Henri Bergson's doctoral thesis, published in 1889, is called Time and Free Will in F. L. Pogson's English translation of 1910. Its claim sits in the author's preface: "all discussion between the determinists and their opponents implies a previous confusion of duration with extensity, of succession with simultaneity, of quality with quantity". Bergson does not take a side in the old way. He argues that both sides share one mistake, a wrong picture of time, and that once it is corrected the objections to free will, the definitions given of it, and "in a certain sense" the problem itself, may disappear (preface, pp. xix-xx). He is not neutral about the result: "All determinism will thus be refuted by experience, but every attempt to define freedom will open the way to determinism" (p. 230).

The question has a new edge now. Bergson grants that we often "resign our freedom" through "sluggishness or indolence", and that when "our most trustworthy friends agree in advising us", their insistence forms "a thick crust which will cover up our own sentiments", so that "we shall believe that we are acting freely" (p. 169). Feeds and recommendations may be a newer source of such advice, and they are not friends. This piece sets out his argument, tests it against the strongest objection to it, and says where 000 parts from him, and what a reader can do to test whether a taste is really theirs.

The shared mistake

Start with the question underneath the whole book. You are about to make a choice. Could you have chosen otherwise? Determinism, Bergson writes, "claims that, given certain antecedents, only one resultant action was possible" (p. 174). Those who believe in free will assume that "the same series could issue in several different acts, equally possible" (p. 175). In everyday terms: either your history, your character and your circumstances left only one way for you to go, or more than one way was really open.

Both sides, he says, start from the same figure: a self that reaches a point and finds two directions "equally open", paths that "become things, real paths" lying ready to be taken (p. 176). Those who believe in free will read the figure as showing that both paths were possible. Determinists read it backwards from the act that was performed, and the same symbolism ends "in proving its absolute necessity" (p. 179). Underneath the two readings is a common postulate: both "take up their position after the action X has been performed" (p. 179). Both questions, he says, come back to one: "Is time space?" (p. 181).

Two ways to picture time

There are two ways to picture time. In the first, time is a line with moments marked off like a ruler, the way clocks and diaries show it. In the second, time is what you live: each moment arrives carrying the ones before it. Think of a disappointment. By the third week it is not the same disappointment it was on the first day, not because it grew, but because everything since has coloured it. (The disappointment is 000's example, not his.) Bergson says deep feelings work this way. Each is "tinged with the colouring of all the others" (p. 164), and deep states "occur once in consciousness and will never occur again" (p. 219).

The trouble starts when we describe ourselves with the ruler. "The greater part of the time we live outside ourselves", he writes, perceiving only "a colourless shadow which pure duration projects into homogeneous space". "Hence our life unfolds in space rather than in time" (p. 231). Freedom, on his account, is "the relation of the concrete self to the act which it performs" (p. 219), and the moment we analyse it from outside, "in place of the doing we put the already done" (p. 219).

Why counting is not time

Chapter I argues that the intensity of a pain, once the bodily reactions we measure it by are taken away, is a quality and not a magnitude: "the intensity of the pain would be a quality, and not a magnitude" (p. 38). On 000's reading that is the ground of the later claim that deep states cannot be broken into comparable elements. Chapter II turns to counting. To count anything you have to set the things side by side, and that is a picture in space. Bergson then asks what could separate time from space if time is pictured as a medium of equal moments. "if space is to be defined as the homogeneous", he argues, "it seems that inversely every homogeneous and unbounded medium will be space" (p. 98). A time of equal, separable moments is then already space. He states the conclusion flatly later: "from the moment when you attribute the least homogeneity to duration, you surreptitiously introduce space" (p. 104). 000 would add that feeds count: likes, streaks, ratings and minutes watched treat a taste as a quantity that can be added up, which is the move Bergson says misdescribes a deep state.

That is the weakest link, because time has an order and a direction, and one might call it homogeneous without making it a spatial medium. Bergson answers the point about order at pp. 101-102. We "could not introduce order among terms without first distinguishing them and then comparing the places which they occupy", so that "if we introduce an order in what is successive, the reason is that succession is converted into simultaneity" (p. 102). To order a series is to set its terms out by place. The doubt is whether an order of before and after needs terms laid side by side, or only a relation between them, and that is a claim about minds that this piece cannot settle. His positive picture is a tune: states that "perceive one in the other, each permeating the other and organizing themselves like the notes of a tune" (p. 105).

The case for freedom

Chapter III turns to free will itself. His reply to the fork picture runs through time. A real decision is not a self standing at a fork. The self "grows, expands, and changes" as it hesitates between two courses (p. 175), so "in reality there are not two tendencies, or even two directions, but a self which lives and develops by means of its very hesitations, until the free action drops from it like an over-ripe fruit" (p. 176).

A free act, then, is one that springs from "our whole personality" (p. 172). Freedom, he says, "is not absolute, as a radically libertarian philosophy would have it; it admits of degrees" (p. 166), and such acts are exceptional (p. 167). He is also generous to the other side. A change in our character, once it is blended with the self, "belongs to us" (p. 172), and the most authoritative education "would not curtail any of our freedom if it only imparted to us ideas and feelings capable of impregnating the whole soul" (p. 167). Bergson does not ask for an act with no history. He asks that the history be the person's own.

Peter and Paul

Then comes the argument the chapter is built to deliver. The determinist says that complete knowledge of a person's antecedents would let us foretell what they will do. Bergson takes the claim apart in two ways. First, take the antecedents as lived. Peter faces a serious decision, and Paul is to foretell it from all the conditions. But the deep states that issue in a free act "express and sum up the whole of our past history" (p. 185), so no detail of Peter's life can escape Paul, and Paul becomes "an actor who plays Peter's part in advance" (p. 187). Give him every detail and "Peter and Paul are one and the same person" (p. 188): there is nothing left to foresee, only something to be done.

Second, take the antecedents from outside, as a list. Then Paul must weigh each one, and he cannot know their weight "without knowing the final act, which is the very thing that is not yet known" (p. 190). Either way the question whether the act could be foreseen from all its antecedents is, in Bergson's words, "devoid of meaning" (p. 189).

Bergson is careful about what this achieves. The discussion "leaves the question of freedom exactly where it was to begin with" (p. 189). It clears away a wrong way of putting the question and does not yet prove freedom. His positive case comes in his account of causality, and he sums it up near the close of the argument, in answer to whether time can be represented by space: "Yes, if you are dealing with time flown; No, if you speak of time flowing" (p. 221).

The objection, and his replies

Now the objection. What Peter and Paul show is that nobody can foretell a free act without living the life that leads to it. A determinist can grant that. Determinism, as the Stanford Encyclopedia of Philosophy defines it, is not a claim about prediction. It says that, given the way things are at a time, "the way things go thereafter is fixed as a matter of natural law", and that, for finite predictors such as ourselves, predictability and determinism are "simply not logically connected at all" (Hoefer, 2023). So being unforeseeable is not the same as being unsettled.

Bergson gives three replies to psychological determinism. The first is about repetition. To say that the same inner causes will reproduce the same effects "is to assume that the same cause can appear a second time on the stage of consciousness" (p. 199), and in duration "the same moment does not occur twice" (p. 200). The deeper claim is that a repeated feeling is a new one: "the same feeling, by the mere fact of being repeated, is a new feeling" (p. 200).

The second reply is about what a law is. If causality means regular succession observed so far, then "by what right" do you apply it to "those deep-seated states of consciousness in which no regular succession has yet been discovered" (p. 203)?

The third goes to the root. Read as binding the future to the present, causality "could never take the form of a necessary principle", and Bergson gives his ground in the same sentence: "the successive moments of real time are not bound up with one another" (p. 208). Causality can be pictured in two ways. Either everything endures as we do, so that the future exists in the present only as an idea and the passage to it is an effort. Or duration belongs to conscious states alone, and things that do not endure as we do carry their future in their present mathematically. Each, taken alone, "safeguards human freedom" (p. 215), and the determinist's habit is to take the principle in both senses at once, passing "by imperceptible steps" from regularity to "mathematical necessity" (p. 216).

Does it hold?

Three points test his case: what lawful fixing needs, what Peter and Paul assume, and what consciousness can report.

A determinist has answers. A law can map one state to the next, and a state may include the whole history so far. Bergson's reply is that a history is not a state in his sense, since it cannot be broken into comparable parts (p. 200). A determinist can add that natural law does not need the "mathematical necessity" Bergson attacks: a law can fix the future without the future being contained in the present. Bergson charges the step from regularity to necessity, but he does not show that no other reading of lawful fixing is possible.

The Peter and Paul argument has a narrower reach than it first seems, and Bergson says so: it dissolves the foreseeability form of the question and leaves freedom where it was. Within that reach it still carries a premise. The second horn assumes that the weight of a deep feeling cannot be read off in advance. A determinist will say that in a lawful system the weight of each antecedent is fixed by the rule that carries it forward, not by the outcome. That is the live point, and the determinist's claim about law is as unshown as Bergson's claim about the weight of a feeling. The further premise that a deep state does not recur is separate, at p. 200, and he asserts it.

The premise of non-recurrence rests on what consciousness reports. Bergson holds it as a datum: "The self, infallible when it affirms its immediate experiences, feels itself free and says so" (p. 183). Libet's experiments (1983) found a readiness potential in the brain beginning several hundred milliseconds before subjects reported deciding to move a wrist, and critics read it as accumulated noise and not a decision (Schurger, Sitt and Dehaene, 2012). Bergson wrote nothing on such experiments. 000's reading, which goes beyond what he says, is that he would set a wrist movement among the habitual acts that are "for the most part insignificant" (p. 168), so the experiments do not bear on a grave choice. The fit is loose, since those acts answer impressions from outside, and a finding that experience misreports what causes our acts would still weaken the datum.

Where Bergson is exposed

At p. 152 Bergson allows that the study of the nerves may reveal "some new kind of energy which may differ from the other two by rebelling against calculation". That is hedged, but in 000's phrase it is a wager on physics, and an exposed one.

The deeper question is what is left that no law governs. If the act is explained by the whole self, that sounds like determination by the self, and Bergson accepts the word: to say that the soul is determined by one of its deep feelings "is thus to recognize that it is self-determined" (p. 165). On Dolbeault's reading he is an incompatibilist who also rejects classical indeterminism, since the picture of a choice among pre-fixed alternatives is what he attacks. The free act is a creation, formed by the whole self, which no law governs: "an indeterminism without the notion of alternative possibilities" (Dolbeault, 2020, p. 109). Whether that answer works, or whether "creation" is luck under another name, is the open question.

000's judgement is this. Bergson shows that foreseeability from antecedents is empty for deep states, and he gives real reasons why necessity read as logical containment cancels duration. What he does not show is that no rule of dependence can hold of a changing self, for that he needs duration to be more than what consciousness reports, and a determinist is free to treat the report as appearance. He himself allows that "if the causal relation still holds good in the realm of inner states, it cannot resemble in any way what we call causality in nature" (p. 200), and says what he takes it to be: a relation that is "purely dynamic", between "psychical force" and "the act which springs from it" (p. 219, from two adjacent sentences).

Where 000 parts from him

Bergson's mark of a free act, which 000 adopts, asks where the act comes from: whether it "springs from the self and from the self alone" (p. 172). 000 agrees with him that a count is a poor account of a person: we can rate a film, a meal or a day and still not have said what it was like. It agrees that freedom is not the default: "we are rarely free" (p. 231). And it agrees that the free act is the one that is yours, not merely one that nobody forced.

Where 000 parts from him is narrower than it first looks. Bergson is confident that the self can be found. We perceive it "by a strenuous effort of reflection", and "we can nevertheless always get back into pure duration" (p. 233). He also knows that influence can sit on the self. A "suggestion received in the hypnotic state is not incorporated in the mass of conscious states", he writes, yet "it will usurp the whole personality when its time comes" (p. 166). That is deliberate shaping by another, and he separates it from an influence that is taken in: "suggestion would become persuasion if the entire self assimilated it" (p. 166). So he does not judge by source alone. None of his examples is an influence that has been absorbed and was also built to work around the person's judgement.

That is 000's doubt, in two steps. The first is about how we know. If influence can be hidden, then a feeling of integration cannot rule it out, since hidden influence is, by its nature, influence the person "couldn't easily become aware of" (Susser, Roessler and Nissenbaum, quoted in Noggle's Stanford Encyclopedia entry). Bergson's mark survives as a definition, because it asks whether the act expresses the whole self and not whether it feels integrated. 000 keeps that definition and adds a second condition, of the kind Christman's history-based account uses and Bergson does not: whether the person would not have resisted the way she came to be as she is, had she seen how she came to be. The condition is non-resistance, not approval: she need not like the history, only not have rejected it. He describes two ways the mistake can show, and 000 takes each as a partial test: one delayed, "it is only by looking back to the past, later on, that we shall see how much we were mistaken", and one at the moment of acting, "something may revolt against it. It is the deep-seated self rushing up to the surface" (p. 169).

Both rest on the very report the doubt undermines, and the second also needs something to revolt against. A taste installed slowly, and made cheap to act on, may give the self nothing to push against, so the absence of a revolt is weak evidence. That is 000's addition, and it is why the practice below adds a check from outside.

The second step is where 000 leaves him. For Bergson, assimilation makes an influence one's own. 000 thinks a system can manufacture the assimilation.

Set two cases side by side. One person's taste in music was formed by a teacher she admired. Another's was formed, patiently, by a system built to hold her attention for its owner's gain. Each endorses the taste wholeheartedly. Every taste has a history, so history alone does not separate them. What can is how the influence works. The teacher's runs through the pupil's judgement: she is persuaded, and could be unpersuaded.

The second step rests on one premise that 000 does not prove: that a system built for attention can work by running around her judgement, through timing and repetition for example, which offer no reasons to accept or refuse. Its only support is the difference in kind: a reason can be weighed and refused, and a cue that carries no reason gives her nothing to weigh. How far real feeds work this way is unsettled (see below), and the contrast holds only so far as the premise does. If it holds, the system works whether or not she would approve, and her endorsement is then weaker evidence of her taste. If it fails, the two cases differ less than this paragraph says.

The test is whether she would have resisted the taste's formation had she seen how it was formed. Wholehearted approval of the result is not enough, and is the very report in doubt. That is not the genetic fallacy: the origin does not discredit the taste, and a taste can survive learning that it was installed. It is a heuristic and not a criterion, since it relies on the same report the first step doubted, and it sits close to historical accounts of autonomy such as Christman's, set out below. It bears on whether the taste is hers, and not only on the supplier's blame.

The argument has costs. Some teachers hide their method, and the dispositions that would resist may themselves have been installed by the same supply. And 000 is a supplier of taste too. It meets the test only if it shapes in the open, and says what it is doing.

This step does not need the claim that feeds actually shape taste. That claim is unsettled: a review calls causal evidence on the effect of recommenders on preferences "a major open problem" (Stray and colleagues, 2022). Philosophers of autonomy have met a similar difficulty: if a person can be manipulated so that each of her endorsements is reasonable, endorsement alone cannot settle whether she is autonomous, and some accounts add conditions on the history of her endorsements, though it is hard to draw the line without "implicitly appealing to the very phenomenon one is trying to explain" (Buss and Westlund). Christman's account makes autonomy depend on "the processes by which she came to be the way she is" (Stanford Encyclopedia).

The regress

Choosing the conditions is itself an act, so the question comes back: what makes that act free? This is the regress behind Galen Strawson's Basic Argument. You do what you do because of the way you are, to be truly responsible for what you do you would have to be truly responsible for the way you are, and you cannot be, since the same is true of the way you were before, and so on back (Strawson, 1994). Bergson did not take up moral responsibility in this book, and Dolbeault says his approach sets the question aside (pp. 94, 109). His own answer begins before Strawson: "Our character is still ourselves", and those "who ask whether we are free to alter our character lay themselves open to the same objection", since they split a person into two selves (p. 172). That is close to the identification reply that compatibilists make to Strawson. Whether it answers the regress or only changes the subject is disputed, and 000 does not settle it here.

000 does not need to settle which reply is right, because the practice asks a different question. The regress is about whether anyone can be free in the ultimate sense. The practice is about autonomy in the ordinary one: whether a particular taste came about through a process you would not have resisted had you seen it. Origin matters to that question locally, in the history of this taste, and not ultimately, in the history of the self that has it. It also means a limit. Bergson says that "any positive definition of freedom will ensure the victory of determinism" (p. 220), and the practice works on the supply, not on the act, so it cannot detect freedom in his sense. The hope is that arranging the ground beforehand makes the next act freer. It is a hope, not something the practice shows.

What to do with this

What follows is 000's, offered as practice and not as doctrine.

  1. Give one thing a duration: hear a whole record from the first note, eat a meal without a screen, watch a film without checking the time.
  2. Let one day go uncounted, then write one page about it in your own words, with no number in it, and keep it.
  3. List five things you chose this year. Beside each write when you began to like it, who or what first put it in front of you, and what you did with it afterwards: what you read, made, gave up or kept going for. Expect 000 to appear on some lists as well as the feeds.
  4. Ask one person who knows you well which of the five they think are yours, and why. Their answer checks your own impression, which the history of a taste may have shaped.
  5. Stop the supply of one for a fortnight, and replace it with something else of equal novelty. If you stop reaching for it, it may have depended on the supply, or on habit or novelty, which does not show it was never yours. If you go on reaching, it does not depend on the supply, though the supply may have planted it. This step shows dependence, not ownership. Steps 3 and 4 bear on ownership, and only as well as your own account allows.
  6. Make one thing from whatever you kept: a meal, a playlist, a page, a drawing. Give it to someone who did not ask for it.

A note on step 5. In one randomised trial with a pre-analysis plan (Allcott and colleagues, 2020), people were paid to deactivate Facebook for four weeks. Later use fell, and the authors say the fall is consistent with habit formation or with learning. In the subsample of iPhone users who reported usage as recorded by the phone's Settings app, the fall was less than half as large and not significant. One person's fortnight is weaker than a trial of many people over four weeks.

None of this makes a person free in Bergson's sense, and he would say that freedom is rare. The practice arranges the ground. Whether a free act follows is not something it can show. When did a choice last feel like an expression of the whole of you, and not like the next item in a series?

Sources

Quotations are from F. L. Pogson's translation of Time and Free Will, as Project Gutenberg prints it, from Bergson's running text only: the marginal summaries in that edition are the translator's, and none is quoted. Page numbers are those of that printing. Joël Dolbeault, "Bergson's Theory of Free Will", Journal of French and Francophone Philosophy 28(2) (2020), pp. 94-115. Stanford Encyclopedia of Philosophy: Carl Hoefer, "Causal Determinism"; Sarah Buss and Andrea Westlund, "Personal Autonomy"; John Christman, "Autonomy in Moral and Political Philosophy"; Robert Noggle, "The Ethics of Manipulation" (for Susser, Roessler and Nissenbaum, "Technology, Autonomy, and Manipulation", Internet Policy Review 8(2), 2019). Galen Strawson, "The Impossibility of Moral Responsibility", Philosophical Studies 75 (1994), pp. 5-24. Benjamin Libet, Curtis Gleason, Elwood Wright and Dennis Pearl, "Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential)", Brain 106(3) (1983), pp. 623-642. Aaron Schurger, Jacobo Sitt and Stanislas Dehaene, "An accumulator model for spontaneous neural activity prior to self-initiated movement", PNAS 109(42) (2012), E2904-E2913. Jonathan Stray and colleagues, "Building Human Values into Recommender Systems" (2022, arXiv 2207.10192). Hunt Allcott, Luca Braghieri, Sarah Eichmeyer and Matthew Gentzkow, "The Welfare Effects of Social Media", American Economic Review 110(3) (2020), read in the NBER working-paper version.