NOTE-185: The moral psychology of obligation

Contribution

The paper gives Darwall-style second-personal accounts of obligation a psychological and evolutionary foundation. The sense of obligation is located in the formation of a shared agent “we”, first joint (dyadic collaboration under joint commitment) and then collective (identification with a cultural group). That “we” both directs and self-regulates the joint activity. The paper collects comparative and developmental findings that 3-year-olds, unlike chimpanzees, act as if bound to collaborative partners (and later to in-group members) in ways they are not bound to others. It then classifies obligation as a “self-conscious motivation”, analogous to guilt.

Key insight

Obligation is agency at the plural level turned back on the individual. When “we” pursue a goal together we also regulate one another’s roles. Because I helped author that “we” and my cooperative identity depends on it, its claims on me feel legitimate rather than merely imposed: a “we > me” self-regulation. Individual instrumental pressure (“to get that apple, I must climb that tree”) becomes cooperative, normative pressure, and guilt is its after-the-fact form.

Assumptions

Premises and authorities:

• The phenomenon to explain is the sense of obligation, a motivation with two features taken from moral philosophy: (1) Special Force, peremptory and coercive (“obligation is a stick”), and (2) Special Social Structure, tied to agreements and promises, with protest and apology on breach (Intro).
• The evolutionary scenario. A feeding-ecology change forced obligate collaboration, which was then selected through partner choice and partner control (protest before exclusion) (§2.1.1). Modern humans (about 150,000 years ago) lived in tribal groups with in-group strangers, which selected for conformity and third-party norm enforcement (§2.2.1). The chronology is declared unimportant and is not defended.
• Ontogeny as maturationally guided learning (Tomasello 2019). Joint intentionality emerges around 1–3 years; group-mindedness emerges around 3, with “a strong evolutionary/maturational component” (§2.2.3).
• Chimpanzee contrast as the null. Behaviours that 3-year-olds show and chimpanzees do not in matched paradigms are taken to be species-unique.
• Interpretive premise. Normative language (“One must pull on the rope”), leave-taking and guilt-driven reparation are read as evidence of felt obligation rather than of enhanced prosocial motivation. §2.1.3 argues this against the “enhanced sympathy” alternative.
• Scope, conceded. The ontogenetic account “cannot be considered to be a universally valid account of the human species so long as the only children for whom we have rich data across ontogeny are from WEIRD cultures” (§2.2.2.4).

Key results

The article reports findings from cited studies; it runs no new experiment.

• Joint commitment effects at about age 3 (§2.1.2.1). With a joint commitment, 3-year-olds resist a bribe to defect more than when playing side by side (Kachel & Tomasello 2019). They wait for, repair for, and take over the role of a partner (Gräfenhain et al. 2013). They protest a partner’s knowing failure but not an ignorant one, in generic normative language (Kachel et al. 2018). They “take leave” when lured away, which 2-year-olds do not (Gräfenhain et al. 2009).
• Sharing (§2.1.2.2). 3-year-olds offer peers about one in four items in dictator games (Ibbotson 2014), but split collaborative spoils about equally. The lucky child equalises about “three-quarters of the time” after collaboration and almost never without it (Hamann et al. 2011). Chimpanzees share equally rarely inside and outside collaboration (Melis et al. 2006, 2011). Children reject generous-but-unequal splits after collaboration (Corbit et al. 2017). They accept unequal outcomes from an agreed fair procedure (Grocke et al. 2015).
• Group-minded obligation at 3–5 years (§2.2.2). The findings are minimal-group loyalty and in-group favouritism, conformity overriding one’s own preference, normative protest at norm violations, and group-relative enforcement of conventional but not moral norms (Schmidt et al. 2012). 5-year-olds enforce self-created norms only on those who agreed to them (Schmidt et al. 2016b). Children expect in-group members not to harm one another, with no in-group/out-group difference for helping (Rhodes 2012), which is read as specific to obligation.
• Against reciprocity and social exchange (§3). Tit-for-tat yields gratitude, not obligation: children who fail to reciprocate do not apologise (Warneken & Tomasello 2013). Social exchange yields obligation only when it is construed as a “we”-regulated agreement.
• Definition (§3). The sense of obligation is “the internalized social/rational pressure from a joint or collective self-regulating agent ‘we’ … to do what ‘we’ expect me to”. Guilt (“we” think of me) is set against shame (“they” think of me).
• Response, R4 (the main concession). Joint commitments and their obligations are graded by explicitness and by how much each party’s contribution matters. This is offered as a gradualist bridge from Bratman-style to Gilbert-style collaboration. Tomasello agrees with Darwall “on one level” that the “we” is idealised (“anyone who would be one of us”) but denies that obligation is anything beyond a psychological phenomenon.
• Response, R5, agenda. Integrate parent-child relations and friendship; map the “many faces” of obligation across roles and societies; relate participatory and third-party routes to group understanding.

Claims

id	claim	strength	support
C1	3-year-olds, unlike chimpanzees, treat collaborative partners with commitment, fairness and normative protest that they do not show to non-partners	moderate	synthesis of cited controlled experiments (§§2.1.2.1–2.1.2.2), largely from the author’s lab; WEIRD samples; the chimpanzee comparisons are matched paradigms
C2	These behaviours reflect a felt obligation rather than enhanced sympathy or prosociality	weak–moderate	informal argument from the pattern: equality not generosity, procedural fairness, normative language, leave-taking (§2.1.3)
C3	Joint commitment (not collaboration as such) is the earliest source of felt obligation	moderate for the behavioural contrast, weak for “source”	the commitment vs side-by-side contrasts (§2.1.2.1); Bratman (e62) shows the causal role is not isolated from the “web of assurances”
C4	Obligation to “anyone who would be one of us” arises from identification with a cultural collective agent, from about age 3	weak	in-group/out-group findings (§2.2.2) plus an evolutionary narrative (§2.2.1); the maturational component is asserted
C5	Obligation is a self-conscious motivation, analogous to guilt, whose force is a threat to cooperative identity	assertion	classificatory proposal (§4), borrowing Korsgaard 1996
C6	Tit-for-tat reciprocity cannot generate obligation	weak	one developmental observation (no apology for non-reciprocation) plus argument; the author conceded in 2016 that reciprocity’s tie to obligation is culturally contingent (note 4)
C7	This is “the most comprehensive account to date of the underlying psychology of the human sense of obligation” (§4)	assertion	no comparison with rival accounts beyond Hume, Darwall, reciprocity and social exchange
C8	A naturalistic, graded account gets from Bratman-style to Gilbert-style collaboration without a “magical leap” (R4)	weak	one stag-hunt study (Siposova et al. 2018) and one implicit-vs-explicit study (Kachel & Tomasello 2019); Pettit’s dilemma (e81) is answered by assertion of gradualism, not by an account of how normative concepts arise

Concepts

• Shared intentionality: the species-unique capacity to form shared agencies. It has two steps: joint intentionality (dyadic “we”, second-personal morality) and collective intentionality (cultural “we”, “objective” morality) (§2).
• Joint commitment: an agreement (explicit, or in R4 also implicit via common ground) that constitutes a normatively structured “we” and gives each party standing to protest (§2.1.3). Used without definition; Gilbert (e71) supplies one.
• We > me self-regulation: the joint or collective agent’s standards regulating the individual, felt as legitimate because self-authored (§§2.1.3, 3).
• Second-personal responsibility: the early, partner-directed form of obligation, reserved for dyads. “Obligation period” requires the collective step (§2.1.3).
• Cooperative / moral identity: how “we” evaluate “me”. Losing it is the threat that gives obligation its force (§§2.1.3, 3, 4).
• Role ideals: impartial performance standards for a role in collaboration, “the first social shared normative standards” (§2.1.3).
• Partner choice vs partner control: selecting partners vs reforming them through protest. Tomasello maps them onto Hume and Darwall respectively (§2.1.1).

Connections

Account of agency held. The account presupposes agency at two levels. Individual agency is goal-directed instrumental rationality plus executive self-regulation, which Tomasello treats as general to primates (§3). Plural agency is a joint or collective agent “we” that directs and self-regulates collaboration, taken from Bratman, Gilbert and Searle. Obligation is what individual agency becomes when it is regulated by a plural agent one has co-constituted. Second-personal agency, Darwall’s standing to make and accept claims, is presented as an output of collaboration rather than a precondition. The agency tag is justified. Plural (shared) agency and self-regulation are the paper’s central explanatory notions, and they sit exactly under the blurb’s “collectives”. Ethics and cognition are also justified.

Link to LIT-117 (Enoch, “Agency, Shmagency”). Tomasello’s account of obligation’s force rests explicitly on Korsgaard 1996, and Korsgaard is a principal target of Enoch’s shmagency objection. Read as a normative account, Tomasello’s view is exposed to the analogue of that objection: why be a member of the “we”, rather than a “shm-member” who does not care about their cooperative identity? The response (R4) sidesteps this by treating obligation as purely psychological. Darwall (e68) is the commentator who presses the gap from the other side: the “we” must be ideal, not actual. This is a reading-level connection. The paper does not cite Enoch.

Among the commentaries, Bratman (e62) and Pettit (e81) give the strongest philosophical objections: the joint-commitment element does the work and is separable from collaboration, and a dilemma lands between a normatively austere Bratman model and a normatively loaded Gilbert model. Gilbert (e71) gives the missing definition of joint commitment. Darwall (e68) argues that obligation is irreducibly normative. Among the psychologists, Rhodes (e84, “outside-in”) and Thompson and Carpendale & Lewis (parent-child) offer rival ontogenies; Killen & Dahl argue that morality is not conformity. None of the works named here is held in this record.

Bearing on the record

It supports no THEORY document in this record and contradicts none. With LIT-117 it could anchor a future theory note on whether normativity can be grounded in (plural) agency. Tomasello is the naturalist, “sense-of” side; Enoch and Darwall mark why that does not settle the normative question.

For ML practice it carries nothing actionable. The dossier’s suggestion that it bears on “whether obligation could arise in artificial collaborators” is the reader’s extrapolation. The paper’s criteria are useful for framing such a question (joint commitment, protest and leave-taking, equality over generosity, and guilt as “we”-evaluation), but the paper makes no claim about machines. Nothing belongs in the Anthology.

Note of 2026-10-16: THEORY-495 now cites LIT-125 as a source.

Limitations

• The article’s evidence is a synthesis. It reports no new data, does not weigh effect sizes, and relies largely on the author’s lab’s paradigms and interpretations.
• WEIRD sampling, conceded in §2.2.2.4. The ontogenetic story is falsifiable by pre-age-3 cultural differences, and those data largely do not exist.
• The step from behaviour to felt obligation is interpretive (§2.1.3). The alternative readings (enhanced prosociality, conformity to adult framing, learned scripts) are argued against rather than excluded experimentally.
• The evolutionary narrative is untested and its chronology is declared unimportant (§2.1.1).
• The account explains the sense of obligation, not obligation. The response concedes that “no explanation of obligation, scientific or otherwise, can tell us what we ought to do” (R4). The abstract’s phrase “a psychological foundation for these unique features” should be read accordingly.
• The in-group scoping of obligation (“Do we have obligations to invading Martians?”, §2.2.3) is contested by Killen & Dahl and by Darwall. By the paper’s own account it is descriptive of an ancestral psychology, not a normative claim.
• My reading covered 4 of the 32 commentaries in full. The characterisations of the other 28 rest on their abstracts and on how R1–R3 summarise them.

Open questions

• Does obligation arise from collaboration or from the web of assurances (Bratman)? This could be settled by an experiment that crosses assurances with joint activity, including assurances without any joint activity.
• Is there a pre-age-3 cultural difference in partner-directed obligation? The account predicts none (§2.2.2.4).
• Would 2–3-year-olds attribute special obligations to parents (Marshall e79)? Tomasello predicts not (R3.1).
• How do the participatory (“inside-out”) and third-party (“outside-in”, Rhodes e84) routes to understanding groups relate? The author calls this “a challenging, but ultimately empirically tractable, question” (R5).
• Can a graded naturalistic story explain how normative concepts arise, not only normative behaviour (Pettit e81)?

Corrections to the seeded skim

• Citation confirmed: Behavioral and Brain Sciences 43, e56: 1–58, doi 10.1017/S0140525X19001742. The dossier’s “e56 … unverified” can be dropped. Target article accepted 19 May 2019, manuscript online 28 May 2019, commentaries accepted 20 September 2019. The published “Cite this article” line gives the year as 2020.
• The dossier’s access note is out of date: the full treatment is openly available as author copies at eva.mpg.de and sites.duke.edu (found by web search). OpenAlex’s “closed / no repository copy” is wrong.
• The dossier asks which philosophical analyses are adopted. They are Darwall’s second-personal standpoint (directed obligation vs “obligation period”, which map onto the account’s two steps), Strawson’s reactive attitudes, Hume’s artificial virtues as the foil, Bratman, Gilbert and Searle on shared intentionality, Nagel (1970) on self-other equivalence, and Korsgaard (1996): “an obligation always takes the form of a reaction against the threat of a loss of identity” (§2.2.3). The Korsgaard link matters for this record; see Connections.
• The abstract’s phrasing “voluntarily place our fate in one another’s hands” understates a distinction the body depends on and Bratman presses. Collaboration alone is not claimed to yield obligation: the body’s own control condition (the child “just playing with the adult spontaneously, in the absence of a joint commitment”, §2.1.2.1) shows no leave-taking. The load-bearing ingredient is joint commitment, not joint intentional collaboration as such (Bratman e62). In the response (R4) Tomasello recasts the account as graded: implicit common-ground commitments yield weaker obligation than explicit ones (Siposova et al. 2018; Kachel & Tomasello 2019).
• The dossier’s scope worry (“empirical cognitive science alone”) is half right. The evidence is empirical developmental and comparative psychology, but the target and the treatment are philosophical: four of the commentators are the philosophers whose accounts it uses, and R4 “Whence normativity?” is an explicit is/ought exchange. It belongs in the record.
• Agency tag: justified; see Connections.