Conclusion: What Philosophy Is For

8.1 The argument in retrospect

I began with Cephalus, who knew what justice was until Socrates showed him a case in which the two things his concept did came apart. I have argued that this is the form of every persistent problem in philosophy, and that seeing this changes what we should think about philosophy's disagreements, its progress, and its point.

The argument had five steps. The concepts at the centre of philosophy's persistent problems are polyfunctional: each does several jobs in our practices, each job imposes its own application condition, and the conditions coincide across the range of cases from which the concept is learned (T1). The problems arise where the conditions come apart, and the thought experiment is the instrument by which such cases are constructed in advance of the world's producing them; this explains why each thesis of an aporetic cluster is compelling, why exotic cases are indispensable, and why the literature accumulates distinctions while the headline questions stand (T2). The positions in a persistent debate are prioritisations of functions, and what remains of the dispute once its factual and verbal layers are removed is a normative question about which function the shared concept should serve, a question that is rationally stable because nothing in the concept fixes the weights, the ends are plural, and no practice has authority over the rest (T3). Progress consists in locating divergence, articulating functions, disentangling them, and relocating the residue to the ethics of the relevant practice; this progress is cumulative, agreed across doctrinal divides, and constitutes understanding, and convergence on the headline is not to be expected (T4). The concept of progress is itself polyfunctional, and the metaphilosophical debate is a divergence case that the account predicts (T5).

The account was defended abductively, as the best explanation of eight features of philosophy's problems, through four detailed case studies and several sketched ones, and against objections from triviality, realism, conventionalism, circularity, demotion, and the apparent counterexamples of philosophy's convergent successes. The objection from circularity it meets only in part, and §7.5 says which part. It was shown to have limits, and to make predictions that could fail.

8.2 Consequences for method

If the account is right, several things follow about how philosophy should be done.

Thought experiments are vindicated, and reinterpreted. The recent scepticism about the case method (Cappelen 2012; Deutsch 2015; Machery 2017) and the recent defences of it (Nagel 2012; Williamson 2007) have shared an assumption: that the verdict elicited by a case is evidence about the extension of a single concept, to be trusted or distrusted as such. On the present account the verdict is a weighting, and the case is a device for discovering how many functions a concept has and where they diverge. So understood, the method is not merely defensible but indispensable: there is no other way to map a divergence region. But the verdicts must be read as what they are. A verdict on a balanced divergence case is not data about "the concept"; it is data about which of the concept's functions the case and its framing have made salient, and about the weighting of the person who gives it. The correct use of a divergence case is to map, not to test a unitary analysis.

Conceptual analysis is reconceived as function-mapping. The traditional project of finding necessary and sufficient conditions for a concept's application presupposes that there is one application condition to find. For polyfunctional concepts there is not, beyond the coincidence range, and the project's history of failure is explained. What replaces it is the mapping of a concept's functions, their application conditions, their coincidence range, and their divergence region, which is what Strawson (1992) called connective analysis and what the record of §6.3 in fact contains.

Experimental philosophy is reconceived as the measurement of weightings. The "restrictionist" reading of experimental findings, on which framing and demographic effects discredit the case method, and the defensive reading, on which the effects are noise to be filtered, both mistake what the effects are. They are measurements of function salience and of the distribution of weightings across a population, and as such they are data about the practices the concept serves. The finding that abstract framing elicits incompatibilist verdicts and concrete framing compatibilist ones is not an embarrassment to be explained away but a discovery about the functional structure of "responsible". A research programme that treated it so would ask which frames engage which functions, how weightings vary with the practices a population participates in, and whether professional training shifts them.

Conceptual engineering is recognised as the explicit form of what philosophy has always done in the residue. The recent literature on conceptual engineering (Cappelen 2018; Burgess, Cappelen, and Plunkett 2020) has been received as a departure from the tradition of analysis. On the present account it is the tradition become self-aware. Every doctrinal position in a persistent aporia was already a proposal for how a shared concept should behave in its divergence region; conceptual engineering says so openly, and asks what constraints such proposals should meet. The account supplies one constraint: an engineering proposal must be made with the map in hand, since a redesign that ignores a function the concept serves will wreck the practice that needed it.

Pedagogy. Persistent problems are usually taught as puzzles with a hidden answer that the student is invited to find. They should be taught as maps: here are the functions the concept serves, here are the cases that separate them, here are the costs of privileging each, and here is the question that remains once the map is drawn. A student who has learned that the compatibilism debate is, in its residue, a question about what blame is for has learned something true and has been placed in a position to think about it. A student who has learned that compatibilism is a thesis awaiting proof has been misled.

8.3 Consequences for the discipline's self-image

Kant's image of the battlefield, and the scientistic envy that has accompanied analytic philosophy since Russell, both rest on a comparison: philosophy is measured against inquiries whose questions have a different structure, and found wanting. The account replaces the comparison with a description. Philosophy is the discipline that maps, and where it can, legislates for, the divergence regions of the concepts that several of our practices load at once. It does this in advance of need, by constructing the cases the world has not yet produced; it does it cumulatively, by individuating functions that stay individuated; and it does it without the authority to impose a weighting, because the practices it serves have no common sovereign. Its persistent disagreements are the shadow of that last fact, and they are the price of serving plural practices honestly rather than legislating for one.

This is not a diminished self-image. It is, I have argued, the one that Socrates and Aristotle held, and it is the one the record of the discipline's actual achievements supports. The tradition's great results, from sense and reference to production and dependence, have the form the account describes, and the tradition's great unsettled questions have the form it predicts. A discipline that has done what philosophy has done does not need to apologise for not having done something its questions did not permit.

8.4 Future work

The account opens several lines of inquiry, which I list in order of the confidence with which I expect them to be fruitful.

An empirical programme. The decomposed survey and the salience manipulation of §7.9 are the most direct tests. A broader programme would measure the functional load of concepts, by counting the functions separable by cases and the practices they serve, and test the predicted correlation with the persistence and depth of the corresponding philosophical problems.

A formal model. The notions of §3.5 admit of formalisation: a concept as a bundle of application conditions over a space of cases, with a correlation structure over the space reflecting the environment of acquisition; the coincidence range and divergence region as regions of that space; functional load as a property of the bundle. Such a model would allow the account's predictions to be stated precisely and would connect it with formal work on learning from correlated features.

A historical programme. The history of philosophy could be reread as the history of conceptual fission: Aristotle's pollachōs legetai, the scholastic distinctions, the early modern separation of primary from secondary qualities, the Kantian separation of the analytic from the synthetic and the a priori from the a posteriori, and the twentieth century's record. The account predicts that each major disentanglement will have been forced by a divergence case and will have been followed by a stable dispute over the residue, and the history could be checked against the prediction.

Applied work. Machine learning is producing divergence cases for "agent", "author", "decision", "responsible", "understand", and "know" faster than any technology before it, and the practices that load these concepts, law, science, education, and ordinary life, have begun to legislate. The account says what the legislation requires: a map of each concept's functions and of what each proposed weighting would cost. Drawing those maps is work philosophy is placed to do and is not yet doing systematically.

Metaethics. The account is neutral on the status of the normative residue. Whether normative questions have objective answers determines whether philosophy's persistent disputes have them, and the account gives metaethics a new reason to matter: it is the theory of the residue.

8.5 Coda

Socrates did not tell Cephalus what justice is. He showed him where his concept ran out, and thereby gave him a choice he had not known he had, and a reason to think about what he wanted justice for. Cephalus, the dialogue tells us, went off to attend to the sacrifices, leaving the argument to his son. It is easy to read this as the failure of the exchange. It is better read as its success. The old man had learned that a concept he had used without difficulty all his life had two voices, and that in the region where they disagreed the question was not what justice is but what it should be for. That is not a discovery about a word. It is the beginning of ethics.

Wittgenstein asked what the aim of philosophy was and answered: to show the fly the way out of the fly-bottle (1953, §309). The account offered here suggests an amendment. The bottle has several exits, one for each of the practices that brought the fly in; the philosopher's task is not to choose one for the fly but to show it that they are there, that they lead to different places, and that where it goes is a question about where it needs to be. That is not the end of philosophy. It is what philosophy, at its best, has always been.