Other-Regarding Preferences¶
Core Idea¶
Other-regarding preferences is the structural commitment that an agent's evaluation of outcomes is indexed not only on the agent's own payoff but on the payoffs received by one or more other agents. The agent's preference function takes as input the joint outcome vector — what I get and what they get — and returns an ordering that systematically responds to both arguments, not only to the self-referential one. The dependence can run in any of several signed directions: positive (I prefer outcomes that benefit them — altruism, kin-directed sacrifice), negative (I prefer outcomes that harm them — spite, envy), reference-comparative (I prefer outcomes close to a fair allocation — inequity aversion), or reciprocity-conditioned (I prefer outcomes that reward those who cooperated and punish those who defected). The structural fact is not that any particular signed dependence holds across all agents, but that the preference function is generically not a function of own-outcome alone.
Once this commitment is named, the canonical own-payoff-maximising agent becomes the special case — the case where the cross-arguments have zero weight — and the empirical distribution of agents across substrates is not concentrated at that point. What other-regarding preferences is not is a specific preference content: it is not the claim that people are nice, that they care about fairness, or that they reciprocate. It is the structural meta-commitment that the preference function has cross-arguments at all, leaving open what sign and shape those arguments take in any given substrate or population. Formally it is the move from a utility function \(u_i(x_i)\) of own-outcome alone to a function \(u_i(x_i, x_{-i})\) of both own-outcome and others' outcomes, with the structure of the dependence — which agents enter, with what signs, under what conditions — as the empirical object of interest.
How would you explain it like I'm…
Caring About Their Cookies
Their Outcome Counts Too
Payoffs Beyond My Own
Structural Signature¶
the evaluating agent — the joint outcome vector (own and others') — the preference function over that vector — the cross-arguments (others' outcomes as inputs) — the signed, conditional dependence on them — the own-payoff-only special case as a measure-zero point
A setting has other-regarding preferences when each of the following holds:
- An evaluating agent. There is an agent with a preference ordering over outcomes — the focal agent whose evaluation is under analysis.
- A joint outcome vector. Outcomes are distributed across multiple recipients, so each state of the world specifies what the focal agent gets and what one or more other agents get.
- A preference function over the joint vector. The agent's evaluation is a function \(u_i(x_i, x_{-i})\) taking the whole outcome vector as input, not \(u_i(x_i)\) taking own-outcome alone.
- Cross-arguments. Others' outcomes genuinely enter the function: the ordering systematically responds to \(x_{-i}\) and not only to \(x_i\). This is the load-bearing meta-commitment — that the function has cross-arguments at all.
- Signed, conditional dependence. The dependence carries a sign and possibly a condition — positive (altruism), negative (spite, envy), reference-comparative (inequity aversion), or reciprocity-conditioned — and which agents enter, with what weights, under what conditions, is the empirical object, left open by the structure.
- The own-payoff-only special case. The canonical self-interested agent is recovered exactly when all cross-argument weights are zero — one measure-zero point in a preference-function space, not a privileged default.
The signature is structural, not psychological: it commits only to the form \(u_i(x_i, x_{-i})\) with nonzero cross-arguments, leaving the sign, shape, and population distribution of those weights as parameters to be estimated per substrate.
What It Is Not¶
- Not a sub-instance of
preferencethat adds nothing. Preference is the umbrella ordering relation over outcomes; other-regarding preferences is the specific content commitment that the ordering depends on others' outcomes — a function \(u_i(x_i, x_{-i})\), not \(u_i(x_i)\). It specialises preference rather than restating it. - Not
cooperation. Cooperation is a behavioural pattern that can arise without other-regarding preferences (repeated-game folk theorem, pure self-interest under the shadow of the future) and that other-regarding preferences can fail to produce (envious or spiteful agents defect). - Not
reciprocity. Reciprocity is one signed, intention-contingent form of other-regarding preference; this prime is the broader meta-commitment that the function has cross-arguments at all, of any sign. - Not
fairness. Fairness is one possible content (a positive weight toward equal allocations); other-regarding preferences also covers spite, envy, and norm-enforcement, which are equally other-regarding but not fair. - Not a
social_dilemma. A social dilemma is a strategic structure defined by the payoff matrix; this prime is about the preference structure over outcomes, which can be present or absent independent of the game's payoffs. - Common misclassification. Reading "other-regarding" as "altruistic" and treating the prime as a niceness claim. An agent who pays to harm a rival has fully other-regarding preferences; the meta-commitment is to nonzero cross-arguments of any sign, not to a positive weight.
Broad Use¶
The cross-argument structure recurs across substrates because the commitment is mathematical rather than substantively psychological. In behavioural game theory, its canonical home, dictator-game donations to anonymous recipients, ultimatum-game rejections of unfair offers, trust-game investments, public-goods contributions, and third-party costly punishment each surface a specific signed dependence and estimate its parameters, because the own-payoff-only special case fails systematically and predictively. In evolutionary biology, kin selection values relatives' reproductive success at a coefficient given by relatedness, reciprocal altruism and eusocial cooperation instantiate the same structure with fitness as the preference function and a related or grouped agent as the "other," and the cross-argument weights are biological parameters with empirical values. In welfare economics and social choice, any social welfare function built from agent preferences must specify whether and how agents' orderings depend on others' outcomes, and the choice changes the aggregate and the policy recommendations. The pattern recurs in mechanism design (allowing other-regarding components changes truth-revelation properties, optimal bid functions, and welfare bounds), in political philosophy (any theory evaluating distributions across persons presupposes preferences over distributions, with the veil of ignorance a procedural way of eliciting other-regarding components), in cooperative AI and multi-agent systems (utility functions with non-self-only arguments to support cooperation), in organisational behaviour (peer-comparison effects on pay satisfaction), and in charitable giving (anonymous versus identified giving, warm-glow versus purely altruistic motivation). The fit holds anywhere agents are modelled with preferences and outcomes are distributed across multiple recipients.
Clarity¶
Other-regarding preferences clarifies by separating the question of self-interest from the question of preference content. Standard rational-choice modelling tends to conflate the two by assuming own-payoff maximisation is the default and "departures" are anomalies. Once other-regarding preferences is named as a structural option, the own-payoff-only agent is recognised as one specific point in a preference-function space, not as the natural starting point — and the empirical question becomes which point in that space a given agent or population occupies, rather than whether they have "departed" from a privileged baseline.
The clarifying force extends to a recurring source of confusion in interpreting prosocial behaviour. The same observed act — giving money to a stranger — can come from positive other-regarding preferences (I value their welfare), reciprocity-conditioned preferences (I expect they would do the same), reputation effects (I value being seen as generous), or warm-glow effects (I value the act of giving regardless of recipient outcome). Naming the underlying parameter space lets the analyst design experiments that distinguish the signed components rather than treating all prosocial behaviour as a single category. The prime also distinguishes itself carefully from neighbours. It is a content commitment about preference, not an alternative to it: preference is the umbrella ordering relation, and other-regarding preferences names a sub-case specifying what enters the preference function. It is distinct from cooperation, which is a behavioural pattern that can occur without other-regarding preferences (the folk theorem) and that other-regarding preferences can fail to produce (envious agents). It is distinct from reciprocity (one signed, intention-contingent form), from fairness (one possible content), and from a social dilemma (a strategic structure defined by the payoff matrix, not the preference structure). Holding these apart keeps the prime from being mistaken for niceness, for cooperation, or for any single one of its signed instances.
Manages Complexity¶
The prime compresses a sprawling literature of named "anomalies" — ultimatum rejections, dictator giving, public-goods contributions, third-party punishment, trust-game investments, inequity aversion, guilt aversion, intention-contingent reciprocity, identifiable-victim effects, in-group bias in helping — into a small parameter space: which other agents enter the preference function, with what signs, and under what conditions. Competing formal models each propose a specific functional form for the cross-argument weights — inequity-aversion models use two parameters for the disutility of advantageous and disadvantageous inequality, others use three, others a different two — and each can be evaluated against the same paradigm suite. A catalogue of disconnected behavioural findings becomes a single function with a few free parameters whose values the experiments estimate.
The compression is operational because it organises the intervention catalogue as well as the explanatory one. Mechanism designers who want to elicit cooperative behaviour can target the cross-argument weights directly: framing manipulations that activate in-group membership shift in-group/out-group weights, anonymity manipulations reduce reputation-channel weights, and partner-versus-stranger matching shifts reciprocity-conditioned weights. Each intervention is recognisable as a move on a specific parameter of the prime rather than as an ad-hoc behavioural nudge. And because aggregate outcomes depend on the distribution of cross-argument weights across a population rather than on the modal type, the prime directs attention to sorting and matching mechanisms: a modest fraction of conditional cooperators can sustain cooperation if matched with similar types. By naming the parameter space, the prime turns the open-ended problem of "why do people behave prosocially" into a bounded reasoning over a small set of weights and their population distribution.
Abstract Reasoning¶
Other-regarding preferences trains a reasoner to interrogate any multi-agent setting through the structure of the preference function. The reasoner asks: whose outcomes enter the focal agent's evaluation, with what weight, and under what conditions? Because this question references only the abstract roles — preference function, joint outcome vector, cross-arguments, signed dependence — it applies to a laboratory game, an evolving population, a welfare aggregation, or a multi-agent learning system without translation, and the kin-selection coefficient in biology and an altruism parameter in a behavioural model are recognised as structurally the same object, a weight on another agent's outcome in the focal agent's evaluation, with the substrate setting only the empirical value.
Several reusable moves follow. The parameter-identification move designs paradigms that isolate specific cross-argument components — a dictator game isolating the purely altruistic component, an ultimatum game isolating responder inequity aversion, a trust game isolating positive reciprocity expectations, third-party punishment isolating norm enforcement without own-victim status. The aggregate-prediction move treats population behaviour as a function of the distribution of cross-argument weights, so the reasoner predicts from the distribution rather than the modal type, making sorting and matching mechanisms load-bearing. The mechanism-redesign move recognises that optimal mechanisms under own-payoff agents differ systematically from those under inequity-averse or reciprocity-conditioned agents, so ignoring the cross-arguments produces predictable implementation failures. And the welfare-construction move recognises that a social welfare function can be built either by aggregating orderings over distributions or by aggregating own-payoff augmented by other-regarding terms, a choice with substantive consequences for what counts as a Pareto improvement. The same reasoning that lets an experimental economist estimate an inequity-aversion parameter lets an evolutionary biologist read a kin-selection coefficient, because both are reasoning about a weight on another agent's outcome.
Knowledge Transfer¶
The transferable content of the prime is that the underlying object is mathematical — a function \(u_i(x_i, x_{-i})\) depending on both own-outcome and others' outcomes — so it travels cleanly across substrates that use different names for it. Experimentalists in behavioural economics, theorists in evolutionary biology, policy analysts in welfare economics, and engineers in multi-agent AI are all working with the same object, and the transferable diagnostic is identical: whose outcomes enter, with what weight, and under what conditions? Once the cross-argument structure is identified for a substrate, the existing toolkit of game-theoretic equilibrium analysis, evolutionary stability, mechanism-design implementation, and welfare aggregation applies, with the substrate-specific cross-argument weights as the new free parameters.
The transfer is deep because the design move — target the cross-argument weights, not just the own-payoff term — ports across substrates that share the structure. The dictator game makes this concrete: a proposer given an endowment unilaterally decides how much to transfer to an anonymous recipient with no power to accept, retaliate, or reciprocate, and the own-payoff-only prediction is zero, while the empirical modal transfer is roughly 20–30% of the endowment. Which cross-argument structure is posited determines the interpretation — positive altruistic weight predicts transfers scaling with perceived benefit, inequity aversion predicts clustering at the equal-split point, norm-following predicts insensitivity to recipient identity — and variant experiments distinguish these by manipulating the relevant parameter: transfers fall when anonymity to the experimenter is increased (implicating reputation weights), rise when the recipient is identifiable or needy (implicating recipient-specific altruistic weights), or cluster at equal split independent of recipient features (implicating norm-following). The same structural pattern recurs in eusocial-insect colonies where workers forgo reproduction for colony fitness (the cross-argument weight biologically anchored by relatedness), in workplaces where coworkers' wages enter individual job satisfaction (an inequity-aversion signature), and in multi-agent reinforcement learning where the cross-argument weight in the reward shapes whether agents converge to cooperative or competitive equilibria. Because the design move — manipulate framing, anonymity, identifiability, partner matching, or social information to shift specific cross-argument weights — is substrate-neutral, a practitioner who has elicited cooperation in one substrate can target the same parameters in another on first contact, and the strip-the-jargon form ("the preference function depends on what others get, not only on what I get") does load-bearing work across experimental economics, evolutionary biology, mechanism design, welfare economics, multi-agent AI, and charitable-giving research.
Examples¶
Formal/abstract¶
The ultimatum game with an inequity-aversion utility function is the pattern in its most explicit formal dress. The evaluating agent is the responder; the joint outcome vector is the pair \((x_i, x_j)\) — the responder's share and the proposer's share of a fixed pie. The own-payoff-only model writes the responder's utility as \(u_i(x_i) = x_i\) and predicts acceptance of any positive offer, because any money beats none. The empirical fact — responders routinely reject offers below roughly 20–30% of the pie, choosing zero over a small but unfair amount — falsifies that model and forces a cross-argument. The Fehr-Schmidt inequity-aversion form makes the dependence precise: \(u_i(x_i, x_j) = x_i - \alpha \max(x_j - x_i, 0) - \beta \max(x_i - x_j, 0)\), where \(\alpha\) weights disadvantageous inequality (the disutility of getting less than the other) and \(\beta\) weights advantageous inequality. This is the signed, conditional dependence made parametric: the responder's evaluation systematically responds to \(x_j\), with a kink at the equal split. The own-payoff-only special case is recovered exactly at \(\alpha = \beta = 0\) — one point in the parameter space, not the default. The model's predictive content is sharp and testable: a responder with \(\alpha\) large enough rejects any sufficiently unfair offer because the inequality disutility exceeds the monetary gain, which locates the rejection threshold as a function of \(\alpha\) and lets the parameter be estimated from the observed minimum acceptable offer.
Mapped back: The responder is the evaluating agent, \((x_i, x_j)\) is the joint outcome vector, the Fehr-Schmidt function is the preference function with explicit cross-arguments, \(\alpha\) and \(\beta\) are the signed conditional weights, and \(\alpha = \beta = 0\) is the measure-zero own-payoff special case — other-regarding preferences with the cross-argument structure fully formalized.
Applied/industry¶
Kin selection in evolutionary biology and reward shaping in multi-agent reinforcement learning instantiate the identical cross-argument structure in unrelated substrates. In a eusocial insect colony, a sterile worker forgoes its own reproduction to raise the queen's offspring — behaviour that own-fitness maximisation cannot explain. Hamilton's rule supplies the cross-argument exactly: the worker's effective preference function values a relative's reproductive success at a weight equal to the coefficient of relatedness \(r\), so an altruistic act is favoured when \(rB > C\) (relatedness times benefit to the recipient exceeds cost to the actor). The "other" whose outcome enters the function is the related individual, and the weight is biologically anchored rather than psychological — but it is structurally the same object as the economist's altruism parameter: a nonzero weight on another agent's payoff. In multi-agent reinforcement learning, a designer who wants cooperative rather than competitive convergence writes each agent's reward as its own task reward plus a weighted term in other agents' rewards; the sign and size of that cross-argument weight determine whether agents learn to share, ignore, or sabotage one another, and tuning it is the direct engineering analogue of the economist's "target the cross-argument weights" move. The same design lever recurs in workplace compensation, where a coworker's wage entering an employee's job satisfaction (an inequity-aversion signature) means pay-transparency and relative-pay structure shift behaviour through the cross-argument rather than the own-wage term. A biologist reading a relatedness coefficient, an ML engineer setting a cooperation weight, and a compensation designer managing relative-pay effects are all reasoning about a weight on another agent's outcome.
Mapped back: The worker, the RL agent, and the employee are evaluating agents; relatedness \(r\), the reward-sharing coefficient, and the inequity-aversion weight are the same signed cross-argument on another agent's payoff; the colony's altruism, the agents' cooperation, and the relative-pay effect are its consequences — other-regarding preferences as one mathematical object across biology, AI, and organisational behaviour.
Structural Tensions¶
T1 — Structural Form versus Preference Content (scopal). The prime commits only to the form \(u_i(x_i, x_{-i})\) having nonzero cross-arguments; it does not commit to any content (niceness, fairness, reciprocity). The characteristic failure is collapsing the meta-commitment into one of its signed instances — reading "other-regarding" as "altruistic" and missing spite, envy, or norm-enforcement, which are equally other-regarding. The diagnostic is to ask what sign and shape the cross-argument takes here rather than assuming a positive weight: an agent who pays to harm a rival has fully other-regarding preferences, and treating the prime as a fairness claim mis-specifies half its range.
T2 — Same Behaviour versus Distinct Cross-Arguments (measurement). A single observed act — giving to a stranger — is consistent with several distinct underlying weights: pure altruism, reciprocity expectation, reputation concern, or warm glow. The failure is inferring a specific preference structure from behaviour that underdetermines it, treating all prosocial acts as one category. The diagnostic is to design paradigms that isolate components — anonymity manipulations strip reputation weights, identifiability manipulations probe recipient-specific altruism, partner-versus-stranger matching isolates reciprocity — so the cross-argument is identified rather than assumed; behaviour alone cannot distinguish the signed channels that produced it.
T3 — Modal Type versus Population Distribution (scalar / local-global). Aggregate outcomes depend on the distribution of cross-argument weights across a population, not on the modal agent — a minority of conditional cooperators can sustain or collapse cooperation depending on matching. The failure is predicting group behaviour from the representative agent, missing that a small fraction of a different type drives the aggregate. The diagnostic is to ask how types are distributed and sorted: the same population produces cooperation or unravelling depending on whether conditional cooperators are matched with their like, so the modal preference is the wrong unit and sorting mechanisms are load-bearing.
T4 — Other-Regarding Preferences versus Cooperation (scopal). The prime is a preference structure, distinct from cooperation, a behavioural pattern: cooperation can arise without other-regarding preferences (repeated-game folk theorem, pure self-interest under future shadow) and other-regarding preferences can fail to produce it (envious or spiteful agents defect). The failure is reading observed cooperation as evidence of prosocial preferences, or expecting prosocial preferences to guarantee cooperation. The diagnostic is to separate the payoff-transforming preference from the strategic structure: ask whether cooperation here is sustained by the preference function or by the equilibrium incentives of repetition, since the two have different interventions.
T5 — Own-Payoff Baseline versus Measure-Zero Point (sign/direction). The canonical self-interested agent is recovered only at the single point where all cross-argument weights are zero — a measure-zero point, not a privileged default from which other behaviour is an "anomaly." The failure is the asymmetry of treating own-payoff maximisation as the natural baseline and other-regarding behaviour as deviation requiring special explanation, which mis-frames the empirical question. The diagnostic is to ask which point in the parameter space a population actually occupies, refusing to privilege the zero-weight point: framing departures as anomalies imports a baseline the structure does not support and biases the model toward the wrong null.
T6 — Self-Interested Mechanism versus Other-Regarding Agents (coupling). Mechanisms optimal under own-payoff agents differ systematically from those optimal under inequity-averse or reciprocity-conditioned ones — truth-revelation, optimal bids, and welfare bounds all shift when cross-arguments are present. The failure is deploying a mechanism designed for self-interested agents onto a population with cross-arguments and getting predictable implementation failure (an auction that assumes no spite, a contract that assumes no fairness concern). The diagnostic is to ask whether the agents' preference function the mechanism assumes matches the population's actual cross-argument structure; a mismatch between assumed and actual weights is a designed-in failure, not a behavioural surprise.
Structural–Framed Character¶
Other-regarding preferences sits just onto the framed side of the structural–framed spectrum — the balanced-hybrid case where its framed label and aggregate of 0.5 reflect every diagnostic reading exactly mid. There is a clean structural core: a move from a utility function of own-outcome alone, \(u_i(x_i)\), to one of the joint outcome vector, \(u_i(x_i, x_{-i})\), with signed, conditional cross-arguments. But the prime is stated in the inherited vocabulary of preference and agency from economics and philosophy, and that lensing pulls each criterion halfway toward framed.
The formal skeleton is real and partially formalized, which keeps it from sliding all the way to framed; the inherited frame is real too, which keeps it off the structural side. Vocabulary partly travels: the cross-argument-utility shape recurs across behavioural game theory, kin selection in evolutionary biology, welfare economics, political philosophy, and multi-agent AI, but the terms "preference," "agent," "utility," "altruism," "spite," and "inequity aversion" come along and carry economic-philosophical content (vocab_travels 0.5). It carries a mild evaluative load: the signed directions are named with morally-tinged labels (altruism, spite, envy, fairness), importing value content even though the structural claim is only that cross-arguments are non-zero (evaluative_weight 0.5). Its origin is behavioural-economic and game-theoretic rather than tied to a single named institution, sitting between formal and human-practice origins (institutional_origin 0.5). It is partly human-practice-bound: kin-selection instances reach into biology where no deliberating agent is present, but the home cases presuppose preference-bearing agents evaluating outcomes (human_practice_bound 0.5). And invoking it imports a modest interpretive frame — read this behaviour as a preference function with cross-arguments rather than as bare correlation — while still recognising a genuine structural feature, dependence on others' payoffs, that is really there (import_vs_recognize 0.5). Because every criterion lands at the midpoint, the prime is a true hybrid that the rubric places just on the framed side of center: a clean cross-argument formalism wrapped in the evaluative, agent-presupposing vocabulary of preference theory.
Substrate Independence¶
Other-regarding preferences is a moderately substrate-independent prime — composite 3 / 5 on the substrate-independence scale. Its domain breadth is real but bounded: the cross-argument-utility structure recurs across behavioural game theory (dictator, ultimatum, trust, public-goods, and third-party-punishment games), evolutionary biology (kin selection, reciprocal altruism, eusocial cooperation), welfare economics and social choice, mechanism design, political philosophy (the veil of ignorance), cooperative AI and multi-agent systems, organisational behaviour, and charitable giving — genuinely distinct fields, but nearly all anchored in the preference substrate, with the one reach beyond deliberating agents (kin selection, where fitness plays the role of the preference function) being the partial exception that keeps it from full breadth. Its structural abstraction is mid because the commitment is mathematical — an agent's utility depends on arguments other than its own payoff, with a signed weight — yet it is stated in the agent-presupposing vocabulary of preference, utility, altruism, spite, and inequity aversion, so the formalism is portable but cannot be cleanly separated from preference-bearing agents. Its transfer evidence earns a 4: the cross-argument weights are concrete, estimated parameters in each home substrate — relatedness coefficients in kin selection, inequity-aversion parameters fit to ultimatum-game data, social-welfare weightings in economics — and the same own-payoff-only special case fails systematically and predictively across them, so the formal structure carries with measured values rather than by loose analogy. The composite sits at 3 because the preference-substrate ceiling pins breadth and abstraction even as the cross-argument formalism transfers with empirical parameters.
- Composite substrate independence — 3 / 5
- Domain breadth — 3 / 5
- Structural abstraction — 3 / 5
- Transfer evidence — 4 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Other-Regarding Preferences is a kind of Preference
The file + dedup flag: a CONTENT SPECIALISATION of preference — the move from u_i(x_i) to u_i(x_i, x_{-i}) with nonzero cross-arguments. preference is the umbrella; the own-payoff-only agent is recovered at zero cross-weights (a measure-zero point). NOT a reparent — preference is the genus, candidate is the more-SPECIFIC child.
Path to root: Other-Regarding Preferences → Preference
Neighborhood in Abstraction Space¶
Other-Regarding Preferences sits in a sparse region of abstraction space (79th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.
Family — Strategic Interaction & Mechanism Design (12 primes)
Nearest neighbors
- Preference — 0.73
- Joint vs. Separate Evaluation — 0.72
- Preference Heterogeneity and Conflict — 0.69
- Social Choice — 0.69
- Revealed Preference — 0.69
Computed from structural-signature embeddings · 2026-06-14
Not to Be Confused With¶
The decisive confusion — and one this prime carries a dedup flag against — is with preference itself, its nearest neighbour at very high similarity. The two are genuinely close, and the relation must be stated precisely rather than denied. Preference is the umbrella ordering relation: an agent's ranking over outcomes, \(u_i(\cdot)\), with no commitment about what enters the ranking. Other-regarding preferences is a content specialisation of that umbrella — the specific commitment that the ranking depends on others' outcomes as well as one's own, the move from \(u_i(x_i)\) to \(u_i(x_i, x_{-i})\). So the candidate is not a rival to preference but a sub-case that fixes a structural feature preference leaves open; the own-payoff-only agent is recovered exactly when all cross-argument weights are zero, one measure-zero point in the preference-function space. The case for keeping it first-class (pending the parent/child resolution the dedup flag defers) is that the cross-argument structure — which agents enter, with what signs, under what conditions — is a distinct empirical object with its own paradigms (dictator, ultimatum, trust games), its own design lever (target the cross-argument weights), and its own failure modes (deploying a self-interested mechanism onto a population with nonzero weights), none of which the bare umbrella of preference articulates. If Phase C folds it into preference, that content becomes a named region of preference-space; if it stays separate, it is the specialisation that makes the cross-arguments the object of study. Either way the discriminating fact is that preference is silent on whether the function has cross-arguments at all, and this prime's entire content is that it does.
A second genuine confusion is with cooperation, because other-regarding preferences are so often invoked to explain cooperative behaviour. But cooperation is a behavioural pattern, while other-regarding preferences is a preference structure, and the two are doubly dissociable. Cooperation can arise with purely self-interested preferences — the repeated-game folk theorem shows that own-payoff maximisers cooperate under a sufficient shadow of the future, sustained by equilibrium incentives rather than by caring about others. And other-regarding preferences can fail to produce cooperation — envious or spiteful agents, whose cross-arguments are negative, defect or sabotage. The failure is reading observed cooperation as evidence of prosocial preferences, or expecting prosocial preferences to guarantee cooperation. The discriminating move is to ask whether cooperation here is sustained by the preference function (agents value others' outcomes) or by the strategic structure (repetition makes defection unprofitable), because the two have entirely different interventions — change the payoffs versus change the preferences.
A third confusion is with reciprocity and fairness, each of which is one instance of the broader meta-commitment. Reciprocity is the signed, intention-contingent form — reward those who cooperated, punish those who defected — and fairness is the reference-comparative form, a positive weight toward equal allocations. Both are genuine other-regarding preferences, but neither exhausts the category, which also contains spite (negative weight on a rival's payoff), envy, and pure altruism. Collapsing the prime into reciprocity or fairness is the most common mis-specification because it reads "other-regarding" as "positive and conditional," missing exactly half the range — the agents who pay to harm others, who are no less other-regarding than the ones who pay to help. The diagnostic is to ask what sign and shape the cross-argument takes here rather than assuming the prosocial instance.
These distinctions matter because they fix the empirical and design questions. Treating the prime as preference loses the cross-argument object; treating it as cooperation confuses a behavioural outcome with a preference cause and prescribes the wrong lever; treating it as fairness or reciprocity mis-specifies the sign of the weight — whereas the prime directs the analyst to identify whose outcomes enter, with what weight, under what conditions, the parameter space that mechanism design and welfare aggregation actually act on.
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.