Assembly Bonus Effect¶

The small-group finding that on the right task types a group's output exceeds even its best member's predicted performance — a positive interaction residual from mutual error-correction and partial-insight combination that no aggregation of individual scores captures.

Core Idea¶

The assembly bonus effect (Collins and Guetzkow, 1964) is the small-group performance finding that a group's output on certain problem types exceeds what would be predicted from any simple aggregation of its members' individual abilities — the interaction itself generates a positive residual that no baseline constructed from individual scores captures. The concept is analytically defined against Steiner's (1972) group productivity framework, which characterizes most group performance as falling below the "potential productivity" that an additive or best-member baseline would predict, because coordination overhead, motivation loss, and social loafing impose process losses. The assembly bonus is the positive counterpart: on problems where the group exceeds even the best individual member's predicted performance, the excess is the bonus. The mechanism involves interactive processes that are genuinely emergent from group composition — mutual error-correction, complementary partial-insight combination, distributed information pooling across differently-informed members, and question-asking chains where one member's partial solution unlocks another's — none of which a lone solver could replicate or that simple aggregation of individual outputs would produce. The effect is task-contingent: it is most reliably observed on intellective or eureka-type problems where there is a demonstrably correct answer, where partial information is distributed across members, or where error-correction by a second perspective adds value; it is least likely on purely additive tasks, disjunctive tasks where the first correct solution wins, or tasks requiring synchronized motor performance where coordination losses dominate.

Structural Signature¶

Sig role-phrases:

the favorable task — a problem whose structure admits interactive solution: a demonstrably correct answer, partial information distributed across members, or error-correction value
the differently-informed members — solvers holding complementary partial insights or distinct perspectives no single member combines alone
the best-member baseline — the demanding reference: the predicted performance of the strongest solo performer (not a weaker additive baseline)
the interactive process — the emergent moves of group work: mutual error-correction, complementary partial-insight combination, distributed information pooling, question-asking chains where one member's partial solution unlocks another's
the positive interaction residual — the bonus proper: output above the best-member baseline that no aggregation of individual scores would predict
the competing process-loss term — coordination overhead, motivation loss, and social loafing, the negative residual pulling output below baseline
the task-contingent sign — task structure sets whether the interaction term goes positive (intellective/eureka, distributed-info, error-correction tasks) or non-positive (additive, disjunctive, synchronized-motor tasks)

What It Is Not¶

Not the slogan "the whole is greater than the sum of its parts." That phrase slides freely between an additive baseline and a best-member one. The assembly bonus is the demanding version: output above the strongest solo performer's predicted score, so it cannot be discharged as members merely not impeding one another, the way a gain over a weak additive baseline can.
Not statistical judgment aggregation (wisdom of crowds). Averaging independent, diverse judgments can also beat the best member, but it does so with no interaction — no mutual error-correction, no partial-insight combination. The assembly bonus requires members to actually interact; the aggregation result is a sibling effect with a different mechanism, and conflating them is a common slip.
Not a general property of groups. It is task-contingent. The surplus is expected only on tasks with a demonstrably correct answer, distributed partial information, or error-correction value; on purely additive, disjunctive, or synchronized-motor tasks the interaction term goes the other way and a process loss is the predicted outcome.
Not a claim that groups outperform individuals. It is one signed term in a two-sided ledger — the positive residual competing against coordination overhead, motivation loss, and loafing. On most task structures those process losses dominate and the group falls below baseline; the bonus names only the cases where the interaction term wins.
Not social facilitation. Being energized (or inhibited) by the mere presence of others changes individual effort or arousal; it is not the emergent combinatorial gain of interactive problem-solving. The assembly bonus is produced by information pooling and perspective-combining among genuinely interacting members, not by an audience effect.

Scope of Application¶

The assembly bonus effect lives across the small-group and team-performance subfields that share the human-collective substrate — interactive perspective-sharing and complementary expertise under group interaction; its reach is within that domain, since the substrate-neutral supra-additive-combination pattern is carried by the parent (synergy_and_antagonism, emergence), not by this name. (Statistical judgment aggregation — wisdom of crowds — beats the same best-member baseline but by averaging independent judgments, so it is a sibling effect that stays outside this map.)

Small-group decision-making and problem-solving — the home turf: Collins and Guetzkow's original eureka/intellective-task setting and Hill's review tradition, where a group on a demonstrably-correct-answer problem can solve faster and more reliably than its best member.
Organizational psychology — the synergy sought in cross-functional and design teams and in jury deliberation, where complementary expertise and mutual error-correction produce output above the strongest solo contributor.
Software-engineering research — the contested pair-programming and group-brainstorming productivity claims, where a real interactive bonus must be distinguished from coordination-loss-dominated drag.

Clarity¶

Naming the assembly bonus restores the missing half of the ledger that Steiner's process-loss framework had quietly closed off. Once "actual = potential − process losses" became the field's default picture, group performance could only ever fall short of a fixed baseline, and the genuine cases where a group beats even its best member had no place to be recorded — they looked like measurement noise or an overgenerous baseline. The bonus names them as a real positive term: a residual the interaction itself produces, above what any aggregation of individual scores predicts. That reframes the central question of the subfield from "how much did coordination cost?" to the sharper two-sided one — "is this task one where interaction generates a surplus, or only one where it imposes a tax?"

The concept also forces a discipline the loose phrase "groups are more than the sum of their parts" lets people skip: stating the baseline the surplus is measured against. A gain over an additive baseline is a weaker claim than a gain over the best individual member; the assembly bonus is specifically the latter, the excess above the strongest solo performer, which is why it cannot be explained away as members simply not getting in each other's way. And it makes the effect's task-contingency the operative variable rather than a footnote: the surplus is expected on intellective or eureka problems with a demonstrably correct answer, on tasks where partial information is distributed across members, or where a second perspective corrects error — and is not expected on purely additive tasks, on disjunctive tasks where the first correct solution simply wins, or on synchronized-motor tasks where coordination losses dominate. The sharp question a practitioner can now ask is therefore not "will the team help?" but "does this task have the structure — distributed information, error-correction value, complementary partial insight — on which a bonus, rather than a process loss, is the predicted outcome?"

Manages Complexity¶

Group performance, observed across the small-group and team literatures, is a tangle of seemingly contradictory outcomes: some teams crawl below what their members could do alone, dragged down by coordination overhead and loafing; some match their best member; and some genuinely exceed even their strongest solo performer. Confronted case by case, each result invites its own story, and the field's process-loss default — actual equals potential minus losses — can only register the first two, leaving the third to look like noise or an overgenerous baseline. The assembly bonus effect completes the ledger and compresses the whole range onto a single signed account: performance equals a stated baseline, plus an interaction-driven gain, minus process losses. The analyst no longer needs a separate explanation for each team; they track two opposing terms — the surplus the interaction can generate and the tax coordination imposes — and read the net deviation from baseline off their balance.

The compression is sharp because the sign of the interaction term is not a continuum to be modeled afresh each time but a function of one readable variable: task structure. The effect is task-contingent, and the contingency is the parameter the analyst tracks. On problems with a demonstrably correct answer, with partial information distributed across members, or where a second perspective corrects error, the interaction term is predicted positive and a bonus is expected. On purely additive tasks, on disjunctive tasks where the first correct solution simply wins, or on synchronized-motor tasks where coordination losses dominate, the interaction term is predicted non-positive and a process loss is expected. So the open-ended question "will this team beat its parts?" collapses to a structural read of the task — does it have distributed information, error-correction value, or complementary partial insight? — from which the qualitative outcome (surplus versus tax) follows, without re-deriving each team's dynamics.

The concept also imposes a discipline that itself reduces analytic load: it forces the baseline to be stated. "More than the sum of its parts" is an unfalsifiable slogan that quietly slides between an additive baseline and a best-member one; the assembly bonus fixes the demanding reference — the excess over the strongest solo performer — so a claimed surplus cannot be explained away as members merely not impeding one another. With the baseline pinned and the task-contingency read as the operative variable, the question a practitioner faces shifts from an unbounded "is collaboration good here?" to a determinate "does this task have the structure on which the interaction term goes positive?" — and the entire two-sided range of group outcomes, from synergy to drag, reads off one equation with a sign set by task type.

Abstract Reasoning¶

The assembly bonus effect licenses reasoning that treats group performance as a signed ledger — a stated baseline, plus an interaction-driven gain, minus process losses — whose interaction term has a sign set by task structure. The analyst reasons from a task's structure to whether the group will exceed or fall short of its baseline, and back from a measured surplus to whether it cleared the demanding best-member reference.

Diagnostic (read surplus or deficit from the two opposing terms; demand the right baseline). The defining inference re-reads a group result against a two-sided account that the process-loss default cannot supply. A team that beats even its strongest solo member is diagnosed as showing a genuine positive interaction residual — emergent mutual error-correction, complementary partial-insight combination, distributed information pooling, question-asking chains — not as noise or an overgenerous baseline. A team that lags is diagnosed as process losses (coordination overhead, motivation loss, loafing) dominating the interaction term. The diagnostic insists on the baseline the surplus is measured against: a gain over an additive baseline is a weaker claim than a gain over the best individual member, and the assembly bonus is specifically the latter — so a claimed synergy is checked against the strongest solo performer and cannot be explained away as members merely not impeding one another. The inference runs net deviation from baseline → the balance of interaction gain versus process loss (against a stated best-member reference), never "the group did well" → unexamined synergy.

Interventionist (select the task and configure the group, predict surplus or tax). Because the sign of the interaction term is set by task structure, the levers are task selection and group composition with forecast consequences. Pose a problem with a demonstrably correct answer, distributed partial information, or error-correction value, and a bonus is predicted; pose a purely additive task, a disjunctive task where the first correct solution simply wins, or a synchronized-motor task, and a process loss is predicted instead. To capture a bonus, compose the group for complementary expertise and diverse partial information and install integration mechanisms that let one member's partial solution unlock another's; to avoid a tax, reduce coordination overhead and loafing on tasks where those dominate. Each manipulation pairs a change in task structure or group configuration with a predicted sign and size of the deviation from baseline.

Boundary-drawing (best-member baseline; task-contingency fixes the regime; interactive process, not aggregation). The concept fixes its scope through several lines. It pins the demanding reference — the excess over the strongest solo performer — so a gain measured against a weaker additive baseline is outside the assembly-bonus claim. It makes task-contingency the operative variable that bounds where a bonus is even possible: the surplus regime is intellective/eureka tasks with a correct answer, distributed-information tasks, and error-correction tasks; the process-loss regime is additive, disjunctive, and motor-coordination tasks. And it is scoped to interactive group process — mutual correction and partial-insight combination — distinct from statistical aggregation of independent judgments and from the broader supra-additive-combination pattern, so a gain produced by mere averaging falls outside this effect even when it beats the best member.

Predictive / branch-ordering. From a structural read of the task — does it have a demonstrably correct answer, distributed information, or error-correction value? — the analyst forecasts the sign of the interaction term and therefore the qualitative outcome before the group runs: a bonus (group beats its best member) on the favorable structures, a process loss (group falls below baseline) on the additive, disjunctive, or motor structures. Surplus versus tax, and which task types yield which, read off one equation whose sign is determined by task type.

Knowledge Transfer¶

Within small-group and team-performance research the effect transfers as mechanism, across the subfields that share the human-collective substrate — perspective-sharing, complementary expertise, and interactive social cognition under group interaction. The signed-ledger framing (baseline + interaction gain − process losses), the demanding best-member baseline, and the task-contingency that sets the sign all carry intact. In small-group decision-making and problem-solving it is Collins and Guetzkow's original eureka-task setting and Hill's review tradition. In organizational psychology it is the synergy sought in cross-functional and design teams and jury deliberation. In software-engineering research it is the contested pair-programming and brainstorming productivity claims. Across these the diagnostic ("does this task have distributed information, error-correction value, or complementary partial insight?") and the levers (select the task structure, compose for complementary expertise, install integration mechanisms) move without translation, because the mechanism — a positive residual produced by interactive group process measured against the strongest solo member — is the same.

A boundary worth marking even inside the collective family: the assembly bonus is specifically a property of interactive group process, and it is not the same as statistical judgment aggregation. The wisdom-of-crowds result — aggregate judgments beating the best individual under independence and diversity — beats the same best-member baseline but does so by averaging independent judgments, with no mutual error-correction or partial-insight combination, so it is a sibling effect rather than this one. Conflating them is a common slip; the assembly bonus requires members to actually interact.

Beyond the human collective the honest reading is shared abstract mechanism, not the named concept. Three more-general patterns the effect instantiates do recur across substrates: supra-additive combination (already carried by synergy_and_antagonism), qualitative novelty from interaction (carried by emergence), and aggregation gains under independence and diversity (carried by the judgment-aggregation treatments). Those parents are what travel — into ensemble methods in machine learning, prediction-market design, and the like. What stays home-bound is the concept's own machinery: the interactive information-pooling and perspective-combining of human social cognition, the contrast against Steiner's process-loss ledger, and the eureka/intellective task taxonomy that fixes the sign. So invoking "an assembly bonus" for an ML ensemble or a distributed physical system is, as the seed flags, distant analogy: the supra-additive shape is borrowed while the human-group interaction mechanism that produces it is dropped — and where the gain is real it is better named by synergy_and_antagonism or emergence than by this effect. The discipline is to carry the supra-additive-combination parent (or the aggregation parent for averaging cases) wherever the substrate is not an interacting human group, and to reserve "assembly bonus effect" for the small-group interaction whose positive residual it actually measures (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

Patrick Laughlin and colleagues supplied a clean demonstration with the "letters-to-numbers" task, in which solvers must deduce which digit each of ten letters stands for by proposing equations, receiving the coded answers, and reasoning from the feedback. In a 2006 study (Laughlin, Hatch, Silver, and Boh, Journal of Personality and Social Psychology), groups of three, four, and five were compared against the best of an equivalent number of individuals working alone. Groups of three, four, and five solved the problems in fewer trials than the best of the same number of individuals — beating not the average solo performer but the strongest one. The task has a demonstrably correct answer and rewards mutual hypothesis-testing and error-correction, so the group's interactive checking generated real gains no lone reasoner and no pooling of separate solo scores would produce.

Mapped back: The letters-to-numbers puzzle is the favorable task — a demonstrably correct answer with error-correction value. Comparing groups against the best of an equal number of solo solvers is exactly the best-member baseline. The members' mutual hypothesis-testing is the interactive process, and clearing the best individual is the positive interaction residual the effect names.

Applied / In Practice¶

Aviation Crew Resource Management (CRM) is a working deployment of the same principle. After accident investigations in the 1970s (notably the 1978 United 173 fuel-exhaustion crash) traced disasters to crews in which a captain's error went unchallenged, airlines restructured the cockpit as an error-correcting team: first officers and engineers are trained and authorized to cross-check, question, and correct the captain, and standardized challenge-and-response callouts institutionalize the checking. The design bet is precisely the assembly bonus — that a crew catching and correcting one another's slips performs more safely than even its most skilled individual pilot flying alone, because flying is a task rich in error-correction value. CRM is now a global standard in commercial aviation and has been adapted into surgical and other high-stakes teams.

Mapped back: Piloting under CRM is the favorable task (high error-correction value); captain, first officer, and engineer are the differently-informed members. Cross-checking and challenge-response callouts are the interactive process, and a crew catching an error the lone captain would have missed is the positive interaction residual measured against the best-member baseline.

Structural Tensions¶

T1: Bonus term versus process-loss tax (one interaction, two opposite residuals). The signed ledger treats the interaction gain and the coordination cost as two competing terms, but they are not independent knobs — the very interaction that pools partial insight and corrects error is the interaction that incurs coordination overhead, motivation loss, and loafing. Adding members and increasing their mutual engagement raises the ceiling of the positive residual and the floor of the process-loss tax at once, so there is no configuration that buys the bonus without paying some of the tax. The construct's clarity comes from separating the terms on paper; its difficulty is that in a real group they ride the same underlying process and cannot be independently maximized and minimized. Diagnostic: Can the interaction that would generate the bonus here be had without also incurring the coordination and motivation costs it necessarily brings — or are the gain and the tax two faces of the same group process?

T2: The demanding best-member baseline (a strong claim that is hard to establish). Fixing the reference at the strongest solo performer is what makes the assembly bonus a real claim rather than the unfalsifiable slogan "more than the sum of its parts" — it cannot be discharged as members merely not impeding one another. But that stringency is double-edged: establishing a genuine bonus requires the counterfactual performance of the best of an equivalent number of independent solvers, which is expensive to measure and often only estimated, and small errors in that estimate can flip the verdict. The baseline that makes the effect worth naming is also the baseline that makes it hard to demonstrate, so claimed bonuses are only as credible as the best-member counterfactual they are checked against. Diagnostic: Is the surplus measured against an actual best-of-N solo counterfactual, or against a weaker additive baseline dressed up as a best-member one?

T3: Task-contingency versus circular classification (favorable because it worked). The construct's compression rests on task structure setting the sign of the interaction term in advance: demonstrably-correct-answer, distributed-information, and error-correction tasks predict a bonus; additive, disjunctive, and motor tasks predict a tax. This is predictive only if task structure is read independently of the outcome. The hazard is retrospective relabeling — a group that beat its best member is declared to have faced a "favorable, error-correction-rich task," and one that lagged a "coordination-dominated" one, after the fact — which converts a falsifiable forecast into an unfalsifiable post-hoc fit. The task taxonomy earns its keep only when the structural read is committed before the result is known. Diagnostic: Was the task classified as bonus-favorable from its structure before the group ran, or is "favorable task" being inferred from the fact that a bonus appeared?

T4: Error-correction versus error-contamination (interaction cuts both ways). The bonus is credited to mutual error-correction — a second perspective catching a slip the lone solver would keep. But the same interactive coupling that lets one member correct another lets one member's error propagate into others', and lets confident-but-wrong members anchor the group; this is exactly the failure mode (groupthink, cascade) that independent judgment aggregation avoids by forbidding interaction. So interaction is not purely corrective: it simultaneously enables the error-catching that produces the bonus and the error-spreading that can sink a group below even its average member. The construct names the upside of coupling while its sibling (wisdom of crowds) is defined by avoiding coupling's downside. Diagnostic: In this group, is interaction letting members catch one another's errors, or letting one member's error and confidence contaminate the rest?

T5: Emergent residual versus residual-of-ignorance (genuine novelty or unmodeled aggregation). The bonus is defined as a positive residual "no aggregation of individual scores would predict" — which makes it, definitionally, whatever the individual-baseline model fails to capture. That framing risks conflating a genuinely emergent combinatorial gain with a mere shortfall of the baseline: a richer model of how individual abilities combine might absorb some "bonus" as predictable after all, shrinking the emergent residual without any change in the group. The construct's claim to emergence is thus partly hostage to the crudeness of the baseline it is measured against, and "irreducible interaction gain" and "gain the additive model was too simple to foresee" are hard to tell apart. Diagnostic: Is the surplus a genuine product of interaction that no model of individual combination could predict, or is it the gap left by an under-specified aggregation baseline?

T6: Autonomy versus reduction (an interactive-group effect, or its supra-additive parents). The assembly bonus is a genuinely named small-group finding with irreducible home cargo — the interactive information-pooling and perspective-combining of human social cognition, the contrast against Steiner's process-loss ledger, and the eureka/intellective task taxonomy that fixes the sign. It transfers as mechanism across small-group, organizational, and software-team research because they share the interacting-human-group substrate. But the thinner patterns it instantiates — supra-additive combination (synergy_and_antagonism), qualitative novelty from interaction (emergence), and, for the averaging sibling, aggregation gains under independence — are what travel to ML ensembles or prediction markets; invoking "an assembly bonus" there borrows the supra-additive shape while dropping the human-interaction mechanism. The tension is between a construct that earns its own study for interactive group process and the recognition that its cross-substrate residue belongs to those parents. Diagnostic: Resolve toward the parents (synergy_and_antagonism, emergence, or the aggregation prime) when the substrate is not an interacting human group; toward "assembly bonus effect" when diagnosing a group that beat its best member through genuine interaction.

Structural–Framed Character¶

The assembly bonus effect sits at mixed on the structural–framed spectrum — off both poles, balancing a genuine emergent-interaction mechanism against a binding to the human-collective substrate and a partly theory-relative definition. The criteria pull in both directions. Evaluative_weight leans structural: despite the positive connotation of "bonus," the effect is analytically one signed term in a two-sided ledger — a positive interaction residual competing against a process-loss tax — so the construct itself is evaluatively neutral about whether a given group helps or hurts; the entry stresses it is "not a claim that groups outperform individuals." Human_practice_bound points framed: the effect is specifically a property of interacting human groups — it requires members who actually pool partial insight, correct one another's errors, and combine perspectives through social cognition — and dissolves off that substrate, which is why the entry marks ML-ensemble or distributed-system invocations as "distant analogy." But, as with other cognition-anchored entries, the practice it is bound to is a natural human faculty (social problem-solving), not a constituted institution. Institutional_origin is mixed-leaning-structural: the phenomenon is a natural empirical finding about group cognition (Collins–Guetzkow 1964), discovered rather than decreed — yet it is defined analytically against Steiner's process-loss framework and a best-member baseline, so its identity as a "positive residual no aggregation predicts" is partly theory-relative (T5 flags that a richer baseline model could absorb some of the residual). Vocab_travels points framed: the interactive-process/process-loss ledger and the eureka/intellective task taxonomy are pinned to small-group psychology and do not float free. Import_vs_recognize is bimodal: within small-group, organizational, and software-team research the mechanism transfers as recognition (one interacting-human-group substrate), while beyond it — ML ensembles, prediction markets — only the supra-additive shape is borrowed by analogy.

The portable structural skeleton is a single idea — supra-additive combination: interaction producing qualitative gain beyond what aggregating the parts predicts — carried by synergy_and_antagonism (supra-additive combination) and emergence (qualitative novelty from interaction), with the judgment-aggregation prime as the parent for the averaging sibling (wisdom of crowds). That skeleton genuinely travels to non-human substrates (ensemble methods, prediction markets), which is what gives the effect its structural substance and keeps it at mixed rather than framed-leaning. But it does not reach the structural side proper, because that skeleton is exactly what the assembly bonus instantiates from those parents, not what makes the named small-group effect travel: the cross-substrate reach belongs to synergy_and_antagonism and emergence, while the distinctive cargo — the interactive information-pooling and perspective-combining of human social cognition, the contrast against Steiner's process-loss ledger, the demanding best-member baseline, and the eureka/intellective task taxonomy that fixes the sign — is the psychology furniture that stays home. Its character: a real, evaluatively neutral, emergent supra-additive mechanism of interacting human groups, defined against a process-loss baseline — structural in the synergy/emergence skeleton it hands off to its parents, but bound by its interactive-human-cognition mechanism and task taxonomy to small-group psychology, leaving it mixed rather than a free-floating prime.

Structural Core vs. Domain Accent¶

This section decides why the assembly bonus effect is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity — there is no separate section for that.

What is skeletal (could lift toward a cross-domain prime). Strip the small-group setting and a thin relational structure survives: when parts interact rather than merely add, the combined output can exceed what aggregating the parts predicts, so the interaction contributes a positive residual that no sum of individual contributions captures. The pieces that travel are abstract — a set of contributing units, an interaction that couples them, a baseline built from the units taken separately, and a supra-additive surplus (or, on other structures, a deficit) that the coupling produces. That skeleton is genuinely substrate-portable, which is exactly why it recurs in the catalog as the general primes the entry instantiates — synergy_and_antagonism (supra-additive combination), emergence (qualitative novelty from interaction), and, for the averaging sibling, the judgment-aggregation prime — and why it reappears in ML ensembles, prediction markets, and other non-human systems. But it is the core it shares, not what makes the assembly bonus distinctive.

What is domain-bound. Almost all the content is small-group-psychology furniture, and none of it survives extraction intact. The mechanism is specifically the interactive information-pooling and perspective-combining of human social cognition — mutual error-correction, complementary partial-insight combination, distributed information pooling, question-asking chains where one member's partial solution unlocks another's — none of which is a general property of coupled parts but a fact about how differently-informed people reason together. The baseline is the demanding best-member reference (the predicted performance of the strongest solo human performer), and the effect is defined analytically against Steiner's process-loss ledger — coordination overhead, motivation loss, social loafing — so its very identity as "a positive residual no aggregation predicts" is relative to that framework. The task taxonomy that fixes the sign (intellective/eureka tasks with a demonstrably correct answer, distributed-information tasks, error-correction tasks versus additive, disjunctive, and synchronized-motor tasks) is likewise small-group-research vocabulary. The decisive test: remove the interacting human group — replace the members with independent units averaged without interaction — and it is no longer the assembly bonus but its sibling (wisdom-of-crowds aggregation) or a looser supra-additive gain; the effect is constituted by the human-interaction mechanism the prime bar asks it to shed.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. The assembly bonus's transfer is bimodal. Within the human-collective substrate the mechanism travels intact — the signed-ledger framing, the best-member baseline, and the task-contingency carry with full content across small-group decision-making, organizational psychology, and software-engineering team research, because each is an interacting human group; the diagnostic and the levers move without translation. Beyond the human collective it travels only by analogy: invoking "an assembly bonus" for an ML ensemble or a distributed physical system borrows the supra-additive shape while dropping the interactive-human-cognition mechanism that produces it, and even the closest human sibling (statistical judgment aggregation) is a different mechanism because it forbids interaction. And when the bare structural lesson is needed cross-domain — interaction can produce a gain beyond what aggregating the parts predicts — it is already supplied, in more general form, by synergy_and_antagonism and emergence (or the aggregation prime for the averaging case). The cross-domain reach belongs to those parents; "assembly bonus effect," as named, carries the interactive-human-cognition mechanism, the process-loss ledger, the best-member baseline, and the eureka/intellective task taxonomy — small-group-psychology baggage that does not and should not travel as a unit.

Relationships to Other Abstractions¶

Current abstraction Assembly Bonus Effect Domain-specific

Parents (1) — more general patterns this builds on

Assembly Bonus Effect is a kind of Synergy and Antagonism Prime

The Assembly Bonus Effect is synergy specialized to interacting human groups whose performance exceeds a best-member baseline through mutual error correction and partial-insight combination.

Hierarchy path (1) — routes to 1 parentless root

Assembly Bonus Effect → Synergy and Antagonism → Nonlinearity

Not to Be Confused With¶

Wisdom of crowds (statistical judgment aggregation). The closest sibling: aggregating many independent diverse judgments (averaging estimates, prediction markets) can also beat the best individual — but it does so by forbidding interaction, precisely to prevent members contaminating one another. The assembly bonus is the opposite: its surplus is produced by mutual error-correction and partial-insight combination among genuinely interacting members. Same demanding best-member baseline, opposite mechanism. Tell: did the gain come from members interacting, pooling, and correcting each other (assembly bonus), or from mechanically averaging judgments made in isolation (wisdom of crowds)?
Steiner's process loss / actual group productivity. The negative residual on the other side of the same ledger — coordination overhead, motivation loss, and social loafing pulling output below the best-member (or additive) baseline. The assembly bonus is only the positive interaction term; net group performance is the two competing terms summed. Confusing the bonus with a group's net output mistakes one signed term for the balance. Tell: are you naming the interaction-driven surplus above baseline (bonus) or the group's net performance after coordination costs are netted out (which may be a loss)?
Groupthink and group polarization. The dark face of the same interactive coupling (T4): interaction lets one member's error and confidence propagate into others', driving the group below even its average member. The assembly bonus credits interaction with catching errors; groupthink is interaction spreading them. The identical coupling produces both, so the sign is not guaranteed. Tell: is interaction letting members catch and correct one another's slips (bonus), or letting a confident wrong view cascade and suppress dissent (groupthink)?
Transactive memory system. A group's division of cognitive labor — who knows what, and who to consult — that lets a team collectively remember and retrieve more than any member. It is a durable structure for distributing knowledge across members, whereas the assembly bonus is a performance residual on a specific favorable task; a transactive memory system is one enabling condition for a bonus, not the surplus itself. Tell: are you describing a standing division of who-holds-what expertise (transactive memory), or a measured output exceeding the best solo performer on a task (assembly bonus)?
Social facilitation. Improved (or impaired) individual performance caused by the mere presence of others — an arousal/effort effect requiring no task interdependence, no information pooling, no combining of perspectives. The assembly bonus needs members to genuinely interact and combine partial insight; facilitation needs only an audience. Tell: is the gain the emergent product of members solving together (bonus), or a lone individual's effort rising because others are simply watching (social facilitation)?
Synergy / emergence (the parents). The substrate-neutral patterns the effect instantiates — supra-additive combination (synergy_and_antagonism) and qualitative novelty from interaction (emergence), which travel to ML ensembles, prediction markets, and coupled physical systems on their own terms. These are not confusable peers but the umbrella that owns the cross-substrate reach; invoking "an assembly bonus" outside an interacting human group borrows the supra-additive shape while dropping the human-cognition mechanism. Tell: if the substrate is not an interacting human group, the portable content is synergy_and_antagonism or emergence, not the assembly bonus effect.

Neighborhood in Abstraction Space¶

Assembly Bonus Effect sits in a moderately populated region (53^rd percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Social Perception & Self-Referential Bias (23 abstractions)

Nearest neighbors

Parkinson's Law of Triviality (Bikeshedding) — 0.87
Guess ⅔ of the Average — 0.84
Reliability Paradox — 0.83
Matching pennies — 0.83
Media Richness Theory — 0.83

Computed from structural-signature embeddings · 2026-07-12