Limitations and ethical considerations:
Since the US-presidency elections of 2016 and the surprising victory of Donald Trump, a sudden public interest in “big data” was sparked, since it was suspected to be key to Trump’s success. More precisely, it was the combination of “big data” and behavioral nudge approaches that was key to the strategy of Donald Trump. This essay examines the powerful synergies of “big data” with behavioral economics on the basis of real life examples from politics and business. Moreover the practical limitations as well as ethical considerations will be discussed.
The election of US-president Donald Trump on November the 9th once again sparked a public debate on big data. However, the Donald Trump-campaign was not the first campaign that was largely supported by big data approaches, since it was actually the 2012 Barack Obama-campaign that first used such an approach. A company called Cambridge Analytica quickly strut itself in, with the victory of Donald Trump. Company officials stated that their approach that combines behavioral insights, big data assessment and ad-targeting was pivotal in bringing about the victory of President Trump. In the German-speaking areas the article “Ich habe nur gezeigt dass es die Bombe gibt” (dasmagazin.ch, 2016) caused sensation on this topic, having been shared over 154,695 times on Facebook alone. On the one hand the article sparked interest on the numerous useful applications of big data, whereas on the other hand the article raised concerns on the ethical use of such approaches. Essentially the concerns are rooted in the allegation of using big data to manipulate people. Moreover a more topical concern revolves around the limitations of big data and the insights that can be drawn from it – respectively the interpretations of these insights Behavioral economics came about through the introduction of psychological insight into the field of mainstream neoclassical economics. Psychologists like Kahnemann and Tversky discovered through various psychological experiments, that the neoclassical assumption of humans as rational decision makers is incomplete. In reality humans neither have access to perfect information, nor are they capable to process all information that they do have. The aim of behavioral economics was not the refutation of mainstream economics but to increase the explanatory power of economics by providing it with more realistic psychological foundations (Camerer & Loewenstein, 2004).
Central to behavioral economics is the belief that increasing the realism of the psychological foundations of economic analysis will advance the field of economics as a whole. Proponents are convinced that the introduction of behavioral insights into mainstream economics leads to better theoretical insights, allowing for more accurate predictions and better policy (Camerer & Loewenstein, 2004). Nowadays many of the insights that were produced by behavioral economists are used by companies and public organizations alike. Marketers today use behavioral insights for commercial purposes, arguably manipulating consumers into buying things they don’t really need. Public organizations and policy makers use behavioral insights to nudge people in making better decisions for themselves, thereby increasing the public good.
This essay aims to explore the synergies of behavioral economics (especially nudging approaches) and the emergence of big data through the rapid digital transformation of Western societies, as well as the everyday lives of people. Furthermore this essay aims to address the limitations and ethical concerns that this development brings about.
The rapid digitalization of society since the emergence of the internet has made it possible to gather large amounts of data. “Big Data” is obviously a buzzword that is used to describe the increased amounts of complex and unstructured data and the technologies that produce and gather these type of data. Much of what is referred to as “big data” is in factbehavioraldata. These are the digital traces that human beings leave behind as they go about their daily activities using computers, smartphones and other digital devices. Some proponents believe that “big data” heralds the “end of theory”, as the sober collection of an exponentially increasing amount of data on human behavior, through an increased amount of digital touchpoints, makes models on human behavior increasingly obsolete. Google’s R&D department extends the famous saying by George Box, that “all models are wrong, but some are useful” (Wired, 2008) by saying that with increased amounts of data, success can be achieved without relying on models at all. This idea is founded in the belief that not even the best models can describe human behavior realistically, but if human behavior is observed and measured close to perfection and analyzed with sophisticated computing power, then the result will be much more accurate than any model could describe. In their aim to find ways to describe and predict human behavior more realistically, the proponents of “big data” resemble the behavioral economists. From my own experience as a former digital marketing professional, “big data” and smart algorithms by themselves are rather useless unless human beings use them for better decision making. With regards to behavioral design approaches versus “big data”, the question is not which approach should be favored over the other, but how behavioral design and “big data” can be combined in a manner to achieve leverage. In “big data” applications, the last-mile problem of actuating change of behavior tends to be left to the assessment of the model’s end user. On the contrary, behavioral nudge applications are often cut and dried affairs applied to entire populations rather than analytically identified sub-segments. It is legitimate to expect better results when both approaches are treated as integral and applied in tandem. Behavioral science principles should be part of the data scientist’s toolkit and vice versa (Deloitte, 2015).
As briefly mentioned in the introduction, behavioral design applications that combine predictive analytics with behavioral nudge applications, have been used for the presidential campaigns of both Donald Trump and Barack Obama. For the Trump campaign, Cambridge Analytica composed a strategy that was built upon the psychological OCEAN-Model (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism). The model postulates that every human character trait can be accurately explained by the aforementioned ‘Big Five’-factors (Goldberg, 1993). According to the model a human being can quite accurately described - his or her needs and fears and likely behaviors can be grasped. For the Trump campaigns this model was triangulated with “big data”-analytics and ad-targeting.
The aim of this strategy was to identify voters that were more likely to be persuaded into voting for Trump. It would have been a waste of resources to have targeted individuals that are either committed Trump voters or dedicated Hillary Clinton supporters. Essentially the core of the strategy was to identify those voters most likely tochange their behaviorif visited by a campaign worker. One example of a promising indicator for a prospect Trump-voter was interest in US-produced cars. Moreover, Trump’s campaign workers were equipped with an app that made it possible to uncover the political attitudes and character traits of prospective voters. Consequently, the campaign workers would have only visited people that were susceptible to the messages of Trump. Based on this information, the campaign workers used tailored interview guides for each personality type. Afterwards the responses were put back into the app and used for refining the algorithms. Additionally, digital advertising was personalized to various voter types, optimizing many variables from message, to images, to colors among other factors to yield best possible results.
Somewhat counter-intuitively, the many contradictions with the speeches of Donald Trump resulted in a vast pool of components out of which very versatile messages could be crafted that fit with a wide array of voter identities.
During the Obama 2012 campaign, very similar approaches that combine nudge tactics and big data were used. Predictive behavioral approaches were used to identify undecided prospects with a high likelihood of responsiveness. For example the consistency principle was used by letting prospective voters to fill in commitment cards with a picture of Barack Obama on them. They were also asked to write up a small plan on how they would go about voting for Barack Obama, also dedicating themselves to a specific time for when they would vote. The so-called consistency principle is in line with psychological research that people tend to behave consistently with to their past behaviors and commitments (Cialdini, 2013). Furthermore, campaign workers used the concept of social proof, in which they informed prospective voters that most people in their peer group also committed to vote for Barack Obama.
Both the Trump and the Obama campaigns demonstrate how together, “big data” analytics and behavioral nudge tactics are powerful approaches. Just as behavioral science can help overcome the last-mile issue of “big data”-approaches, perhaps “big data”-approaches can help with the last-mile issue of behavioral economics- in certain contexts, useful nudges can manifest as digital data products (Deloitte, 2015).
The so-called internet-of-things offers many possibilities to combine “big data” approaches with behavioral nudge tactics. Self-tracking-devices like Jawbone-Up make use of peer effects to nudge people into doing more sport, e.g. by communicating to the users that their friends are doing more sport than them. Data on energy consumption of people can be used to nudge people into saving more energy. The effect of social proof could be used to tell the individual person that they consume more energy than their neighbors. The UK insurance company ingenie uses black-box data in order to calculate risk scores, thereby making use of the peer effects to inform drivers via an app, on the riskiness of their driving behavior. Incentive structures that reward safe driving by lowered insurance premiums nudge drivers into driving more safely.
Limitations and ethical considerations:
Thaler & Sunstein (2008) refer to these types of approaches as choice architecture. The idea behind such approaches consists of designing policies and programs that consider the mechanisms of human psychology. These approaches are not believed to restrict choices, since options are arranged and presented in ways that help people make daily choices that are in line with their long-term goals. Contrasting hard incentive structures of classical approaches to behavioral nudges, the latter is a softer approach for prompting desired change of behavior (Thaler & Sunstein, 2008).
Nudge approaches are included in discussions of “big data”. It has already been established that the majority of so-called “big data” is in fact behavioral data. This type of data is controversial for reasons not only limited to basic privacy concerns. Behavioral data that is gathered in one context can be repurposed for use in other contexts where inferences on preferences, psychological traits and attitudes are done with such precision that it can be unsettling for many, invoking fears of an Orwellian society.
- Quote paper
- Alexander Ritter (Author), 2017, The symbiosis of big data and behavioral insights. Applications and ethical considerations, Munich, GRIN Verlag, https://www.grin.com/document/379640