​Chapter 14

Freedom stays a postulate, and the word needs a precise sense before it can carry weight.

The postulate is a primitive relation. Chance also supplies the absence of a cause, and the postulate claims more: a question stood open, the settling of it established the sufficiency of a ground, and the author came into being. The relation is positive and basic, offered without analysis into components. Somebody may reasonably decline to accept it. Accepting it while holding that nothing has been added beyond indeterminism is not.

Kant is the nearest precedent. Theoretical reason cannot prove freedom, and the practical position cannot be occupied without treating oneself as capable of initiating action; a person deliberating is not forecasting his own behavior, and the two attitudes cannot be held toward the same choice at once. The practical asymmetry is accepted here, the derivation from the moral law is not, and the free act stays in the world. It occurs in the physical event, at a particular table and a specific moment. Kant purchased freedom by moving it out of the appearances, and the purchase is refused, the cost of refusing it being the empirical indistinguishability already conceded. Nothing later restores the purchase. A Name is performed in a settling, and a settling occurs in the world where everything else occurs.

The natural objection is that a postulate of this kind is empty. It predicts nothing, forbids nothing, and leaves an observation where it was; a description that makes no difference to what happens has the appearance of a decoration. The objection would land if the postulate were competing with physical explanation for the same work. It is not. It changes what the event is taken to be, and the content of that claim is definite: a question was resolved rather than terminated, a ground was accepted as sufficient rather than found to be weightier, an answer was given rather than an outcome produced. Unverifiable and empty are different failings, and only the first has been shown.

The same structure appears in scientific theory choice. Evidence does not select a theory. Any body of data is compatible with multiple accounts of it, and a failed prediction can be accommodated by adjusting something apart from the hypothesis under test — an auxiliary assumption, a background theory — so that no experiment refutes a hypothesis on its own. Duhem showed this about physics, Quine generalized it, and Kuhn drew the consequence for what scientists actually do: when two frameworks are in play, no algorithm ranks them. Simplicity, scope, fruitfulness, and accuracy all bear on the choice and do not order it, since they conflict and no rule assigns their weights. What closes the question is that someone closes it. A scientist commits to a research program. The commitment is neither read off the evidence nor arbitrary: he can give reasons, though they do not compel. The scheme is the same: an open question, grounds in view, none of them sufficient, and a settling that establishes sufficiency. Whoever says the postulate makes no difference to anything is committed to holding that theory choice in science is either mechanical or random, and it is visibly neither.

James saw what follows when proof is unavailable on either side. Declining to choose is a choice, made under the uncertainty and carrying its own risk; the man who suspends judgment has taken a side while calling it neutrality. The determinist wagers that behind every settling lies a sufficient cause, not yet found, awaiting a finer instrument. The libertarian wagers that in some settlings the causal chain genuinely fails to close and authorship occurs. Causation is also a wager that has forgotten it is a wager.

The determinist has behind him a long record of missing causes found, and every reason to expect the record to continue. The libertarian has nothing comparable in predictive reach. What he has is imputation: a judgment that holds a person answerable for a deed, a verdict that distinguishes the guilty from the unlucky, an ethical demand that presupposes the act was someone's. Imputation is a settling: grounds are taken as sufficient and a question is closed by someone. It therefore requires no prior verification that this act was authored, and the unverifiability established above leaves it intact. It requires that authorship be possible at all.

Sapolsky, who follows the chain without flinching, dismantles responsibility along with freedom — guilt disappears, and with it the distinction between a crime and a symptom. The compatibilist stops short of that and keeps responsibility without authorship: a man is answerable when the deed issued from him in the right way, and no undetermined joint is needed. Such a verdict reaches a nature. It sorts the dangerous from the harmless, tells us whom to restrain and whom to release, and does the work of protection. It cannot mark the difference between a man who settled the question and a man in whom it was settled — the guilty and the unlucky come out the same, since both acted from what they were. The practice that treats them differently depends on that primitive. The two are level at the foundation: each rests on a primitive it cannot derive, and each is backed by a practice too large to abandon.

A wager costs something either way. Accepting the postulate means holding a claim no evidence will confirm and no evidence will refute, which is an uncomfortable commitment, easily mistaken for a claim about mechanism. Declining it means treating a settled question as a terminated process, an avowal as a report on a prior state, and an act of taking responsibility as a description of how the machinery ran. The atrocity in the newspaper takes the treatment easily — a monster, a sick man, a product of what was done to him, the deed dissolved into its conditions. A friend who lies for money does not. The determinist wants an answer from him and not from his greed, and the wanting is addressed to somebody his own account does not contain. Nothing follows from the inconsistency about how the world is put together. This is the price of the second wager, paid by people who have not agreed to pay it.