The Process of Self Discovery Has Now Been Technologically Outsourced

In 2015 a team of researchers at Cambridge and Stanford reported that a computer program, given the Facebook likes of a person, could judge his personality more accurately than his friends and family could. With ten likes the program beat a work colleague. With seventy it beat a friend, with a hundred and fifty a parent, and with three hundred it beat a spouse. The finding unsettled many readers, because it suggested that the oldest of human projects, the effort to know oneself, was being done better by software than by the people closest to us. The topic puts the point as a claim, that self discovery has now been technologically outsourced.

The injunction to know oneself was carved, according to Pausanias, at the temple of Apollo at Delphi, and Socrates built his philosophy on it. In India the Upanishads ask who is the knower and the sage Ramana Maharshi taught the question Who am I? as a method. In all these traditions self knowledge was an arduous task of reflection, conversation and discipline. It was also personal. No one else could do it, though a teacher might help.

Today a large part of the work is handed to devices. A watch records how we sleep, how our heart behaves and how many steps we take. The Quantified Self movement, named in 2007 by the writers Gary Wolf and Kevin Kelly, encouraged people to track their moods, diets and habits to learn about themselves. Streaming services close the year by telling listeners what kind of person they are, according to what they played. DNA tests promise to reveal ancestry and disease risk, and online questionnaires sort users into personality types in minutes. We consult these outputs as we once consulted an astrologer.

Some of this is useful. A person who thinks he sleeps well may learn from a record that he does not. A bank statement shows what a person values more honestly than his account of himself, and a health app may expose a pattern that no amount of introspection would. Psychologists know that introspection is unreliable, since people often misreport their reasons and are blind to their own biases. A machine that observes behaviour without flattery can correct the story that we tell about ourselves. In this limited respect, outsourcing is a gain.

A machine that observes behaviour without flattery can correct the story that we tell about ourselves.

Yet the mirror is not neutral. The data that devices collect are the ones that can be recorded. A watch counts steps but not the conversation that made the walk worth taking. An algorithm reads clicks, and clicks reflect what caught the eye at a weak moment and not necessarily what a person cares about. The profile it builds is of the consumer, designed to predict what he will buy or watch. In 2018 it emerged that the consulting firm Cambridge Analytica had used data on tens of millions of Facebook users to target political messages by personality, which showed what the commercial interest in knowing us can do.

The mirror also shapes what it reflects. Eli Pariser pointed out in 2011 that personalised feeds create a filter bubble, in which a person sees more of what he has already liked. Over time the self that the system predicts becomes the self that it produces, as recommendations narrow taste and reinforce habit. A person who is told that he is an introvert, an analyst or a certain type may begin to act accordingly. Labels such as those of the Myers Briggs test, whose scientific standing is poor, can harden into identity. The discovery turns into a prescription.

There is a deeper loss. Self knowledge in the older traditions was not a set of facts about oneself but a transformation that occurs through effort. A man comes to understand his courage by having to be brave, and his capacity for patience by being tried. These are learnt through experience and struggle, and cannot be delivered as a report. Sherry Turkle warned in Alone Together that people are turning to machines for companionship and self reflection precisely because the machines make no demands, and that something essential to being a person withers when we avoid the friction of other humans and of ourselves.

India offers its own illustration. Matrimonial sites ask users for their interests, income and horoscope, and algorithms propose the partners who match, so that a decision in which families once debated character at length is guided by a score. Students choose careers by taking aptitude tests online, and recruiters screen candidates with software that reads their résumés and sometimes their voices and faces. In each case an external system pronounces on who a person is and what suits him, and the person tends to accept the verdict because it carries the authority of data. Self knowledge also has a moral dimension that no profile can provide, since a person learns who he is by what he does when it costs him.

Education and culture should respond accordingly. Young people who grow up with devices that tell them who they are need practices that teach them to find out for themselves, such as keeping a diary, taking solitary walks, reading widely and talking to a mentor. A school might teach students to treat an algorithmic profile as a hypothesis to test and not as a verdict to accept. The same caution applies to institutions that use such profiles in hiring, lending and policing, where the person described may have no chance to answer back.

It would be too simple to say that technology has taken over the task. It has taken over the measurement, while the interpretation remains ours, and each person must decide what the numbers mean and what to do about them. But the boundary is slippery. When a person consults his phone to learn how he feels, as many do, the question has already shifted. There is a danger that those who rely on the mirror will lose the ability to see without it, and no outside observer, human or machine, will be positioned to tell them what has gone missing.