frameworks / bovens-adaptive-preferences · article
frameworks/ · 2,914 w · 13 min · ✎ dialogue

Bovens (1992) — Sour Grapes and Character Planning

Luc Bovens's "Sour Grapes and Character Planning" (The Journal of Philosophy, Vol. 89, No. 2, February 1992, pp. 57–78) is a canonical paper in the adaptive-preferences literature. Writing as a graduate student at Minnesota (the paper is "a shortened version of the first chapter of my Ph.D. dissertation, Reasons for Preferences"), Bovens sets out to solve a puzzle Jon Elster raised in Sour Grapes (1983): why do we call one preference shift irrational (the fox deciding the grapes are sour after he fails to reach them) and another preference shift rational (the poker player who decides to stop cheating and, through habituation, comes to enjoy fair play)? Both involve changing what you want in response to what you can get. Both can be described in the same words. Yet our intuitions about them differ radically. Bovens argues Elster's own solution — that the rational case (character planning, CP) is intentionally acquired while the irrational case (sour grapes, SG) is not — fails on its own terms. He then constructs a better criterion grounded in Donald Davidson's theory of practical reasoning. For Jason's deep-ambition-book-thesis, specifically the "letting go vs. giving up" chapter framed in motivation-and-goals and eudaimonia-vs-hedonia, Bovens supplies the sharpest available philosophical vocabulary for the exact problem the chapter is about: how to tell, from inside, whether your preference shift is a cope or a genuine reorientation.

The Puzzle

Bovens opens with an ancient tension in the Stoic Lebensregel — "mold one's desires such that they come to match whatever it is that life has in store." The opening epigraph from Marcus Aurelius makes the point directly: "Fit yourself into accord with the things in which your position has been cast, and love the men among whom your lot has fallen, but love them truly." Bovens admits the advice is "well-taken in certain cases" but "strictly off the mark in others." He offers two caricatures on either side:

  • Irrational adaptation (Aesop's fox). "The fox in La Fontaine's fable has an appetite for grapes, though when he finds out that he cannot reach them, he claims that grapes are too sour for his canine taste anyway." Bovens is careful: on his reading of the fable, "the fox does not change his beliefs about the sourness of the particular grapes in question, but rather changes his tastes concerning the sourness of grapes in general." This is pure sour-grapes (SG).

  • Rational adaptation (the reforming poker player). "Imagine a poker player who enjoys the game tremendously because of the opportunities for cheating involved. She comes to realize that cheating is morally reprehensible and decides to start a better life on this score. Initially she does not find fair-play poker terribly exciting. Nonetheless, in her quest for Aristotelian virtue, she is committed to becoming the kind of person who enjoys fair play. And in order to carry through this project of CP, she chooses the Aristotelian route of habituation." This is character planning (CP).

Bovens generalizes beyond the moral case: the poker player may also choose to reform her preferences because cheating is no longer feasible — say, the casino installs proper lighting. "It may well turn out that, after a few games, she will indeed come to enjoy fair-play poker quite a bit better than her former style."

The puzzle: "both the fox and the poker player adjust their preferences in reference to what they believe to be feasible alternatives." Same description. Different verdicts. What structurally distinguishes them?

Why Elster's "Intentionality" Answer Fails

Elster's solution was to appeal to autonomy via intentionality. Preferences are rational only if rationally acquired, and rationally acquired only if intentionally acquired. CP is intentional (caused by a metadesire — a desire about what to desire); SG is caused by a non-conscious "affective drive" that reduces frustration from unsatisfiable desires. Drives aren't mental states the agent has first-person knowledge of, so SG adaptation is a "mere doing," not an "action," and therefore cannot be rational.

Bovens's Section II and III dismantle this in two moves:

  1. First-person knowledge is not a necessary condition for intentional action on Davidson's own theory (which Elster claims to accept). Rational-choice theory routinely explains actions by unveiling preferences the agents don't consciously endorse — and Elster is committed to rational-choice explanations being paradigmatic intentional explanations. So Elster can't hold both (i) first-person knowledge as a requirement for intentional action and (ii) rational-choice explanations as intentional. The position is internally inconsistent.

  2. Intentional acquisition is neither necessary nor sufficient for rationality of preferences. Bovens gives counterexamples: you can acquire preferences through learning (enrolling in a wine-tasting course to develop a taste for dry wines) or through habituation (years of North American living that shift peanut butter up your sandwich-spread ranking), and these shifts can be fully rational whether or not they are intentional. "Intentional preference acquisition may have very little to do with the ascription of rationality to preferences."

The second move is the important one for how the paper gets used downstream. It generalizes the CP family well beyond the Aristotelian habituation case: learning, socialization, and non-reflective absorption of culture can all produce rational preference changes. Bovens writes: "One can hardly deny that at least some preferences so acquired are autonomous in character." This matters for the Deep Ambition reader whose values shifted through becoming a parent, getting older, or watching a friend die — preference shifts that are neither pure SG nor explicit CP projects but are nonetheless clearly rational.

Bovens's Positive Theory — the All-Things-Considered Judgment

Bovens's alternative runs parallel to Davidson's theory of akrasia (weakness of will). Preferences are rankings defined over state-of-affairs (SOA) tokens — for instance, "my owning this Volvo now" vs. "my owning this Austin now." Unlike desires, preferences are nonrelational: they are not indexed to a particular reason. You can desire A over B (for one reason) and desire B over A (for another reason), but you cannot have a preference for A over B and a preference for B over A simultaneously. A preference is the output of arbitration across all reasons.

Bovens builds up the machinery carefully:

  • Criterial judgments — relational rankings of SOA tokens relative to one reason pair (e.g., "considering country of origin, the Volvo is better than the Austin").
  • All-things-considered judgment — the relational ranking relative to all the agent's relevant reasons. Arbitration across all criterial judgments.
  • The rationality constraint (R'). "A preference for SOA token s over SOA token t is rational only if it results from proper arbitration over all criterial judgments defined over s and t, implied by the reason pairs that the person takes to be relevant to the ranking at hand."

This is the Davidsonian "principle of continence" — an action is rational only if informed by the agent's all-things-considered judgment — applied to preferences instead of to actions. Akratic preference is the new technical object.

The SG/CP Distinction, Precisely Drawn

With (R') in place, Bovens draws the distinction. Let s and t be feasible alternatives, and suppose s is no longer feasible.

  • Sour grapes (typical case). The agent adjusts her preference from s over t to t over s, but none of the criterial judgments, none of the underlying reasons, and none of the relative importance weights change. Assuming beliefs and reason pairs are fixed, the all-things-considered judgment is still that s > t. The new preference contradicts the unchanged all-things-considered verdict. It cannot have resulted from proper arbitration. It is akratic with respect to preference, and therefore irrational. "The adjusted preference cannot be informed by the all-things-considered judgment, i.e., it cannot result from proper arbitration over all criterial judgments. And this is what accounts for the irrational character of preferences acquired through SG."

  • Character planning (typical case). The agent adjusts her preference and her reasons. Either she shifts some of the criterial judgments (e.g., now believes fair-play poker is genuinely more fun), or she changes the relative importance attached to each consideration. The all-things-considered judgment is now in flux along with the preference. "Hence, over the course of the CP move, my preference in flux remains informed by the all-things-considered judgment in flux. In other words, my preference results at all times from proper arbitration over the relevant criterial judgments. And this is why there is nothing irrational about preferences acquired through CP."

Bovens illustrates with a job example, which is especially useful for Jason's coaching work. Imagine wanting a job with "well-paid overtime work" and "challenging career opportunities." That new job becomes infeasible. In the SG case, you keep holding that the new job would be better on every dimension that matters to you, yet you now prefer your actual job. Your preference is irrational — it's saving you from frustration at the cost of coherence. In the CP case, you actually come to value the charm of leisure time and the absence of promotion pressure; your ranking of reason types itself shifts; the preference then follows rationally from the revised reason pairs.

The Continuum and the Gray Zone

Bovens's criterion doesn't force a binary verdict. In Section V he shows the theory predicts the intuitive gray zone: between typical SG (no reason changes) and typical CP (complete reason restructuring), there is a continuum of cases where the agent shifts some criterial judgments or some importance weights but not enough for the all-things-considered verdict to complete the flip. These cases are genuinely neither — and Bovens is clear that "many real-life or fictional cases of preference adjustment are actually located on this continuum."

He also accounts for three auxiliary facts about our SG-vs-CP intuitions:

  • Timing. SG shifts tend to be "at the spur of the moment," while CP is "a project of change" involving many adjustments over time. A sudden flip raises suspicion of SG precisely because restructuring a full set of reasons takes time.
  • Self-deception temptation. SG invites a description in which the agent secretly still holds the old preference and is merely deceiving herself. Bovens thinks this is often an overreach — there are genuine cases where introspection would confirm the new preference — but the temptation to read SG as self-deception is persistent and his theory explains why.
  • The principle of charity. We are under pressure, when attributing preferences, to assume that a person holds preferences that are rational given their reasons. SG is precisely the case where we cannot satisfy this third presumption along with the introspective and behavioral presumptions. "My theory of rational preferences drives a wedge between SG and CP without describing the former phenomenon in terms of self-deception. At the same time, it can explain why there is a prima facie plausibility to such description."

Why This Matters for Deep Ambition

The Deep Ambition reader's central phenomenology — described in the "Treading Water" material from outputs/2026-04-18--letting-go-vs-giving-up-psychology-synthesis.md — is that she cannot tell, from inside, whether her shifted values are wisdom (letting go) or defeat (giving up). Bovens gives this exact problem a technical name and a diagnostic structure:

  1. The question is not "did your preferences change?" They obviously did. The question is whether your reasons changed along with them. If your ranking of what kinds of considerations matter — relational depth vs. status, craft vs. prestige, present-with-family vs. career acceleration — has actually restructured, you're in CP territory and the new preference is rational. If only the ranking of alternatives has flipped while all the underlying reason-weights remain where they were, you're in SG territory and the new preference is post-hoc rationalization.

  2. Self-interrogation has a specific form. Not "do I really want this?" (introspection underdetermines the answer, as Bovens notes) but "what are the reason pairs I take to be relevant here, and has their relative importance to me actually shifted, or am I holding them constant and just flipping the ranking?"

  3. Timing as diagnostic. Bovens's point about SG being typically abrupt while CP is a long process over time maps onto Jason's coaching observation that authentic value shifts tend to involve a "project of change" — often years of events, reflection, and recommitment. A sudden post-failure flip is structurally suspicious. A slow drift accompanied by many small reorderings is structurally trustworthy.

  4. The gray zone is real, and treating it honestly is the mature move. Bovens's continuum analysis vindicates the "you can't fully tell from inside" stance. Neither "it's all sour grapes" (hyper-vigilant self-accusation) nor "it's all character planning" (premature redemption) is the right default. The question is how far along the restructuring actually is.

  5. This is the philosophical underpinning for authentic-pride-patterns. That coaching exercise surfaces the operating patterns the client actually values in action — i.e., the reason-pairs and their relative weights — rather than relying on their summary self-report of "what I want now." It's a practical workaround for the problem Bovens identifies: you can't verify a preference shift from the preference alone; you need access to the underlying reasons and their arbitration. Authentic-pride moments are empirical windows onto that underlying structure.

  6. It links directly to self-determination-theory. SDT's internalization continuum (external regulation → introjected → identified → integrated) is, in Bovens's vocabulary, a map of how far criterial judgments and their weights have been restructured. Integrated regulation is the full CP endpoint — you don't just prefer the new path, your whole reason-pair set has absorbed it. Introjected regulation is closer to the SG failure mode dressed up in aspirational language: the behavior flipped, but the underlying reasons haven't.

Key Direct Quotes for Book Citation

  • On the problem: "What is puzzling about SG and CP is that both phenomena can be described in the same way, namely, both the fox and the poker player adjust their preferences in reference to what they believe to be feasible alternatives. Nonetheless, while the fox is an epitome of irrational preference acquisition, the same judgment could hardly be passed on to the poker player." (p. 59)
  • On why SG is irrational: "The adjusted preference cannot be informed by the all-things-considered judgment for the ranking at hand... this is what accounts for the irrational character of preferences acquired through SG." (p. 74)
  • On why CP is rational: "Over the course of the CP move, my preference in flux remains informed by the all-things-considered judgment in flux... And this is why there is nothing irrational about preferences acquired through CP." (p. 74)
  • On timing: "The SG move typically is a response at the spur of the moment, while CP typically is more of a long-term process. If preferences are abandoned at the spur of the moment, there is a presumption of irrationality, which there is not if such occurs gradually." (p. 77)
  • Against self-deception as the distinguishing move: "My theory of rational preferences drives a wedge between SG and CP without describing the former phenomenon in terms of self-deception. At the same time, it can explain why there is a prima facie plausibility to such description." (p. 78)

Positioning Against the Literature

  • Against Elster (1983). Bovens accepts Elster's diagnosis of the puzzle but rejects his autonomy/intentionality criterion as both internally inconsistent and extensionally wrong.
  • Parallel to Davidson on akrasia. Bovens extends Davidson's Principle of Continence from actions to preferences. The akratic preference — a preference that flouts the agent's own all-things-considered verdict — is his technical novelty.
  • Ancestor to Callard's aspiration. Agnes Callard's Aspiration: The Agency of Becoming (2018, already in the Deep Ambition bibliography via deep-ambition-book-thesis) resolves the SG/CP problem from a different angle — aspiration as the agency of coming to hold values you do not yet hold, via acting-as-if. Bovens and Callard are telling the same story from different sides: Bovens from the internal structure of rational preference, Callard from the phenomenology of becoming.
  • Grounds for eudaimonia-vs-hedonia. The eudaimonic/hedonic distinction aligns with Bovens's reason-pair restructuring: eudaimonic orientation is the mature CP state where intrinsic reasons have taken on higher weight; hedonic-treadmill chasing is closer to SG dynamics where the ranking shifts without the reason-weights restructuring.
Thread · 0 replies+ add reply
no replies yet — be the first to write back to this article.