Reputation¶
Core Idea¶
Reputation is the aggregated, publicly available record of an agent's past behavior that propagates beyond the parties to the original interactions and governs how third parties treat the agent in future ones, a structure the economics of repeated interaction first made precise when Klein and Leffler (1981) modeled how the present value of a future reputation premium disciplines current quality. [1] Its defining structure is information flow across two dimensions at once: across time, because conduct in the past constrains treatment in the future, and across observers, because behavior toward A becomes a signal that B, C, and D act upon. The agent's present incentives are therefore shaped not by the party currently in front of them but by the shadow their conduct casts on a wide and largely anonymous future audience. Reputation thereby converts a series of nominally one-shot encounters into something with the strategic texture of a repeated game, the mechanism Kreps and Wilson (1982) formalized when they showed that even a small probability of being a "tough" or honest type can sustain cooperative play in finitely repeated games. [2]
What makes reputation a distinct structural primitive rather than a loose synonym for "good name" is that it is a third-party-propagated stock: it lives in the minds and records of observers who were never party to the original interaction, it accumulates and decays over time, and it can be consulted by a stranger who must decide how to treat the agent before any direct interaction has occurred. The prime names the general shape — observed conduct, transmitted across an audience, stored, and fed back as altered treatment — wherever that loop appears, whether the agents are firms, nations, species, web pages, or scientists.
How would you explain it like I'm…
What People Say About You
Track Record People Share
Shared Record of Past Behavior
Structural Signature¶
Reputation encodes a structural pattern: observed conduct → third-party transmission → aggregated standing → altered future treatment → updated incentives. It separates the act (a single observable behavior), the channel (the observers and records that carry it forward), the stock (the standing that accumulates in others' memories), and the feedback (the way that stock changes how the agent is treated and therefore what the agent does next). The loop closes when anticipated future treatment reaches back to discipline present conduct, the self-enforcing structure Nowak and Sigmund (2005) identified as the engine of indirect reciprocity. [3]
Recurring features:
- Aggregated public record of an agent's past behavior
- Standing that propagates beyond the original interacting parties
- Track record that strangers consult before any direct dealing
- Behavioral history compressed into a transferable token
- Shadow of the future cast onto present incentives
- Asset that is slow to build and fast to destroy
- Third-party information stock that sustains cooperation without enforcement
The structural insight is robust across substrates: a firm guarding brand equity, a nation cultivating a reputation for resolve, a cooperating organism accruing "good standing," a seller accumulating eBay feedback, and a web page gathering inbound links all exhibit the same loop in which past behavior, made observable to a future audience, becomes a stock that governs subsequent treatment, a convergence Resnick, Zeckhauser, Friedman, and Kuwabara (2000) documented when they showed that online reputation systems re-implement the indirect-reciprocity logic biologists had independently described. [4]
What It Is Not¶
Reputation is not a private belief held by a single party. A buyer's individual confidence in a particular seller is a dyadic assessment; reputation is the population-level stock that exists even for observers who hold no belief yet because they have never interacted with the agent. The defining feature is that the standing is consultable by a stranger — it is information held in common, not a relationship held between two parties.
Reputation is not the same as actual quality or character. It is an inference about underlying type or behavior, and the inference can be wrong, stale, or manipulated. An agent can have a sterling reputation built on a record that no longer reflects current conduct (reputation lag), or a poor reputation generated by an unlucky run of observable outcomes that does not track true type. The prime names the propagated record and its effects, not the ground truth the record is supposed to summarize.
Reputation is not intrinsically good or deserved. A reputation for ruthlessness, for unreliability, or for cheating is still a reputation: the structure is symmetric with respect to valence. The loop sustains cooperation only when the channel reliably transmits the relevant behavior and the audience rewards good standing; the same loop can entrench a bad name, propagate slander, or lock an agent out of opportunity on the basis of a single damaging episode. Treating reputation as a synonym for "good reputation" smuggles in a value judgment the structure does not make.
Finally, reputation is not a single act of communication. People sometimes use "reputation" loosely for any attempt to manage how one is seen, but a one-off message or display is a signal; reputation is the accumulated, third-party-propagated history against which such signals are read, and which gives them their long-run stakes. The prime is the stock and its dynamics, not the momentary act of broadcasting.
Broad Use¶
Economics & markets: Brand reputation lets buyers infer unobservable quality before purchase; firms invest in costly-to-build, easy-to-destroy reputational capital precisely because the anticipated stream of repeat business disciplines present quality, the quality-assurance role Shapiro (1983) modeled by showing that a reputation premium can support prices above marginal cost as a bond against cheating. [5] Warranties, return policies, and certifications are instruments for transmitting or substituting for reputation.
Evolutionary biology: Indirect reciprocity sustains cooperation among non-kin when helping behavior earns the helper "good standing" that third parties later reward; the cooperation is carried not by direct payback from the recipient but by the propagated record itself, a mechanism whose stability under image-scoring dynamics Nowak and Sigmund (1998) demonstrated in their foundational simulations. [6]
Computer science & distributed systems: Reputation scores aggregate strangers' prior experiences so that parties with no shared history can transact — eBay feedback, web-of-trust key endorsement, and PageRank-style link authority all compute a standing from distributed observations, with PageRank treating an inbound hyperlink as a transferable endorsement of a page's authority, as Page, Brin, Motwani, and Winograd (1999) formalized. [7] Adversarial concerns — Sybil attacks, collusive rating, whitewashing through identity change — are reputation-system attacks on the channel and the stock.
International relations: A state's reputation for resolve, or for honoring treaties and alliance commitments, shapes how adversaries and allies respond to its current moves; the credibility of deterrence and of promises rests on a propagated track record, the dynamic Schelling (1960) made central to his theory of commitment and bargaining. [8]
Academia & open source (non-obvious): Citation counts, h-indices, and public commit histories function as portable reputation that allocates attention, hiring, funding, and trust across institutions the agent has never directly dealt with — the standing travels with the record rather than with the relationship, an attention-allocation effect Merton (1968) named the "Matthew effect" when he traced how prior eminence accrues disproportionate future recognition. [9]
Clarity¶
Naming reputation as a distinct prime separates three things that ordinary language runs together: the standing of an agent (a stock, held in many observers' minds and records), a single act of signaling (a flow, emitted to convey type), and the dyadic stance of trust (one party's willingness to be vulnerable to another). The distinction lets practitioners say something that is otherwise hard to express precisely: that an agent can be trusted by someone who has no direct experience of them at all, purely on transmitted reputation, and conversely that reputation can be damaged by behavior toward someone other than the party now deciding how to treat the agent. The harm to standing propagates through the audience, not through the injured party's own response. This separation also clarifies whose information reputation is: not the agent's (who may dispute it), not any single counterparty's (who holds only a fragment), but the aggregate carried by the observing population.
The clarity is practically load-bearing in mechanism design. Once one sees reputation as a stock fed by observable conduct and consulted by future strangers, the design questions become sharp: which behaviors are made observable, to whom, through what channel, with what memory and what decay, and how is the stock protected against forged or strategic inputs? Dellarocas (2003) argued that this reframing is exactly what lets online platforms engineer trust among strangers at scale — by deliberately constructing the observation-transmission-aggregation channel that physical communities once supplied through gossip. [10]
Manages Complexity¶
Reputation compresses a long, dispersed behavioral history into a compact, transferable token, sparing each new counterparty the prohibitive cost of monitoring the agent from scratch or learning its type through costly trial. Instead of every newcomer re-running the experiment of dealing with an unknown agent, the population maintains a shared summary that any member can consult cheaply. This lets large populations of strangers coordinate as if they had a common memory, even though no individual holds the whole history. Milgrom, North, and Weingast (1990) showed how exactly this kind of institutionalized reputation tracking — the medieval law-merchant courts that recorded and transmitted traders' conduct — let merchants who would never meet again trade honestly across the fairs of Europe, substituting a reputation channel for the impossible task of bilateral enforcement among thousands of strangers. [11]
The complexity-management payoff is also a complexity relocation: the cost does not vanish but moves into the channel and the aggregator. Maintaining a trustworthy reputation system — verifying inputs, resisting manipulation, deciding how to weight old versus recent conduct — is itself demanding, which is why reputation institutions (rating agencies, credit bureaus, platform feedback systems, peer review) are durable, contested, and frequently captured.
Abstract Reasoning¶
Recognizing reputation as a prime enables folk-theorem reasoning in settings that look one-shot but are not. Because defection is observed and propagated to future counterparties, the threat of reputational loss can sustain cooperation without any external enforcer, as in the community-enforcement equilibria Kandori (1992) proved can support cooperation in large populations whose members rarely meet the same partner twice. [12] The reasoning move is to ask, of any apparently isolated interaction: who is watching, what will they remember, and whom will they tell? If the answer is "a future audience that will adjust its treatment," the strategic structure is repeated-game-like even when the named parties never interact again.
The same lens exposes characteristic failure modes as structural rather than incidental: reputation lag (the stock reflects outdated behavior), gaming (agents invest in the observable proxy rather than the underlying conduct), whitewashing (an agent sheds a bad record by acquiring a fresh identity), and the cliff-edge collapse in which a long-accumulated record is destroyed by a single revelation. Tirole (1996) formalized the dynamics of such collapse and recovery, showing how an individual's reputation can be hostage to the past behavior of a group and why rehabilitation after a damaging episode is slow and uncertain. [13]
Knowledge Transfer¶
The biology of indirect reciprocity transfers directly to platform design: build a mechanism that makes past behavior observable to future counterparties, ensure the channel is hard to forge, and cooperation becomes self-enforcing without a central enforcer. This is not a loose analogy but a documented pipeline — online reputation-system designers explicitly drew on the indirect-reciprocity literature, and the structural correspondence (observed act, transmitted standing, third-party reward) is exact. Conversely, the economic insight that reputation is an asset with hysteresis — slow to build because it requires a long observable record, fast to lose because a single contrary signal can dominate the prior — transfers to diplomacy (a state's hard-won credibility for resolve evaporates with one backed-down threat) and to personal-credibility management. [1] The transferable reasoning is a checklist of channel properties: observability of conduct, fidelity of transmission, persistence and decay of the stock, and resistance to manipulation, which Bolton, Katok, and Ockenfels (2004) tested experimentally by comparing how different feedback-channel designs produce systematically different cooperation rates among the same population of traders. [14]
Examples¶
Formal/abstract¶
Repeated-game / folk-theorem setting: Consider a population of agents who are randomly matched in pairs to play a one-shot Prisoner's Dilemma, and who will almost never meet the same partner again. In a true one-shot game, defection is dominant and cooperation collapses. Now add a reputation channel: each agent carries an observable "image score" recording whether they cooperated or defected in recent interactions, visible to whomever they are matched with next. A strategy of "cooperate with partners in good standing, defect against those in bad standing" can now be an equilibrium: defecting earns a short-term gain but lowers the agent's score, inviting future partners to defect against them. The cooperation is carried entirely by the third-party-propagated record, not by any bilateral relationship. Mapped back: This is the prime's loop in its purest form — observed conduct becomes a transmitted standing that alters future treatment and so feeds back to discipline present conduct. The model isolates exactly the structure (a consultable stock summarizing past behavior across observers) that the prime names, stripped of any particular domain content; markets, alliances, and platforms are this same loop with different channels.
PageRank as computed reputation: PageRank treats the web as a population of pages whose "standing" is an aggregate of others' endorsements. A hyperlink from page A to page B is read as A vouching for B, and B's authority is the recursively computed sum of the authority flowing in through its inbound links. No page asserts its own importance; importance is a propagated stock constructed from the distributed behavior (linking) of the whole population. Mapped back: The pattern is reputation with hyperlinks as the observable conduct and the link graph as the transmission channel: a page's treatment by the ranking algorithm (where it appears in results) is governed by an aggregated record of how third parties have behaved toward it. Link-spam and link-farms are precisely channel attacks — attempts to forge the observable conduct that feeds the stock — which is why the structural failure modes of human reputation reappear, unchanged, in the computational substrate.
Applied/industry¶
Online marketplace feedback (eBay-style): A buyer must decide whether to send money to a seller they have never dealt with and will likely never deal with again. The marketplace supplies a reputation system: every past transaction leaves a rating, the ratings aggregate into a visible score and a written history, and the buyer consults this stock before transacting. Sellers, anticipating that future buyers will read the record, have an incentive to fulfill honestly even on a transaction whose own counterparty has no future leverage. The platform must defend the channel — preventing fake reviews, collusive rating rings, and "whitewashing" via fresh accounts — because the stock is only useful if its inputs are faithful. Mapped back: The platform has deliberately engineered the reputation loop that close-knit communities get for free through gossip: it manufactures observability (mandatory post-transaction ratings), transmission (a public profile), a persistent stock (the cumulative score), and feedback (buyers' altered willingness to transact). Every design lever — how long feedback persists, how much recent behavior is weighted, how identity is verified — is a choice about the channel and stock properties the prime identifies.
Sovereign credibility and deterrence: A state issues a deterrent threat — "if you cross this line, we will respond." Whether the threat works depends less on the state's present capabilities than on its reputation for actually carrying out such threats, built from a record of past behavior that adversaries and allies have observed and propagated. A state that has backed down before finds its current threats discounted; a state with a record of following through can deter with words alone. The reputation is an asset the state spends and replenishes: each crisis is partly a transaction in credibility, and a single conspicuous capitulation can collapse a standing built over decades. Mapped back: This is the prime operating with nations as agents, the international audience as the transmission channel, and credibility-for-resolve as the stock. The hysteresis property is vivid — credibility is slow to build (it requires a long observed record of follow-through) and fast to lose (one backed-down threat can dominate the prior) — which is exactly the asymmetry the economic treatment of reputation predicts.
Structural Tensions¶
T1: Reputation rewards the observable proxy, not the underlying conduct it is supposed to summarize. Because the stock is built from what observers can see and record, agents face an incentive to invest in the observable signal rather than the substance behind it. A firm may polish its review profile while quality slips; a scholar may chase citable, fashionable work over careful but uncitable work; a seller may game the rating mechanism. The tighter and more consequential the reputation channel, the stronger the pull toward optimizing the measure rather than the thing measured. The structure that makes reputation useful — compression of conduct into a transmissible token — is the same structure that lets the token detach from the conduct.
T2: Reputation is slow to build but fast to destroy, and this asymmetry is not symmetric in its costs. Building a stock requires a long observable record; a single damaging revelation can dominate the entire prior and collapse the standing overnight. This hysteresis disciplines behavior usefully (the threat of cliff-edge collapse keeps agents honest) but also makes reputation fragile and unforgiving: an agent who has behaved well for years can be ruined by one episode, sometimes one they did not commit, and rehabilitation is slow and uncertain even when the original record was overwhelmingly good.
T3: The channel that makes cooperation self-enforcing is also an attack surface. Reputation sustains cooperation only when the transmission channel faithfully carries the relevant behavior. But the same channel can be forged (fake reviews, link farms), flooded (Sybil identities), captured (a rating agency paid by those it rates), or weaponized (coordinated slander). Strengthening the channel's reach and consequence raises the payoff to subverting it. Every reputation system therefore runs an arms race between the agents whose conduct it is meant to summarize and the manipulators who target the channel instead.
T4: Reputation lets strangers coordinate, but only by importing the audience's biases. The promise of reputation is that a stranger can treat an agent appropriately without direct experience. But the stock the stranger consults is whatever the observing population chose to record and propagate, which carries that population's blind spots, prejudices, and uneven attention. The Matthew effect — prior standing attracting disproportionate further recognition — means reputation tends to amplify initial advantage rather than track current merit. Reputation can thus systematically misallocate trust and attention while appearing to be a neutral summary of behavior.
T5: A bad reputation can be as self-reinforcing as a good one, trapping agents below a recovery threshold. The same feedback loop that rewards good standing punishes bad standing, and the punishment can foreclose the very interactions through which an agent might rebuild. An agent denied transactions, links, alliances, or citations because of a poor record gets no further chance to generate the contrary signals that would update the stock. Whitewashing — discarding the spoiled identity for a fresh one — becomes individually rational, which in turn forces reputation systems to penalize newcomers, raising the entry cost for genuinely new and well-intentioned agents.
T6: Reputation presupposes a stable, trackable identity, yet identity is exactly what strategic agents can manipulate. The loop only closes if today's agent is reliably linked to the record of yesterday's conduct. When identity is cheap to change or hard to verify, the stock detaches from the agent: cheaters shed bad records, impostors borrow good ones, and the standing no longer governs the agent it was built from. Reputation systems are therefore forced into a costly identity-binding problem, and the strength of any reputation effect is capped by how firmly conduct can be pinned to a persistent identity.
Structural–Framed Character¶
Reputation sits toward the structural side of the structural–framed spectrum, with some framing: it is the aggregated, publicly available record of an agent's past behavior that propagates beyond the original parties and governs how third parties treat the agent in future interactions. Its core is information flow across two dimensions at once — across time, as past conduct conditions future treatment, and across observers.
The underlying pattern is substrate-neutral, value-neutral, and recognized rather than imported: the same information-flow structure is indirect reciprocity in evolutionary biology and PageRank-style propagated standing in computing. What tilts it toward framing is that the concept took shape in sociology and presupposes agents and observers — it cannot be defined wholly without a population that tracks and transmits records, leaving its origin and practice-dependence partly engaged. The recognized core dominates the social wrapper, so it reads mixed but structural-leaning.
Substrate Independence¶
Reputation is a highly substrate-independent prime — composite 4 / 5 on the substrate-independence scale. Its structure — past behavior propagating as a signal beyond the original parties to govern future treatment — transfers genuinely across social settings (markets, international relations), biology (indirect reciprocity sustaining cooperation), and computation (eBay feedback, webs of trust, PageRank-style endorsement). The transfer evidence is explicit and cross-substrate: the note that biological indirect reciprocity directly informs platform mechanism design shows the crossing is real, not analogical. It lands at 4 rather than 5 because the pattern presupposes observers and memory, so it does not reach physical substrates.
- Composite substrate independence — 4 / 5
- Domain breadth — 4 / 5
- Structural abstraction — 4 / 5
- Transfer evidence — 4 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Reputation is a decomposition of Feedback
Reputation is the specific shape feedback takes when the output is an agent's conduct and the return path runs across time and across observers. The agent's behavior at t feeds into a public record; that record shapes how third parties treat the agent at t+1, which feeds back into the agent's incentives for behavior at t+1. The loop closure feedback requires is realized through the information channel third parties maintain, and the present value of future reputation premiums disciplines current quality through the same input-from-prior-output structure feedback names.
Path to root: Reputation → Feedback
Neighborhood in Abstraction Space¶
Reputation sits among the more crowded primes in the catalog (2nd percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.
Family — Cooperation, Trust & Institutional Bonds (19 primes)
Nearest neighbors
- Institution — 0.87
- Informal Enforcement — 0.85
- Cooperation — 0.85
- Conformity — 0.85
- Competition — 0.84
Computed from structural-signature embeddings · 2026-05-29
Not to Be Confused With¶
Reputation must be distinguished from Trust, which is its nearest neighbor and the prime most often conflated with it. Trust is a truster's relational stance toward a specific party under conditions of vulnerability: it is held by one agent, directed at another, and constituted by a willingness to accept exposure to that party's discretion. Reputation, by contrast, is a population-level information stock about an agent that exists independently of any particular truster and that even a complete stranger can consult. The two are causally linked but structurally distinct: reputation is often an input to trust — I trust you partly because of your reputation — but I can trust you on grounds that have nothing to do with reputation (direct experience, kinship, a costly hostage, contractual recourse), and your reputation exists whether or not any given person trusts you. The cleanest test is the locus: trust lives in a relationship between two parties; reputation lives in the memory of an observing crowd. A seller can have an excellent reputation that a particular skeptical buyer refuses to trust, and a friend I trust completely may have no reputation at all because no third party has ever had occasion to record their conduct. Treating the two as one collapses the distinction between a dyadic disposition and a propagated stock, and it obscures the central, surprising claim the reputation prime makes — that an agent can be treated as trustworthy by someone who has never interacted with them and holds no relationship with them whatsoever, purely on transmitted standing.
Reputation is not Signaling, which is its structural parent and a frequent source of confusion. A signal is a single, typically costly act emitted by an agent to convey otherwise-unobservable information about its type — a peacock's tail, a warranty, a credential, a deliberately burned bridge. Signaling is a flow: an action taken in the moment, addressed to an audience, intended to move beliefs. Reputation is the stock that such flows accumulate into and against which they are read. The relationship is generative: individual signals are the deposits, and reputation is the balance. But the prime's distinctive content is everything that the single-act framing of signaling omits — the third-party propagation (a signal emitted to A becomes part of a record consulted by B), the persistence and decay of the accumulated stock over time, and the long-run stakes that the existence of the stock confers on each individual signal. A signal can be honest or dishonest in the instant; reputation introduces the time dimension in which dishonesty has delayed, aggregated consequences. One can model a reputation as the history of signals, but one cannot reduce reputation to any single signal, and it is precisely the cross-time, cross-observer aggregation — not the momentary emission — that does the explanatory work in repeated-game and indirect-reciprocity accounts.
Reputation is not Legitimacy, with which it is sometimes blurred in political and organizational contexts. Legitimacy is a normative property: the accepted rightfulness of an authority's claim to rule, to issue binding decisions, or to command compliance, grounded in shared beliefs about what makes power proper (legal procedure, tradition, consent, charisma). Reputation is a positive property: a track-record inference about how an agent has actually behaved, usable by anyone observing past conduct regardless of any question of rightful authority. The two diverge sharply at the boundaries. An authority can possess legitimacy — be widely accepted as having the right to rule — while having a poor reputation for competence or honesty; conversely, an actor with no claim to authority at all (a private firm, a reviewer, a foreign state) can have a powerful reputation. Legitimacy answers "does this agent have the standing to be obeyed?"; reputation answers "given what this agent has done before, how should I treat it?" One is about the normative ground of authority, the other about the empirical extrapolation of behavior, and an institution can be losing one while gaining the other.
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.
Notes¶
Reputation operates at many scales and over many channels, and the prime is agnostic about which: the agent may be an individual, a firm, a nation, a species lineage, a web page, or an algorithm; the channel may be word-of-mouth gossip, a credit bureau, a platform's rating database, the citation graph, or the link structure of the web. What is invariant is the loop — observed conduct, transmitted across an audience, stored as a stock, fed back as altered treatment. When analyzing a concrete case it is usually most productive to name the four components explicitly: what conduct is observable, who carries it forward and how faithfully, where and how long the stock persists, and how the stock changes treatment.
A recurring confusion is between reputation and the true underlying property it is supposed to track (quality, type, reliability, resolve). The prime names the propagated record, which is an inference and can diverge arbitrarily from the truth. Much of the practical action in reputation systems is in managing this divergence — designing channels so that the stock tracks conduct closely, weighting recent behavior to limit lag, and binding records to persistent identities so manipulators cannot detach standing from agent.
Reputation also carries an implicit assumption that the observing population shares enough of a common interpretive frame to read conduct the same way. When audiences fragment — different communities recording different behaviors, or reading the same behavior with opposite valence — an agent can hold simultaneously good and bad reputations in different sub-populations, and the unitary "reputation" the prime names dissolves into several. This is increasingly visible in networked environments where reputation is computed differently by different platforms and communities.
References¶
[1] Klein, B., & Leffler, K. B. (1981). The role of market forces in assuring contractual performance. Journal of Political Economy, 89(4), 615–641. Models how the present value of a future reputation premium disciplines current quality and supports the hysteresis (slow-to-build, fast-to-lose) character of reputational capital across markets and personal credibility. ↩
[2] Kreps, D. M., & Wilson, R. (1982). Reputation and imperfect information. Journal of Economic Theory, 27(2), 253–279. Shows that even a small probability of a "committed" type can sustain cooperative play in finitely repeated games — the formal reputation effect. ↩
[3] Nowak, M. A., & Sigmund, K. (2005). Evolution of indirect reciprocity. Nature, 437(7063), 1291–1298. Identifies indirect reciprocity — third parties rewarding "good standing" — as the self-enforcing engine that closes the reputation loop in evolutionary settings. ↩
[4] Resnick, P., Zeckhauser, R., Friedman, E., & Kuwabara, K. (2000). Reputation systems. Communications of the ACM, 43(12), 45–48. Documents how online reputation systems collect, aggregate, and distribute feedback, re-implementing the indirect-reciprocity logic across the computational substrate. ↩
[5] Shapiro, C. (1983). Premiums for high quality products as returns to reputations. The Quarterly Journal of Economics, 98(4), 659–679. Derives an equilibrium price-quality schedule in which a reputation premium supports prices above marginal cost as a bond against quality-cutting. ↩
[6] Nowak, M. A., & Sigmund, K. (1998). Evolution of indirect reciprocity by image scoring. Nature, 393(6685), 573–577. Foundational simulations showing image-scoring dynamics make third-party-rewarded cooperation evolutionarily stable without direct payback. ↩
[7] Page, L., Brin, S., Motwani, R., & Winograd, T. (1999). The PageRank citation ranking: Bringing order to the web (Technical Report 1999-66). Stanford InfoLab. Formalizes an inbound hyperlink as a transferable endorsement of a page's authority, computing standing as a propagated stock from distributed link behavior. ↩
[8] Schelling, T. C. (1960). The Strategy of Conflict. Harvard University Press. Introduces strategic pre-commitment and commitment devices as deliberate self-binding mechanisms; the contrast with inadvertent lock-in is structural — both produce future-self constraint, but commitment devices are sought while lock-ins emerge as side effects of locally reasonable choices. ↩
[9] Merton, R. K. (1968). The Matthew effect in science. Science, 159(3810), 56–63. Sociological formalization of cumulative advantage in the reward system of science: well-known scientists accrue disproportionate credit, attention, and resources, and successive recognition compounds — a non-market instantiation of increasing returns operating without prices. ↩
[10] Dellarocas, C. (2003). The digitization of word of mouth: Promise and challenges of online feedback mechanisms. Management Science, 49(10), 1407–1424. Argues platforms engineer trust among strangers at scale by deliberately constructing the observation-transmission-aggregation channel. ↩
[11] Milgrom, P. R., North, D. C., & Weingast, B. R. (1990). The role of institutions in the revival of trade: The law merchant, private judges, and the Champagne fairs. Economics & Politics, 2(1), 1–23. Shows how institutionalized reputation tracking (law-merchant courts) let medieval merchants who would never meet again trade honestly among strangers. ↩
[12] Kandori, M. (1992). Social norms and community enforcement. The Review of Economic Studies, 59(1), 63–80. Proves that community enforcement can sustain cooperation in large populations whose members rarely meet the same partner twice. ↩
[13] Tirole, J. (1996). A theory of collective reputations (with applications to the persistence of corruption and to firm quality). The Review of Economic Studies, 63(1), 1–22. Formalizes the dynamics of reputation collapse and slow, uncertain rehabilitation, including how an individual's standing can be hostage to a group's past. ↩
[14] Bolton, G. E., Katok, E., & Ockenfels, A. (2004). How effective are electronic reputation mechanisms? An experimental investigation. Management Science, 50(11), 1587–1602. Tests experimentally how different feedback-channel designs produce systematically different cooperation rates among the same population of traders. ↩