Premature Selfhood
Who gets there first: the system or the self?
In The Future of You, published in 2021, I wrote about a possibility that disturbed me. What if systems became good enough at reading our behaviour to tell us not only who we were, but who we were going to be, perhaps before we knew it ourselves?
I was writing then about facial analysis, behavioural prediction and the growing capacity of technology to infer things about us from signals we barely noticed we were producing. Yuval Noah Harari had posed a particularly uncomfortable version of the problem. Imagine an algorithm noticing, from your glances and your attention, something fundamental about your identity before you had recognised it yourself. Should it tell you? Should it gradually steer you towards its conclusion?
At the time I understood this mainly as a question about privacy, manipulation and the outsourcing of identity. I now think there is another problem underneath it, and it is a problem of timing.
In the second essay of this book I described how the system's model of my future gets there before I do. Here I want to follow that thought inward, from the future of the world to the future of the self. The self before the self.
I call this Premature Selfhood: the condition in which a system’s model of who we might become begins to shape us before we have had the chance to become for ourselves.
Selfhood takes time. We become who we are through experience, relationships, mistakes, reversals, curiosity, memory, imagination and encounters with things we did not know we wanted. We contradict ourselves. We change our minds. We surprise other people, and sometimes ourselves. There is no particular reason why the person I appear likely to become at fourteen, thirty-four or sixty-four should be the person I eventually become.
That openness matters. But prediction introduces another actor into the process. A system does not need to wait for me to become. It can infer. It can take what I have already done, combine it with patterns found in millions of other people, and construct a probable version of what I might do next.
That future self might be commercially valuable, administratively useful or statistically accurate. But it is still a model.
Premature Selfhood begins when that model stops merely describing a possibility and starts taking part in my becoming. The earlier essays showed how predicted futures can shape a person's access to credit, the scrutiny their neighbourhood attracts and the environment in which their preferences develop. Premature Selfhood is what happens when the future being modelled is not the economy or the city, but me.
From prediction to becoming
Prediction is not the problem in itself. Human beings predict one another constantly. Parents imagine what their children might become, teachers recognise aptitude, and friends tell us they always knew we would end up doing whatever we eventually do. Sometimes they are right.
But human predictions are usually understood as interpretations. They can be argued with, resisted or disproved.
Computational prediction can acquire a different kind of authority. It may be quantified and repeated at scale. It may be embedded in a system, connected directly to a decision and hidden from the person being assessed. And it can become personalised enough to feel intimate.
That matters because a prediction about the self can become part of the environment in which the self develops. The recommendation changes what I encounter. The score changes what I can access. The personalised assistant changes which options are placed in front of me.
The predicted self begins to shape the lived self.
This is where Premature Selfhood meets the four conditions of futurity. An outsourced future is imagined elsewhere. An invisible future may operate without my knowing it. An imposed future starts changing what is possible for me now. Premature Selfhood is what can happen to the self inside those conditions. The future being imposed is not only a future about the economy, the climate or society. It may be a future about me.
Before I have become
The word premature matters. It does not simply mean early. It means before the proper time. A premature conclusion is drawn before the evidence is in. A premature decision closes something that might usefully have stayed open.
Premature Selfhood is therefore not just an early prediction of identity. It is the premature closure of becoming.
The model says: this is the sort of person you are. Or: this is the sort of person you are becoming. And then systems begin organising themselves around that proposition.
The danger is not necessarily that the prediction is wrong. It may be remarkably accurate. The harder question is what happens when accuracy itself becomes an argument for authority.
If a system predicts my preferences better than I can articulate them, should it decide what I see? If it detects a characteristic before I recognise it myself, does that knowledge belong to the system before it belongs to me? And if an AI knows enough about my past to simulate a plausible future version of me, when does assistance become authorship?
These are questions about agency, because becoming has always contained a degree of opacity. I do not fully know who I will become. Neither do you. That uncertainty is not simply an information gap waiting for better technology to close it. It may be part of what makes agency possible.
AI changes the scale
Generative AI sharpens all of this, because the system no longer merely calculates a probability. It can talk, remember and explain. It can synthesise years of information into an account of who I appear to be, recommend what I should do and, more and more, act on my behalf. It can generate plausible versions of what comes next.
This creates an unusual asymmetry. The machine can draw on a record of my past behaviour larger than anything I can consciously remember. It can compare me with patterns across populations I will never encounter. And it operates continuously, without forgetting, fatigue or embarrassment.
That does not make its model me. But the better the model becomes, the easier it is to confuse predictive accuracy with identity. A persuasive enough model of the self can begin to feel less like a proposition and more like a revelation.
Perhaps this is why the old language of fate becomes unexpectedly relevant. People have always sought to know themselves through accounts of the future. Astrology, prophecy and tarot all offer versions of what lies ahead, and part of their power is that once a future has been named, we start noticing ourselves in relation to it.
AI gives that ancient impulse a new authority. Not the mystique of the oracle, but the authority of data. The prediction appears to come not from intuition or divination but from computation.
The right to remain unfinished
There is an obvious response to all of this: predictions are useful. Of course they are. Predictive systems can detect disease earlier, anticipate risk and help people make better decisions. The answer cannot be to abandon prediction. The question is what relationship we want with it.
In The Future of You, I contrasted systems that make citizens ever more transparent to institutions with Audrey Tang's very different philosophy: making the state transparent to the citizen. That principle belongs here too. If systems are going to construct futures about us, those futures need to stay visible and contestable to the people expected to live inside them.
What has been inferred, from what, and with what confidence? What follows from it? Can I reject it? Can I change? Can I surprise the system?
In other words, we need to protect something predictive technology has little reason, by itself, to value:
the right to remain unfinished.
Keeping Becoming open
This is why Premature Selfhood closes this book. Too Much Future asked what happens when we are saturated by futures generated elsewhere. Four Conditions of Futurity asked what happens to agency when those futures become outsourced, invisible or imposed. Half an Hour of Future was an attempt to reclaim a small space in which we can take part in imagining what comes next.
Premature Selfhood asks what is at stake if we do not.
What suffers when systems become so good at anticipating us that they write our futures before we have had the chance to live them? The answer may be becoming itself.
The danger is not that machines will eventually know us perfectly. They will not. The more immediate danger is that their imperfect versions of us are acted on before we have had the chance to contradict them, and carry enough authority to shape who we become.
A future self is supposed to be somewhere I have not reached yet.
We should be careful about letting the system get there first.