​Chapter 2

If John is free, the freedom is in the settling. The determinist holds that John closed the examination where he did because of what he was before — the temperament that made him capable of overriding a prohibition under pressure, the upbringing that ranked need above property, the history that made this option feel decisive and that one feel thin. Schopenhauer set the formula: a man can do what he wants, but he cannot want what he wants. The examining is the doing; what made a conclusion feel final is the wanting; and the wanting was not wanted.

A ground that shaped John can enter the spread and become revisable. John was raised in a church that named theft a sin against God, and the prohibition went in early and deep; at the table he closed the examination against it. If the settling merely executed what was written, the writing was there and the deed contradicts it. Whatever the settling is, it can take a ground that formed it, hold it out as revisable, and let a conclusion stand that the ground forbids.

The determinist answers that the prohibition's defeat in this case does not mean that the contest had no winner. Hunger was present in John too, and the fear of dying was carved deeper than any catechism; a resentment carried since childhood may have been waiting for its occasion. What looks like a man overturning his formation is a part of the formation prevailing over another. No external force overturned it — the outcome fell out of the material that happened to contain more than the catechism. Revision of a ground is not escape from all grounds; it is the grounds settling among themselves, and what looks like a closing performed is the place where that outcome surfaced.

Strawson tightens this into a regress. To be ultimately responsible for how he closed the case, John must be responsible for the character that closed it that way — for the standards by which a consideration counted as sufficient. To be responsible for those standards, he must have set them; to have set them, he must have done so for reasons, which were themselves standards he would then have to be responsible for having. The regress does not terminate. At any level the question returns, and an answer requires an earlier act of self-formation that John was not present to perform. Frankfurt's higher-order desires do not escape it — a desire about a desire is another member of the spread, and the same question runs on it.

Even if the values were carried to the table, the capacity to weigh them may still be called John's. He can reason, hold a conclusion up and judge it; the operation is the freedom, whatever material it operates on. The faculty has a provenance of the same kind as the content: genes laid down the architecture that lets John reason at all, experience tuned how it runs, and education fixed what counts as a good reason. A finely wired thermostat, capable of checking more inputs and correcting further back, is a thermostat whose wiring was done elsewhere. The capacity to judge is as caused as the material it judges.

Active inference gives Strawson's regress a naturalistic form. On Friston's picture, the brain has no central decider. It runs a hierarchy of models: each predicts the level below, updates from prediction error, and receives its weighting — precision — from the level above. The weights are reassigned without a homunculus doing the job, since every level does to the level below it what was done to it from above.

Ask what revised the weighting at any level and the answer names the level above. Ask at the top and the hierarchy terminates in priors the organism did not set — laid down by evolution, by development, by the history of what it has encountered. The regress is not infinite, which was not the difficulty. The issue is that each rung runs on a setting supplied by another rung, and the last rung relies on a parameter nobody set. A hierarchy is a more exact description of the thermostats nested in thermostats, worked out to the point of predicting behavior, and it arrives where the crude picture landed: at a bottom that was not chosen.

Another reading removes history from the settling. Suppose the considerations came close enough to balance — B21 and B22 nearly matched, and the evaluation could have closed on either side. Then the side on which it closed was no more than the face a die showed. No upbringing required it, no wiring; the closing was genuinely open, and open is all it was. Chance is nobody's deed. This clears the closing of the causal chain and puts nothing in its place — no John in front of it.

Freedom would need a third reading, on which the settling is posited: laid down by John, in the act of closing, neither carried in nor fallen. Reflection cannot end what it holds open, and a free settling would have to close the deliberation without being produced by any consideration within it.

The first two interpretations appear exhaustive because both place the settling on the axis of derivation. Either the closing was produced by what came before John decided, or it was not; if it was, the chain holds, and if it was not, there is a chance. The third reading cannot be a point on that axis. It answers a different question: by whom, rather than where from. The settling was posited by John, and John becomes its author. The relation that matters joins the act to its owner. Chance and freedom share their back face — neither is derived from the prior state — and part at the front: chance has no author.

That the settling is not another member of the spread follows from the structure: a consideration extends the reflection rather than ending it. Whether the settling was derived from the prior state, and whether it had an author, do not follow from anything shown so far. The scene does not distinguish a closing John performed from one that happened in him. From outside, the deed looks the same; from within, the feeling of authorship cannot establish which occurred.