Evidence is the layer where Interaction makes contact with strategic reality. It defines what forms of interdependent behavior can be observed, how they are measured, how incentives and information flows manifest empirically, and how reliable those observations are. Every empirical claim about markets, negotiations, auctions, contracts, networks, or mechanisms rests on the integrity of this layer. Evidence specifies the observable signals produced by strategic environments—prices, bids, quantities, communication patterns, timing of moves, belief updates, institutional outcomes—and the measurement systems that translate these signals into analyzable data. It includes the operational definitions that tie concepts like “market power,” “information asymmetry,” “strategic response,” or “equilibrium behavior” to concrete empirical procedures; the protocols that govern how strategic data are collected in the field, laboratory, or simulations; the formats in which raw interactional behavior appears; and the calibration, identification strategies, and error analysis required to ensure that observed behavior reflects genuine strategic forces rather than noise or mismeasurement. This section establishes Interaction’s empirical backbone: the standards and constraints that allow its observations to be credible, reproducible, comparable across settings, and meaningful for interpreting how real-world agents behave in interdependent systems.
Interaction (Markets, Strategy & Mechanisms) – Evidence – SAT
| Element | Interaction (Markets, Strategy & Mechanisms) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Scope Category | 2.1 Observable Phenomena | 2.2 Measurement Systems | 2.3 Operational Definitions | 2.4 Data Acquisition | 2.5 Data Character & Format | 2.6 Reliability & Calibration | ||||||
| Sub-Item | Interaction – Observables | Interaction – Detection Limits | Interaction – Units | Interaction – Instruments | Interaction – Definitions | Interaction – Procedures | Interaction – Protocols | Interaction – Sampling | Interaction – Data Types | Interaction – Resolution | Interaction – Calibration | Interaction – Error CharacterizationInteraction – Error Characterization |
2.1 Observable Phenomena
Observable Phenomena define the empirical interface of Interaction—what strategic signals real environments produce and what the discipline can reliably detect. In multi-agent systems, observables are the measurable manifestations of interdependent behavior: bids in auctions, posted prices, traded quantities, negotiation offers, timing of moves, communication patterns, contract terms, rule compliance or violation, strategic experimentation, belief-revealing actions, market adjustments, and the institutional outcomes generated by mechanisms. These signals reveal how agents respond to incentives, how information is transmitted or withheld, and how rules shape equilibrium and non-equilibrium behavior.
Detection limits mark the boundaries of what Interaction can currently observe: unspoken beliefs, private information, uncommunicated intentions, latent coordination attempts, and off-path strategic reasoning often remain inaccessible except through indirect inference. Together, observable phenomena and detection limits set the empirical horizon of the field—the line between what can be evidenced through real markets, experimental data, field observations, and mechanism-level logs, and what must remain theoretically posited until measurement systems advance. They define what strategic behavior can be directly witnessed, what must be inferred, and what cannot yet be resolved.
Observables are the measurable signals through which strategic interaction becomes empirically visible. They are the concrete manifestations of interdependent behavior that researchers can record, quantify, or infer from data. In markets and strategic environments, observables include prices, bids, quantities traded, contract terms, waiting times, communication patterns, timing of moves, acceptance or rejection behavior, compliance rates, deviations from prescribed rules, network connections, and the realized allocations or payoffs that result from mechanisms or negotiations. These signals form the empirical bridge between theory and reality: while models may posit beliefs, incentives, private information, or equilibrium reasoning, only the components that produce detectable effects can be directly studied.
Identifying observables focuses empirical research on aspects of interaction that can be captured through data—whether in field markets, laboratory experiments, platform logs, or institutional records. It also highlights the limits of current measurement: private beliefs, internal reasoning processes, or unobserved negotiations may be theoretically central but observationally inaccessible, motivating the development of inference techniques or new modes of data collection. In Interaction, observables are the raw empirical inputs that reveal how real agents respond to incentives, how information flows, and how rules shape outcomes; they are the evidentiary foundation on which hypotheses about strategic behavior stand or fall.
Detection Limits in Interaction recognize that the empirical tools used to observe strategic environments have finite resolution—there are aspects of interdependent behavior that cannot be directly seen, measured, or inferred beyond certain thresholds. Unlike physical sciences where limits arise from optical or instrumental constraints, Interaction’s detection limits stem from informational opacity, institutional design, data granularity, and the strategic concealment of private variables.
Some strategic signals are observable—bids, prices, quantities, messages, timing—but others remain below the detection threshold: private beliefs, off-equilibrium reasoning, unrecorded negotiations, hidden deviations, tacit collusion, or latent expectations. Even when institutions generate data, resolution limits may prevent distinguishing whether behavior is driven by strategic motives or noise. In platform markets or auctions, for example, logs may capture actions but not the unobserved counterfactual strategies or belief updates that generated them.
Understanding detection limits is essential for correct interpretation: failure to observe collusion, signaling, coordination, or incentive responses does not mean they are absent—only that they fall below the observational power of the current measurement system. These limits also guide empirical design, pushing researchers to develop better identification strategies, richer experimental protocols, or more transparent mechanisms that surface deeper strategic signals.
In Interaction, acknowledging detection limits ensures empirical humility and prevents the field from overinterpreting sparse or noisy data. It frames conclusions with the understanding that evidence is bounded by what institutional, observational, and inferential tools can reveal—and motivates continual innovation to push those boundaries outward.
2.2 Measurement Systems
Measurement Systems specify how Interaction converts strategic observables into analyzable, quantitative form. Because the domain studies interdependent human and institutional behavior rather than physical signals, its “units” and “instruments” consist of standardized economic quantities, structured data protocols, institutional records, platform logs, experimental frameworks, and inferential tools that translate strategic actions into comparable measurements. Units include prices, quantities, bids, payoffs, probabilities, belief indices, strategy frequencies, and compliance rates—scales that allow different observations of strategic behavior to be meaningfully aligned. Instruments include market data systems, auction logs, negotiation transcripts, experimental economics platforms, mechanism-level telemetry, network-mapping tools, and econometric or game-theoretic identification methods that extract structural information from observed interactions.
Together, these systems constitute the operational machinery through which Interaction produces empirical claims. They determine what strategic patterns can be captured with precision, what forms of interdependence can be quantified, and what aspects of behavior remain too coarse, noisy, or indirect to analyze. Measurement Systems thus constrain and enable the domain: they shape data quality, comparability across environments, reproducibility of findings, and the reliability of inferences drawn about incentives, beliefs, equilibria, and institutional effects.
Units in Interaction are the standardized quantitative scales that express strategic and economic measurements in a consistent, interpretable form. Unlike physical sciences that rely on meters or seconds, Interaction uses economic and behavioral units that describe interdependent behavior: currency units for payoffs and prices, quantity units for traded goods or effort levels, probability units for beliefs, time units for timing of moves or adjustment speeds, frequency units for strategy distributions, and rate units (such as markup percentages or bid increments) that capture relative strategic intensity. These standardized units form a shared measurement language across markets, experiments, and mechanism implementations, allowing strategic outcomes to be compared across contexts.
Units ensure clarity and prevent misinterpretation—data expressed in different currencies, probability scales, or time resolutions must be properly converted, or analyses of incentives, equilibrium behavior, or efficiency may be distorted. They also underpin dimensional consistency in models: equilibrium conditions must align payoff units with probability units; mechanism rules must use compatible scales for bids, transfers, and allocations; and empirical estimates must map behavioral responses onto coherent units of measurement.
In Interaction, a “measurement without a unit” is effectively unusable—whether examining bids in an auction, prices in a market, or belief updates in a signaling game. Units make the domain empirically portable, comparable across studies, and rigorously interpretable.
Instruments in Interaction are the procedural and technological tools that translate strategic observables into recorded data. Because the domain studies human and institutional behavior rather than physical signals, its instruments are not telescopes or microscopes but market data feeds, platform logs, auction interfaces, contract repositories, negotiation transcripts, behavioral experiments, surveys, monitoring systems, and mechanism-embedded telemetry. These tools capture bids, prices, quantities, communication events, timing patterns, rule compliance, allocation outcomes, and other measurable expressions of interdependent behavior.
Each instrument has capabilities and limitations. Market data feeds may show prices but not private beliefs; auction platforms record bids but not off-platform collusion; laboratory experiments reveal controlled strategic responses but may miss field-level complexities; surveys capture stated preferences that may diverge from revealed behavior. Calibration in this context means validating data pipelines, ensuring timestamps are accurate, verifying that transaction logs reflect true behavior, and checking that experimental protocols produce stable, replicable results.
The credibility of empirical claims in Interaction depends on the quality of these instruments. Poorly designed data systems can obscure strategic signals; miscalibrated logs can distort timing inferences; incomplete institutional records can understate true incentive effects. Different aspects of strategic behavior require different instruments: field experiments for causal inference, game logs for strategy reconstruction, audits for compliance detection, and communication traces for information-flow analysis.
In Interaction, the instrument chosen shapes not only the data obtained but also the phenomena that become visible—just as a high-speed camera reveals motion otherwise unseen, institutional and platform instruments reveal strategic dynamics that would be invisible without systematic recording.
2.3 Operational Definitions
Operational Definitions in Interaction bind the domain’s conceptual vocabulary—strategic behavior, incentives, information asymmetry, market power, equilibrium response, mechanism performance—to the empirical procedures used to measure them. They specify what a concept means in observable terms and how that meaning is produced in practice. For example, “market power” might be operationalized through markup measures or residual demand elasticities; “strategic signaling” through identifiable shifts in action patterns conditional on information; “equilibrium behavior” through convergence of strategies or stable best-response cycles; “compliance” through documented adherence to mechanism rules; and “information asymmetry” through measurable differences in what agents observe or infer.
Procedural clarity ensures that these definitions translate into repeatable workflows: how bids are coded, how belief updates are inferred, how timing thresholds are identified, how deviations from equilibrium predictions are classified, or how allocation efficiency is computed. Without explicit operational definitions, Interaction’s core constructs would remain theoretical abstractions detached from empirical reality.
Together, these definitions eliminate ambiguity, enforce reproducibility across markets, experiments, and institutional settings, and ensure that every theoretical concept corresponds to an observable or inferable procedure. They anchor the empirical meaning of strategic phenomena and make measurement possible, comparable, and scientifically defensible.
Operational definitions in Interaction tie the meaning of strategic concepts directly to the procedures that detect or measure them. Because strategic behavior is abstract—beliefs, incentives, deviations, information asymmetry, equilibrium, market power, collusion—each must be defined in terms of concrete, observable operations so that researchers know exactly how to recognize, code, or calculate them.
For example:
- “Market power” may be operationalized as a markup over marginal cost or as residual demand elasticity.
- “Strategic deviation” may be defined as an action that violates a mechanism’s prescribed strategy or departs from equilibrium best responses.
- “Information asymmetry” may be operationalized as measurable differences in what agents observe, such as private-value draws in auctions or differential access to platform data.
- “Signaling behavior” may be defined as statistically detectable action shifts conditional on private information.
- “Equilibrium behavior” may be defined through convergence patterns, stability criteria, or alignment with predicted best-response mappings.
Operational definitions remove ambiguity: the concept means precisely the output of the specified measurement or inference procedure. They make empirical work reproducible—any researcher following the same protocol should classify strategies, signals, or market outcomes in exactly the same way. They also ground hypothesis testing in measurable terms: theories of signaling, screening, collusion, or bidding behavior only make contact with evidence when their constructs have operational definitions that map cleanly onto observable data.
In Interaction, a strategic concept is only as scientifically meaningful as the procedure that detects it. Operational definitions ensure that the domain’s abstract ideas remain anchored to observable, codifiable reality.
Procedural Clarity in Interaction requires that every operational definition be accompanied by explicit, reproducible steps detailing how strategic data are obtained, coded, or inferred. Because Interaction studies behaviors that unfold through choices, incentives, and information flows, clarity about procedures is essential: researchers must specify precisely how bids are recorded, how communication events are classified, how deviations are detected, how beliefs are inferred, how allocation outcomes are computed, and how market or mechanism logs are processed.
Step-by-step methods anchor the domain’s empirical validity. For example:
- To measure bid shading, a procedure may specify how to estimate private-value distributions and compute deviations from truthful bidding.
- To measure strategic delay, a protocol may define the timing window, data source, and threshold that classify an action as delayed.
- To measure compliance, the method must detail how platform rules are encoded, how violations are detected, and how ambiguous behaviors are resolved.
- To infer beliefs or information sets, the procedure may specify which modeling assumptions or econometric tools reconstruct hidden strategic variables.
Procedural clarity ensures that any researcher following the same protocol will reproduce the same classification of strategies, the same identification of signals, and the same measurement of incentive responses. It transforms conceptual definitions—such as “collusion,” “screening,” or “strategic adjustment”—into concrete, testable operations. In a domain where unobserved reasoning, private information, and institutional complexity can cloud empirical interpretation, procedural clarity is what prevents ambiguity and secures reproducibility.
2.4 Data Acquisition
Data Acquisition in Interaction governs how empirical evidence about strategic behavior is obtained. Because the domain studies interdependent decisions occurring in markets, negotiations, auctions, platforms, and institutional settings, acquisition must cope with decentralized information, strategic concealment, heterogeneous agents, and noisy or incomplete records. Protocols specify the standardized procedures through which interactional data are collected—field experiments, laboratory designs, platform telemetry, institutional audits, market data feeds, contract repositories, or structured observation of strategic events. Sampling determines which agents, interactions, or institutional conditions are measured, and how representative they are of the strategic environment as a whole.
Together, these components shape the empirical foundation of Interaction: what data are captured (bids, prices, actions, signals, timing, communication), how consistently they are recorded, how systematically environments are observed, and how confidently the resulting evidence can be generalized beyond the measured context. In a domain where strategic behavior can vary dramatically across settings, data acquisition determines not only what is known, but how reliable that knowledge is, and whether patterns observed in one strategic environment meaningfully extrapolate to another.
Protocols in Interaction specify the methodical, standardized procedures used to collect data on strategic behavior. Because interactional evidence comes from markets, mechanisms, negotiations, platforms, and controlled experiments, protocols must detail how agents are sampled, how environments are initialized, how actions are recorded, how instruments (platform logs, auction interfaces, experimental software, institutional data systems) are calibrated, and how observations are coded. They define the exact sequence of steps through which bids are collected, communication events are captured, rule compliance is monitored, payoffs are computed, timestamps are validated, and ambiguous behaviors are classified.
These protocols ensure reproducibility: any researcher following the same procedures should obtain comparable measurements of strategies, signals, and outcomes. Standardization reduces confounding factors—such as inconsistent instructions in an experiment, differing market conditions across field sites, or variations in platform data extraction—so that observed patterns reflect genuine strategic forces rather than artifacts of collection. Protocols also make the methodology transparent: others can scrutinize, replicate, or refine the procedure to strengthen reliability.
In Interaction, where behavior is sensitive to incentives, framing, and environmental structure, well-defined protocols are essential. They enforce empirical discipline, minimize noise from data-gathering idiosyncrasies, and allow strategic evidence to be compared across sessions, markets, mechanisms, or institutions.
Sampling in Interaction concerns how researchers select the subset of agents, interactions, markets, mechanisms, or institutional contexts from which strategic data are drawn. Because it is rarely feasible to observe every participant or every instance of interdependent behavior, sampling rules determine which interactions enter the dataset and how representative they are of the broader strategic environment. Methods may include random sampling of agents in a market, stratified sampling across different mechanism types, systematic sampling of negotiation rounds or bidding events, or temporal sampling of platform activity at specified intervals.
Representativeness is crucial. Strategic behavior can vary sharply with incentives, information conditions, institutional rules, or competitive intensity. A sample drawn only from high-frequency traders, peak-demand auctions, cooperative negotiation contexts, or well-functioning institutions may misrepresent the full distribution of strategic responses. Biased sampling can obscure collusion, exaggerate rationality, understate frictions, or distort estimates of mechanism performance.
Carefully designed sampling strategies—paired with transparent reporting—allow researchers to quantify uncertainty, evaluate generalizability, and determine how far findings extend beyond the observed subset. In Interaction, where environments differ widely across markets, mechanisms, and contexts, sampling design directly shapes the scientific validity of empirical conclusions.
2.5 Data Character & Format
Data Character & Format in Interaction specifies the structural form in which strategic observations are captured and the granularity at which interdependent behavior is preserved. Because Interaction studies choices, incentives, information flows, and institutional outcomes, its data take characteristic shapes: time-stamped action sequences, bid logs, price histories, allocation records, communication transcripts, network adjacency structures, strategy-frequency matrices, event logs from mechanisms or platforms, and outcome distributions from experiments or field interventions.
Data format determines whether the evidence appears as:
- time series of market adjustments,
- panel data on agents across repeated interactions,
- discrete event logs from auctions,
- communication or message sequences,
- network graphs encoding relational structure,
- mechanism-level rule applications,
- payoff or allocation matrices,
- coded behavioral classifications.
Resolution determines how much detail survives: whether bids are recorded to the millisecond or rounded to seconds; whether message content is captured verbatim or coded into categories; whether market transitions are logged continuously or in periodic snapshots; whether strategic actions are grouped or disaggregated at the individual level.
Together, Data Character & Format govern what patterns can be detected and which analyses are valid. Fine-resolution logs may reveal subtle signaling or timing strategies; coarse-resolution data may obscure them. Structural formats determine whether equilibrium testing, behavioral classification, network inference, or mechanism-performance analysis is feasible. Ultimately, the format and resolution of strategic data shape how faithfully evidence reflects the underlying interdependent phenomena and what kinds of insight Interaction can extract from them.
Data Format in Interaction refers to the structural shape in which strategic observations are captured and recorded. Because the domain studies interdependent behavior—bidding, pricing, communication, negotiation, allocation, rule-following—its data appear as structured sequences of actions, time-indexed records of market states, logs of mechanism events, transcripts of communication, or coded representations of behavioral outcomes. The form matters because it determines what kinds of patterns can be detected and what analytical tools are appropriate. A continuous stream of bids reveals adjustment dynamics that a coarse summary would erase; a transcript preserves signaling structure that a simple action label cannot convey; a detailed log of rule applications makes incentive effects empirically tractable in ways that aggregated outcomes cannot. Choosing the right format ensures that strategic detail is neither lost nor misrepresented and that the resulting evidence supports the kinds of inferences Interaction requires. When researchers specify a data format, they define the interpretive frame of the entire study—what variation is preserved, what structure is visible, and how faithfully the data reflect the underlying interactive phenomena.
Resolution in Interaction refers to the level of detail at which strategic behavior is captured—how finely the data record timing, magnitude, communication content, institutional responses, or patterns of adjustment. High temporal resolution might document bids or messages millisecond by millisecond, revealing coordination attempts or rapid strategic shifts that disappear in coarser logs. High informational resolution may capture the exact content of communications rather than collapsing them into categories, allowing subtler signaling structures to appear. Conversely, low resolution aggregates actions across time, players, or contexts, obscuring the fine-grained incentives and dependencies that define strategic interaction. Resolution determines not only what can be seen but what can be inferred: collusion, signaling, experimentation, or off-equilibrium reasoning may be detectable only when the data preserve sufficient granularity. At the same time, excessively fine resolution may introduce noise or overwhelm analysis without adding insight. Researchers in Interaction must choose a resolution that preserves the essential dynamics of interdependent behavior while avoiding unnecessary complexity. When interpreting evidence, the resolution of the data sets the limits of what claims can be made; phenomena invisible at the chosen granularity cannot be ruled out, only noted as undetectable within those constraints.
2.6 Reliability & Calibration
Reliability & Calibration in Interaction secures the credibility of the data on which all strategic analysis depends. Because the domain observes behavior filtered through platforms, institutions, experiments, and market infrastructures, calibration means ensuring that these recording systems accurately reflect the actions and outcomes they claim to measure. Timestamps must align with real event order, bid logs must capture genuine submissions rather than artifacts of software or network delay, contract records must reflect true executed terms, and experimental interfaces must implement incentives exactly as specified. Reliability concerns the stability and repeatability of these measurements across sessions, markets, or institutional contexts; a mechanism log that records bids inconsistently, or a negotiation transcript system that misses certain messages, undermines any inference about strategic behavior. Calibration prevents systematic drift in data—misaligned clocks, rounding errors, truncated communication records, or misclassified actions—while reliability ensures that repeated observations under equivalent conditions yield equivalent data. Error analysis quantifies the noise, bias, and uncertainty that remain, acknowledging that in strategic environments some elements—beliefs, private information, covert coordination—are never observed directly. Together, reliability and calibration anchor Interaction’s empirical foundation, ensuring that its measurements are not merely collected but verified, stable, and defensible, so that any conclusions drawn about incentives, information flow, or equilibrium behavior rest on evidence that can be trusted.
Calibration in Interaction involves aligning the systems that record strategic behavior with known reference points so that their outputs reflect true actions, timing, and institutional outcomes rather than artifacts of the recording process. Because Interaction relies not on physical instruments but on platforms, experiments, market infrastructures, and institutional logs, calibration means verifying that bid timestamps correspond to actual submission times, that mechanism rules are executed exactly as specified, that payoff calculations match true incentive structures, that communication records capture all relevant messages, and that market feeds report prices and quantities without distortion. Without calibration, recorded data may drift—clocks may desynchronize across servers, bids may be rounded or truncated, messages may be lost in transmission, or platform interfaces may apply hidden transformations. Such drift systematically biases the evidence on which strategic inference depends. Regular calibration routines—synchronizing clocks, validating log integrity, comparing platform outputs against controlled test cases, cross-checking experimental payouts, and verifying that rule execution matches design—anchor the data to reality. Calibration ensures that what appears as a strategic delay is not a timestamp error, that what looks like a deviation from equilibrium is not a logging glitch, and that the numerical values used in analysis truly correspond to the incentives faced by agents. It provides confidence that the evidence reflects actual strategic behavior within known error bounds, making it an indispensable component of reliable empirical practice in Interaction.
Error Analysis in Interaction addresses the unavoidable imperfections in data arising from the complexity of observing strategic behavior. Even with well-calibrated platforms, consistent logging systems, and controlled experimental designs, evidence reflects both random noise and systematic distortion. Random errors appear as unpredictable variation in bids, timing, communication patterns, or recorded quantities—arising from network latency, stochastic decision jitter, environmental variation, or the inherent volatility of strategic adjustment. Systematic errors arise when platform interfaces introduce consistent delays, when institutional logs omit certain categories of actions, when coding schemes misclassify behavior, or when incentives are implemented imperfectly. Identifying and quantifying these errors is essential because strategic phenomena—collusion, signaling, equilibrium deviations, learning dynamics—often manifest as subtle patterns that can be easily obscured or mistaken when uncertainty is high.
Error analysis in Interaction requires assessing the magnitude of timing uncertainty, classification ambiguity, missing data, aggregation distortion, and institutional misreporting, and expressing these uncertainties through appropriate measures such as confidence intervals, robustness checks, sensitivity analyses, or alternative coding schemes. By distinguishing noise from systematic bias, researchers can correct distortions where possible and, at minimum, remain transparent about the limits of inference. This discipline is indispensable: a strategic effect cannot be trusted if it lies within the error margin of the measurement process, and an observed deviation from equilibrium cannot be interpreted if uncertainty overwhelms the signal. Incorporating rigorous error analysis prevents overconfident conclusions, clarifies the evidentiary strength of findings, and anchors empirical claims to the actual resolution and reliability of the data.