Behavioural Science Meets Data Science

There has been much recent scrutiny of the real-time, real world experiments that Facebook has been conducting with its users. Randomised controlled studies conducted by Facebook have identified measurable changes in the “emotional content” of users’ posts according to the amount of negative and positive messages they were strategically exposed to (for more on Facebook’s emotional contagion study click here and here). Meantime, Facebook’s “Voter Megaphone” campaign, which promotes voting by revealing the names and faces of friends who have already cast votes (and was portrted to have increased voter turnout by some 340,000 in the 2010 US elections) has generated further controversy. While the promotion of voting appears to be a good use of social media, the fact that the Voter Megaphone project was also part of a study, and was thus only applied to certain users, raised ethical questions about the real world political impacts of this behavioural manipulation. I have reflected elsewhere about the various issues that the use of RCTs in behavioural trials raise. While Facebook’s use of RCTs clearly increases the stakes of this debate (with a voting nudge being able to instantaneously reach 160 million people in the US alone), in this post I want to reflect more broadly on the emerging connections that are evident between behavioural science and data science.

In an excellent piece for Forbes, Parmy Olson recently considered the broader implication of big data and smart technologies for behaviour change. According to Olson, “Now the proliferation of connected devices–smartphones, wearables, thermostats, autos–combined with powerful and integrated software spells a golden age of behavioral science. Data will no longer reflect who we are–it will help determine it.” (Olson, 2015). It was that phrase, “Data will no longer reflect who we are–it will help determine it” that really got me thinking. The idea that integrated technologies will increasingly enable us to close the loop between learning about what people are doing and being able to shape what they are doing, at unprecedented demographic scales, surely heralds a qualitative and quantitive sea change in behavioural governance.

It is important to note that the exploitation of the connection between technology, data and behaviour change is nothing new.  Olson traces the history of this behaviour changing industry to Sun Micro-systems in the 1990s who recognised the potential of testing different forms of early internet browsers among the same group of users, in order to refine and improve its products. Google has, of course, taken this form of user experimentation to new levels, as it constantly tests subtle modifications of its platforms on millions of users. But something appears to be shifting in the contemporary use of internet technologies. Whereas Sun Micro-Systems and Google were initially interested in testing and refining their on products, and associated consumer interactions with those products, Facebook and others are now using these technologies to change more varied forms of real world behaviours that are practised well beyond the consumer product relation. Social media outlets  are also able to exploit the behavioural power of social networks and herds in ways that early software designers could not have imagined.

There is now an established field of academic inquiry that explores the role of Computers as Persuasive Technologies (named Captology). What is clear is that the commercial development, adaptation and application of these new opportunities is currently moving much faster than academic and political scrutiny of their impacts. A range of new start-ups are fusing the insights of internet optimising experiments with those of the behavioural sciences. It is now possible for energy companies, supermarkets, fitness firms, and financial groups to continuously test their latest use of behavioural insights on their clients consumption patterns and behavioural habits. As household technologies (such as smart energy meters, smart TVs, and mobile health monitoring wrist bands) become ever more interconnected, the potential for behavioural manipulation and experimentation becomes ever greater. The vision of the smart city, with its ability to monitor public behaviours and habits at larger scale, could change the geographical scope of these developments further.

To observe such developments is one things, but to explain why they require critical scrutiny is another.   For me the need to subject the emerging nexus between big data and behavioural science to critical analysis is a cut and dried argument. The application of the behavioural sciences within public policy making (something which we discuss regularly on this blog) raises important constitutional and ethical questions (concerning who has the right to wield psychology power, to what ends, and with what degree of public accountability and disclosure). It is one thing, however, when these behavioural insights are being applied in fairly generic ways by governments to encourage us to pay our taxes on time, register for organ donations, or save for our retirement; it is quite another matter when new forms of behavioural power can reach us every second of everyday, while being reflexively imbued with the coded knowledge of past conducts, habits and behavioural proclivities. It is also quite another thing when we can be subject to behavioural trials and experiments, the knowledge of which can then be used to change our behaviours further down the line, without our knowledge or consent.

Various frameworks have been suggested to develop a more critical account of big behavioural data science. Parmy Olson has, for example, suggested the idea of the Guinea Pig Economy to capture the ways which we are incessantly having our behavioural patterns tested and monitored. I have suggested the notion of the experimental citizen as a way of drawing attention to the impact of unseen experiments on our rights as political subjects. Increasingly, I think that it will become important to draw on the insights of critical surveillance studies and emerging analyses of the political economies of big data and contagion. These perspectives should lead to the notions of accountability, consent, ownership and access entering the lexicon of the guinea pig economy. As knowledge of our behaviour, and how it can be reflexively used to shape our future behaviours (and deployed to transform the behaviours of those around us), becomes the object of proprietorial struggles, these are discussions we can ill afford to delay.