Big Data and Dear Life

I am superior, sir, in many ways, but I would gladly give it up to be human. – Lt. Cmdr. Data

Big data have – or ‘has’ – become big indeed, not just in the sheer amount to petabytes out there, but also in our imagination and consciousness. The concept of big data is closely linked to the concepts of artificial intelligence, deep learning, neural networks, and other quasi-neuroscientific anthropomorphisms. The rise of the machines, if we can imagine it, is foreshadowed in memorable works of science-fiction. The Terminator, who taught that anger is more useful than despair, also gave us reason to despair when he declared that human psychology was but one of his subroutines.

Advocates of the blessings of big data ask us to let data guide our lives. Seth Stephens-Davidowitz, a former data scientist at Google, loves big data with an almost fetishistic fervor, and he eagerly describes its blessings in his new book Don’t Trust Your Gut.

Trust, Stephens-Davidowitz counsels, the big data. There is enough information to choose wisely, reliably, and validly when looking for a mate, a place to live, a career to build, and happiness to enjoy. Whether Stephens-Davidowitz, who describes himself as a data geek and who is hailed as a prophet of the data revolution by a dust-jacket enthusiast, has succeeded to practice what he preaches is not entirely clear.

There is much to be puzzled about with regard to Don’t Trust (see Krueger & Grüning, in press, for a review). but lest this essay become an unnecessarily big one, I will focus on two issues: samples and chance.

Samples

It is a truism in statistics that large samples provide more reliable estimates of population characteristics than do small samples, and this is so even if the population is not composed of random events, that is, if instead the population can be broken up into discrete and differentiated subcategories. We can ask how much a mammal weighs on average, but should we? The answer might be useful when we compare mammals with insects and reassure ourselves that the former are heavier. But even a child can spot a problem with lumping dogs with dolphins. Should we not want to learn the average weight of fairly homogeneous categories?

The dog-dolphin example generalizes to human categories, although things get fuzzier and dependent on the social categories du jour. Stephens-Davidowitz never asks if big data are to be preferred no matter what, that is, to be preferred even when a smaller sample from a narrower subcategory has greater specificity and is more useful for decision-making. As Steven Pinker (2021, p. 167) put it "We have no choice but to use human judgment in trading off specificity [dogs] against reliability [mammals]."

Chapter 7, Makeover: Nerd edition, is revealing. Here, Stephens-Davidowitz asks if data can help to increase personal attractiveness. He could have considered what large data sets say about how men and women (note: not humans in general) are perceived given their features. There are plenty of data available and much of it predates the so-called data revolution. We have long known, for example, that men – but not women – with beards are perceived as more dominant and attractive (Dixson & Vasey, 2012).

Stephens-Davidowitz instead did a self-study, using an app to modify his photographic self- portrait, and using a quick survey to see how he came across. He found that he would benefit from having a beard and glasses. As to the glasses, his example allows that there is a place for idiographic, that is, person-focused, research. One would hope that there are ways for the author to look better while still being himself as opposed to becoming one with a population mean. In short, advice drawn from data sampled from heterogeneous populations ignores a person’s idiosyncratic personal and social world. Big data is better than small data, but only if the sampled population has been carefully identified and justified.

Chance

If you trust big data to run your life, intuition and local advice take a back seat. And so does chance. When you can “google it” because the answer is out there, uncertainty and whimsy melt away. And so it is surprising that Stephens-Davidowitz approvingly notes a study by Steven Levitt (2021) who found that people are happier if they leave big life decisions, such as quitting an unloved job, to the toss of a coin.

Levitt, who rose to fame by combining folksiness with shock value in his book Freakonomics (Levitt & Dubner, 2005), follows in the footsteps of The Dice Man (Rhinehart, 1971; see Krueger, 2010, for a briefing). Let randomness take a hand in guiding your life; relax and feel happier!

Participants in Levitt’s study visited a website where they described a decision they were struggling with (e.g., to quit smoking, to adopt a child). They learned that a virtual coin would be tossed and that they should, if possible, act as instructed by the coin. Participants complied with the coin more for unimportant than for import decisions, but even for the latter there was a small effect. Six months later, those who had followed the coin’s instruction reported being more satisfied than those who had ignored the coin.

Levitt concludes that people are overly cautious when making life decisions, and that one reason why chance can help is that it liberates people from some of the responsibility and anticipated regret they might otherwise experience. Randomness gives permission to do what is necessary.

There is a loose end here, which is worth mentioning. The happiness effect is seen in the comparison between those who were both instructed by chance to make a decision they desired and feared and who then in fact made that decision and those who were not instructed to make the decision and who did not make it. This comparison conflates the verdict of the coin (do it! or don’t do it!) with the eventual choice (I did it! vs I didn’t do it!). Is the action the critical element or is it the action contingent on the permission given by the coin? Possibly but improbably, the increase in happiness occurred only among those participants who made the difficult choice when the coin told them not to.

Trust Essential Reads

What the Flint Water Crisis Revealed About Trust

Algorithm Aversion: Who Trusts Algorithms to Make Decisions?

Referring to Levitt’s work in a footnote, Stephens-Davidowitz calls it a “data-driven life hack” (p. 242). A single toss of a coin or cast of a die is not data but only a datum, and it is not big. Levitt’s gambit, clever and liberating as it is, involves a sort of deliberate ignorance of the kind of big data Google might serve up. Perhaps big data has something to say about whether one should move to Florida, but it can’t tell you whether you should.