When the Future Becomes a Verdict
Written while reading Carissa Véliz's Prophecy (2026)
I paused at two sentences about Carissa Véliz's book Prophecy: "Democracy needs uncertainty to thrive" and philosophy can be an "antidote to prophecy." I understand well the wish to know what is coming. Uncertainty can be heavy to bear, especially when we must make decisions that have consequences for others. And yet there is something unsettling about the thought of a future that is already fixed.
Véliz traces the history of prediction from the oracles of antiquity to artificial intelligence. An oracle did not merely give people a picture of what was to happen. The words shaped what they did. In the story of Oedipus, the characters try to escape the prophecy, but their actions become part of the path toward it. Véliz's concern is not that an algorithm works the same way as an ancient oracle. It is the question of what power we grant to whoever claims to know the future.
A calculated probability can yield useful knowledge. It can help us see patterns we would otherwise have missed. But its meaning changes when the calculation enters an institution that can make decisions about a person's life. It then becomes not merely something we know. It becomes something someone does with what they believe they know.
I recognize this difference from my work with people and assessments, even though the technology at the time was different. Think of a family that comes into contact with child protective services. A concern must be taken seriously. Information must be investigated, and the child must be protected if necessary. Now imagine that a system assigns the family a high risk score based on past records and similarities to other cases. The score may be intended as support for a professional judgment. But if it follows the family into every meeting, it can also shape which questions are asked, what is interpreted as a sign of danger, and how much the family must do to be believed.
I am describing a hypothetical situation here, not a specific case. What matters is the mechanism: the institution's assessment has consequences already while it is being tested. The family no longer faces only a question of what is happening now, but an expectation of what it might go on to do. If that expectation is allowed to determine how each new sign is understood, it can become difficult to show that it was wrong.
An estimate is granted authority through the decisions built upon it. AI can make this transition faster and less visible, because the result arrives as a score, a ranking, or a recommendation. The numbers can be precise in form even when the basis is uncertain. Who chose the data? What counts as a good outcome? Which circumstances in a person's life could not be recorded? Whoever is affected by an assessment must be able to ask such questions and receive answers.
This does not mean the use is irresponsible. A warning can make us aware of a danger we would not otherwise have seen — but it must still be examined, with room for professional judgment and for the possibility that this particular person may act differently than those to whom they have been compared. Especially in work with children and families, we cannot let a probability become a characterization of who someone is.
Véliz's critique also applies to the prophecies about AI itself. The technology is sometimes presented as an inevitable solution, other times as an almost inevitable catastrophe. Warnings about genuine danger should be taken seriously. But here too we need to ask who is making the claim, what it is based on, and who gains greater room to act if we believe it. I am left with an impression from one of Véliz's more recent texts: that the future is shaped by what people do, and does not lie ready-made, waiting to be read off by technology companies.
Now I understand better why a democracy needs uncertainty. A choice must be able to have an outcome other than the one expected. A citizen must be able to change their mind. A person assessed by an institution must be able to show something other than what the numbers indicated. If we treat the outcome as given in advance, the conversation about what we ought to do is gradually replaced by a conversation about how we should adjust to what is supposedly going to happen. This is how I currently read Véliz's democratic argument.
Even when I write, I notice the difference on a small scale. AI can suggest a phrasing or point out an ambiguity. Sometimes this helps me move forward. Other times I first discover what I think when I object to the suggestion. If I let the first well-formed text determine the direction, something of the work of thinking would disappear. As a writing person, I need the same capacity to hesitate, object, and change my mind that I also want to safeguard on behalf of others.
Perhaps this is what philosophy can be an antidote to: the habit of hearing a prediction as a verdict. Philosophy cannot promise us that everything will turn out well. It can make us stop and ask what an estimate is based on, whom it empowers, and what we can still do. The oracle spoke as though the future were already known. Our responsibility begins where we still dare to answer.
No comments:
Post a Comment