictive potential of dataism, there remains only limited systematic critical reflection on this opportunity space. One exception has been the pioneering work of Yeung on hypernudging(2016). According to Yeung,“Big Data driven nudging is […] nimble, unobtrusive and highly potent, providing the data subject with a highly personalised choice environment – hence I refer to these techniques as ‘hypernudge’. For Yeung the hypernudge is the idealised meeting place of data and behavioural science: the point at which these sciences find an optimal collaborative form of expression. It is, for want of a better term, nudging on steroids. But if the hypernudge reflects an unprecedented escalation of the reach and impact of both the behavioural and data sciences, it is clear there is much to critically scrutinise in this emerging action space. Nudging has been critiqued as a form of opaque manipulation of human action, and a largely uncontestably intrusion by the psychological state into people’s everyday lives (Whitehead et al 2017). The hypernudge tends to up the critical stakes. The fairly static exploitation of the collective unconscious within nudges is combined with the intellectually opaque territory of algorithmic government (Zuboff, 2015). Furthermore, the governmental orchestration of behaviour change begins to move from the public sphere into the proprietorial realms of data brokers and platforms associated with the corporate world. According to Yeung these developments are exposing “[t]he inability of the liberal political tradition to grasp how commercial applications of Big Data driven hypernudging implicate deeper societal, democratic and ethical concerns” (Yeung, 2016: 17).
For some time now nudge techniques have provided policy-makers and behaviour change entrepreneurs with a basis for challenging liberal norms of freedom. In particular nudges have provided the grounds upon which the harm-to-othersprinciple—that has been the historical yardstick against which legitimate government intervention into citizens’ everyday lives has been determined—could be challenged. If personal freedom can be preserved (largely through the maintenance of some form of choice), then advocates of nudge suggest that there is renewed scope for governmental intervention in harm-to-self issues (such as personal health and financial matters). Within these nudge assumptions is an often overlooked, but not insignificant, challenge to liberal norms regarding the basis of personal freedom. But static nudges can still feel well-intentioned and fairly harmless. While hidden within the choice architectures of everyday lives, they were still subject to the usual checks and balances of liberal government and appear reasonably easy to resist. But the fusion of nudges and dataism clearly change the nature of the balance between behaviourism and freedom. Hypernudges are more potent, opaque, and persistent than their static counterparts. They have the ability to be constantly revised and reapplied, while hiding behind the proprietorial armature of emerging forms of surveillance capitalism (see Zuboff, 2016). Hypernudging also has the ability to reach new scales of influence that traditional nudges could only dream of.
While there are now critical frameworks of analysis that have begun to make sense of the extractive economic logics of dataism (see for example Zuboff’s (2015) theory of surveillance capitalism), critical political analyses of the fusion of the behavioural and data sciences are limited. Building on Yeung’s pioneering work on hypernudges, I think that there are several lines of inquiry that a critical political analysis of the dialectics between nudges and dataism could take:
- At the most obvious level, there is the direct impact that nudges and dataism are having on democratic elections (from Facebook’s Voter Megaphone Project, to the operations of Cambridge Analytica).
- There are important questions of algorithmic accountabilitythat the hypernudge raises. If the fusion of behavioural and data science is likely to involve new public private partnerships, how are the opaque operations of proprietorial algorithms going to be held to democratic account.
- The operations of hypernudges raise important questions concerning implicit exchanges and transactions in personal freedom. The Big Data Industry’s economic model is premised on the exchange of personal privacy for free data services. But in the world of hyper- and micro-nudges these exchanges of personal freedom become more complicated. We are not only trading privacy, but autonomy, as our data can be fed back to us through choice architectures and prompts that seek to actively shape our decision-making and conduct. It is important to ask critical questions about the conditions under which these exchanges of autonomy for data and other services (perhaps reduced health insurance premiums if we allow ourselves to be quantified and nudged) are made, and the extent to which they corrode actually existing liberal freedom.
- As dataism provides greater certainty about our actions, and nudges are able to shape our conduct more directly, it is also important to consider the combined impacts of these process on trust and freedom within the contractual interactions of everyday life. According to Zuboff (2015), trust between different social actors is a critical condition for the effective functioning of society and the preservation of freedom. Without trust totalitarianism may become inevitable. Trust is, of course, a condition of freedom, and the ability to reasonably err from a social agreements made in trust is a vital condition for human autonomy and flourishing. But what if the certainty of data science and the power of hypernudges does not require trust, just compliance (achieved through full surveillance and behavioural manipulation)? What then for human freedom?
This list of research areas is, of course, by no means exhaustive. It is, however, indictive of the significant scope and ethical importance of research in this area. It is also important to recognise that the fusion of data and behavioural sciences does not inevitably lead to a form of big data cyber-dystopia. The work of Acquisti, Brandimarte, and Loewenstein (2015), has pursued a more progressive dialogue between the behavioural sciences (particularly behavioural economics) and data science. Their work has sought to apply the insights of the behavioural sciences to better understand peoples’ susceptibility to a loss of online privacy. It is, perhaps, ironic that behavioural science can simultaneously provide the basis for undermining freedom in an age of smart tech, as well as a framework for understanding human frailties to online privacy exploitation that can be used to guard against the dangers of dataism.
It is to be hoped that the coming years will see both the growth of critical studies of the interactions between behavioural and data sciences, and the more progressive fusion of these two powerful epistemological projects.
Acquisti, A. Brandimarte, L and Loewenstein, G. (2015) ‘Privacy and human behaviour in the Information Age’ Science: Vol. 347 no. 6221 pp. 509‐514
Beer, D. (2019) The Data Gaze (London, Sage).
Davidson, J. (2015) ‘Plenary address e A year of living ‘dangerously’: Reflections on risk, trust, trauma and change’ Emotion, Space and Society 18: 28-34.
Dow Schüll, N. (2016) ‘Data for life: wearable technology and the design of selfcare’ Biosocieties 1-17
Guszcza, J. (2015) ‘The last mile problem: How data science and behavioural science can work together ‘ Deloitte Review Issue 16: 65-79.
Risdon C. (2017) ‘Scaling Nudges with Machine Learning’ Behavioural Scientist October.
van Dijck, J. (2014) ‘Datafication, dataism and dataveillance: Big Data between scientific paradigm and ideology’ Surveillance and Society 12: 197-208.
Whitehead, M et al. (2017) Neuroliberalism: Behavioural Government in the 21stCentury (Abingdon, Routedge).
Yeung, K. (2016) ‘’Hypernudge’: Big Data as a Mode of Regulation by Design’ TLI Think! Paper 28/2016.
Zuboff, S. (2015) ‘Big other: surveillance capitalism and the prospects of an information civilization’ Journal of Information Technology 30: 75-89.