Contribution
Before this essay Bratman had an account of self-governance at a time, as guidance by a coherent, plan-infused standpoint. This essay gives the account over time. Its new contributions are two arguments for the “plan-theoretic model” of the cross-temporal links, the “shared agency model” behind the second, and the end of one’s own diachronic self-governance, which he uses to explain willpower and steadfastness without making them brute tidiness or rigidity.
Key insight
What makes a person’s agency over time hers, and self-governed rather than merely managed, is the structure people have when they act together. Bratman’s intention now that I do A later is meant to work through my later intention. My later intention is meant to follow through on the earlier one, and the sub-plans are meant to mesh. That is how two people who share an intention are related. Treating oneself across time as a partner, not a bargaining opponent, explains why abandoning a resolution or shuffling between options can be a lapse in self-governance even when each choice is self-governed at its own time.
Assumptions
- Synchronic self-governance as standpoint-guidance. “Self-governance involves guidance of thought and action by the agent’s relevant practical standpoint … and the standpoint itself needs to be sufficiently coherent to constitute a clear place where the agent stands.” This avoids “a little person in the head who is pulling the strings”. It is taken from earlier work.
- Over-time/at-a-time coordination. “An individual’s self-governance over time needs to accommodate her self-governance at each relevant time.” This is presented as a “metaphysical constraint”, not a norm of rationality.
- Bratman’s theory of shared intention (Shared Agency, 2014): (A) interlocking intentions, (B) intended mutual responsiveness and meshing sub-plans, (C) correct belief in persistence interdependence, (D) expectation of success, (E) common knowledge. These are offered as sufficient conditions for robust shared intention, not necessary ones.
- Hierarchical, partial plans. Continuity can hold at a higher level of plan while sub-plans change.
- Ends are not intentions. “Not all intentions are ends”, and “it is common to have ends one knows are not co-realizable”.
Key results
The essay’s model (§9), in four parts:
- Planned temporally extended activity with a time-frame, which “will commonly include one’s relevant later responses to one’s then-earlier efforts”.
- Synchronic self-governance at relevant times or small intervals along the way.
- Plan-theoretic interconnections of intention, which induce shared-agency-analogue interconnections.
- Diachronicalized standpoints, which include a conditional, reflexive end in favour of one’s own diachronic self-governance.
Its applications:
- Sartre’s young man (§2, §7). Shuffling between options he judges non-comparable is self-governed at each time but “diminishes the extent to which he is diachronically self-governing”. Resistance to shuffling is supported by the end even if the non-comparability remains.
- Temptation (§6–7). Two elements, a prior resolution against a present evaluation, cannot explain why the resolution speaks for the agent. A third element is needed: the end of diachronic self-governance. It re-shifts the standpoint first and the evaluation second. Sobel’s “premium on steadfastness” and Rabinowicz’s plan-influenced preferences are rejected as brute tidiness. Kelly’s end of redeeming one’s past actions is rejected as not general enough.
- The toxin (§8). Sticking with an intention to drink Kavka’s toxin is excluded because at “plan’s end” one would regret having followed through. In temptation one would regret having given in. The plan’s temporal footprint extends to that later reflection.
- An asymmetry (§6). There is no parallel direct argument for an end of synchronic self-governance. The argument for one is indirect, through the diachronic end.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Diachronic self-governance requires cross-temporal links of intention beyond self-governance at each time | argued from cases (Ulysses, Sartre’s young man) | §§1–2 |
| C2 | The settling function of intention supports the plan-theoretic model | argued, then conceded “not conclusive” against Ferrero’s objection | §3 |
| C3 | Planned temporally extended activity contains intrapersonal analogues of conditions (A)–(E) of shared intention | argued condition by condition; the analogy is “at most only partial”, with no two-way causation and condition (C) only partly met | §5 |
| C4 | An end of one’s own diachronic self-governance, in the standpoint, makes some willpower and resistance to shuffling self-governed | argued by inference to the best explanation; the end is posited, its presence in agents is not shown | §§6–7 |
| C5 | Anticipated later regret, within the plan’s time-frame, separates self-governed willpower from following through on the toxin intention | argued from cases | §8 |
| C6 | Bargaining between time-slices (McClennen, Ainslie) cannot model diachronic self-governance, because there is no reciprocity across time | asserted in two sentences, with reference to an earlier essay | §4 and note 19 |
Concepts
- Shuffling: switching between options judged non-comparable at successive times with no change in evaluation. The term is from Richard Kraut; Bratman’s earlier name was “brute shuffling”.
- Plan-theoretic model: the links characteristic of diachronic self-governance are those of planned temporally extended activity.
- Shared agency model: those links are intrapersonal analogues of the links in shared intention, so that the agent acts “together” with herself.
- Diachronicalized standpoint: a standpoint that includes the conditional, reflexive end of one’s diachronic self-governance.
- Plan’s end: the temporal footprint a plan specifies, commonly including later reflection on earlier follow-through.
- Snowball effect: an initial stage of executing an intention creates new reasons to continue, such as a non-refundable ticket. Here it is extended to side bets made for that purpose.
Connections
The essay builds on Bratman’s own work: the 1987 planning theory (LIT-593), the synchronic account of “Three Theories of Self-Governance” (LIT-596), the Lockean reading of planning agency in “Reflection, Planning, and Temporally Extended Agency” (in Structures of Agency, LIT-590) and Shared Agency (LIT-488). Frankfurt’s “where (if anywhere) the person himself stands” is the starting point (Frankfurt’s “Identification and Wholeheartedness” is cited, which another reader has filed). The essay is set against McClennen’s and Ainslie’s bargaining models (LIT-586 is Ainslie’s), Holton’s account of willpower, and Velleman’s “intellectualist” alternative (LIT-295 is Velleman on action). It mentions narrative unity only to subordinate it to acting together with oneself (compare LIT-595). Korsgaard (LIT-587) gives a normative reading of Bratman’s Lockean claims.
Bearing on the record
- THEORY-029. This essay is first-hand evidence on how Bratman answers the regress, though for the over-time case. His stopping point is structural: a coherent standpoint, made partly of plans and policies, plus a reflexive end. Whether that end can be disowned is the question THEORY-029 asks. The essay allows that the end “may not always be sufficiently important” and can be “trumped” by other ends. That makes it one weight among others, not a guarantor of ownership. That fits THEORY-029’s claim that no attitude secures ownership by its order or kind alone. The essay does not engage the regress directly. THEORY-029’s promote_when names “Bratman’s self-governing policies”, which are in LIT-596, still unread.
- THEORY-040 (manipulation). The model is entirely structural: present standpoints and their cross-temporal links. It has no historical condition. So it is exposed to the manipulation objection as THEORY-040 states it. The essay does not discuss manipulation. Korsgaard (LIT-587) reports that Bratman, asked about a post-hypnotically installed intention, said it would be the agent’s own. That fits a structural view that bites the bullet.
- A THEORY, filed as THEORY-076 (Proposed). Diachronic self-governance is constituted by intrapersonal analogues of shared intention, not by narrative unity or intertemporal bargaining. Sources: this essay; against it, LIT-586 (bargaining) and LIT-595 (narrative). It is Proposed until Shared Agency (LIT-488) is read, since the analogy’s terms come from there.
It carries no instruction for machine-learning practice.
Limitations
- The analogy is partial by Bratman’s own account. There is no two-way causation across time, and the interdependence condition (C) holds only “in some cases”.
- The end of diachronic self-governance is posited by inference to the best explanation. Nothing shows that ordinary agents have it, and it can be outweighed.
- The settling-function argument is conceded to be inconclusive.
- The reply to the bargaining models is two sentences long.
- The draft, not the published text, was read.
Open questions
- Whether the conditional, reflexive end can be specified without presupposing the self-governance it is an end of, or whether it carries the regress THEORY-029 records into the diachronic case.
- Whether empirical work on temptation and precommitment (the behavioral-integration literature, Ainslie) can tell the bargaining model and the acting-together model apart, for instance by whether resolve depends on anticipated later regret.