Overview

The Strategic Interaction Framework (T18) is the territory of operations that model a situation as a game between two or more agents — where each agent’s choice depends on what they expect the other agents to choose — and analyze likely play, equilibria, signaling, and incentive design. The framework is game-theoretic in the formal sense: it names players with payoffs in their actual value terms (not what they claim to want), classifies the game on its four structural dimensions (timing, information, duration, sum), identifies equilibria with the solution method explicit, audits the credibility of declared threats and promises, tests at least one alternative game structure, and produces strategic recommendations grounded in the game structure rather than in heuristic intuition.

T18 runs two resident modes at Tier-2. strategic-interaction (the founder) covers two-to-n-player situations across the full equilibrium-concept toolbox (Nash, subgame perfect, Perfect Bayesian, repeated-game cooperation) and the credibility-audit toolbox (cheap talk, sunk costs, commitment devices, future shadow). mechanism-design (Mechanism and Incentive Analysis, added 2026-06-01) handles the case where hidden information (adverse selection — hidden type, pre-contract), hidden action (moral hazard — hidden effort, post-contract), and the incentive structure are the crux rather than the observable moves — naming and classifying the asymmetry, tracing the distortion, grounding the named mechanisms (winner’s curse, signaling, screening, principal-agent), and (when asked) designing the contract, auction, or rule set that aligns incentives while satisfying both the participation and incentive-compatibility constraints. The signaling game variant remains deferred per CR-6 and is surfaced as a flag.

The framework’s load-bearing intellectual content is the four-dimension game classification, the method-naming requirement, the credibility audit, and the alternative-structure test. The four-dimension classification determines which equilibrium concept applies — backward induction for sequential complete-information games, Perfect Bayesian for incomplete-information games, repeated-game folk theorems for indefinitely repeated games. Skipping a dimension is a hard verification failure because the wrong equilibrium concept produces the wrong answer. The method-naming requirement says equilibria must be asserted with the solution method named and traceable from players, payoffs, and classification. The credibility audit tests every declared threat or promise: cheap talk, backed by sunk costs, backed by a commitment device, or backed by future shadow — treating cheap talk as credible is a hard failure mode. The alternative-structure test guards against classification-locked analysis (what if this is actually repeated rather than one-shot, or has private information rather than complete?).

The framework deliberately resists three patterns. Hyperrationality drift — counteracted by the bounded-rationality (Simon) foundational lens. Stated-payoffs-as-actual-payoffs — counteracted by the requirement to state payoffs in each player’s actual value terms. Static-frame trap — one-shot equilibrium applied to what is genuinely a repeated game, missing cooperation the future shadow makes self-enforcing.

The framework answers questions like: What will they do if we do X? Is this threat credible? Are we in a repeated game or one-shot? What incentives are actually driving the other party — what they say, or something else?

Systemic context

Strategic Interaction is a Tier-3 cognitive territory in the Ora reasoning lattice, paired with three sibling territories at canonical disambiguation points. T13 ↔ T18 (Negotiation ↔ Strategic Interaction) — active negotiation guidance with integrative possibility is T13; game-theoretic equilibrium analysis with formal payoffs is T18. The two coexist because some negotiations are best understood as games (zero-sum bargaining with incomplete information) while others are best understood as integrative-search problems (interest-based, multi-issue, with creative problem-solving available). T2 ↔ T18 (Interest and Power ↔ Strategic Interaction) — descriptive interest and power analysis without modeling interaction is T2 (Cui Bono); equilibrium analysis with credibility assessment is T18. T7 ↔ T18 (Risk and Failure ↔ Strategic Interaction) — analyzing where the strategic structure could fail is T7 (pre-mortem-fragility / fragility-antifragility-audit); analyzing the equilibrium itself is T18. T18 routes within-territory to mechanism-design when the question becomes about hidden information, hidden action, or designing rules under which agents will produce a desired outcome (mechanism and incentive analysis); it still surfaces a deferred-mode flag for signaling when the question is specifically about asymmetric-information signaling games (Spence-style costly signals). The framework is a building block for downstream Decision Architecture when the strategic interaction is one component of a larger decision the user is evaluating.

Ora’s output

Example question: A larger competitor has just publicly threatened to undercut our pricing if we expand into their core market. They’ve said they’ll match any price we set, indefinitely. We’re considering whether to expand anyway. Is the threat credible, and what’s our best move?

Ora’s response (running T18 in strategic-interaction):

The situation is a two-player strategic interaction with a declared threat that determines the value of expansion. T18 is the right territory and strategic-interaction is the founder mode. I’ll work through the four-dimension classification, the equilibrium analysis with method named, the credibility audit on the declared threat, and at least one alternative-structure test.

Players and payoffs in actual value terms. You (smaller entrant): stated payoff is revenue from expansion; actual structure includes reputational gains from demonstrated willingness to compete with a larger incumbent (positive even in the loss case) minus opportunity cost of capital. Competitor (larger incumbent): stated payoff is core-market protection; actual structure includes profit from the core market minus the cost of price-matching across their entire revenue base (not just the slice you would take), plus deterrence value to other potential entrants, minus regulatory and antitrust attention. The deterrence value to other entrants is the load-bearing piece they are not stating out loud.

Four-dimension game classification. Timing: sequential — they made the threat first; you decide; they respond. Information: incomplete on both sides — neither knows the other’s pain threshold; perfect on actions when taken. Duration: repeated — you will interact in this and adjacent markets for years. Sum: mixed but lean positive-sum if expansion grows the total market.

Equilibrium analysis (method named). The relevant concept is Perfect Bayesian equilibrium for the sequential incomplete-information game, with repeated-game considerations layered on top. Method: pair strategies with beliefs about types; check incentive compatibility and Bayes consistency. The repeated-game layer adds future-shadow — strategies that sacrifice current-period payoff for reputation in future periods can be self-enforcing when the discount factor is high. The equilibrium turns on the threat’s credibility: if fully credible, your expected payoff is negative; if not credible, positive.

Credibility audit. The threat is “match any price, indefinitely.” Cheap talk? Yes partially — verbal statements have no cost. Sunk costs? No clear sunk-cost commitment visible. Commitment device? Possibly — public statement is a soft commitment device but moderate; market participants read “we’ll match prices” as competitive posturing more often than binding. Future shadow? This is the interesting one. Carrying out the threat is short-term costly but long-term valuable for deterring future entrants. They may do it partially or selectively to preserve deterrence value even if locally costly.

Net assessment: the threat is partially credible — they will likely respond with price competition but probably not sustained indefinite matching. Likely pattern: sharp early cuts (3-9 months), then de-escalation as cost becomes visible against limited share recovery. The “indefinite” framing is overstated; the future-shadow value is real but bounded.

Alternative-structure test. What if this is one-shot rather than repeated? Then the threat is closer to cheap talk and your expected payoff rises. What if there is a third potential entrant watching? Then the competitor’s incentive to demonstrate credibility rises sharply and the threat becomes more credible.

Strategic recommendations grounded in the game structure. (1) Expand, but stage entry into a smaller adjacent segment first to limit the price-war attack surface. (2) Build your own commitment device — public statements, dedicated capacity, fixed-cost investments visible to the competitor — to raise your retreat cost and shift their calculation. (3) Position your entry to draw regulatory or antitrust scrutiny on aggressive price-matching; their actual payoff includes that cost. (4) Monitor for third-entrant signals; if one moves first, the threat becomes more credible. (5) Plan for the de-escalation handshake; do not over-invest past the point they signal de-escalation.

Confidence per finding. High on the four-dimension classification and equilibrium concept. Medium-high on the credibility assessment — the partial-credibility verdict is robust; the specific 3-9 month timing is an estimate from similar response patterns. Medium on recommendations — they follow from the analysis but execution depends on resource and cost-structure detail I do not have.

The threat is partially credible (not fully, not cheap talk); the equilibrium turns on a Perfect Bayesian assessment with repeated-game layering; the alternative-structure tests surface two ways the analysis could be wrong; the recommendations follow from game structure rather than competitive-strategy heuristics.

Commercial AI comparison

Comparison content auto-populates when the comparison-refresh framework runs against this question. Drafters do not author this section.

Brief comparison commentary

Auto-populates with the comparison content above.

How to use this framework

You can run the Strategic Interaction pattern with any AI of your choice. The composition is single-pass through the strategic-interaction mode.

The prompt:

[Paste the framework specification.]

Run strategic-interaction on this situation.

Situation: [Plain prose description of the multi-agent strategic situation.]

Players (optional): [Named players with their stated and (if you suspect) actual payoffs.]

Declared threats or promises (optional): [Specific commitments either side has made — these get audited for credibility.]

Time horizon (optional): [Whether you expect to interact with these players once, repeatedly over months, or indefinitely.]

The AI runs the four-dimension classification, names the equilibrium method explicitly, audits credibility of any declared threats or promises, tests at least one alternative game structure, and produces specific strategic recommendations grounded in the game structure. The output has six required sections.

For best results:

  1. Name the actual payoffs, not the stated ones. What the players will say they want is often different from what they will play for. The framework asks you to surface the unstated drivers — the deterrence value to third parties, the reputational stakes, the regulatory cost of certain moves, the long-term positioning that explains short-term irrationality.
  2. Be honest about the duration. One-shot vs. repeated is the dimension that most often gets misclassified. If you expect to interact with these players for years, even if this specific decision feels one-shot, the repeated-game framing applies.
  3. Don’t suppress the alternative-structure test. The framework will surface the most plausible misclassification possibility (e.g., “what if this is actually a multi-player game rather than two-player?”). Take the alternative seriously rather than dismissing it as edge-case thinking.
  4. Read the credibility audit carefully. The most common failure mode in strategic analysis is treating cheap talk as commitment. The audit’s purpose is to break that habit; resist the urge to assume the other side will do what they say.

The framework is deliberately tool-agnostic. The four-dimension classification, the equilibrium-method-naming requirement, the credibility audit, and the alternative-structure test are conceptual disciplines that survive the lift to any environment. The output is decision-tree-friendly when the game is sequential and matrix-friendly when simultaneous.

Other examples

  • A negotiation with a coalition counterpart — A complex deal where the counterpart is three internal stakeholders. Multi-player game: each stakeholder a separate player with separate payoffs; sequential, incomplete-information, repeated, mixed-sum. Equilibrium identifies the pivotal coalition member and their likely reservation point; credibility audit surfaces which positions are posture and which are binding. Recommendation: structure the proposal to address the pivotal player’s actual payoff, not the coalition’s stated position. Demonstrates negotiation preparation handing off to T13.

  • A signaling-game variant on hiring competition — Multiple employers, sought-after candidate, asymmetric information on both sides. Strategic-interaction with deferred-mode flag for signaling. Equilibrium identifies cost-of-signal distinguishing serious from filler offers (response speed, customization, willingness to accept counters); credibility audit surfaces which “best and final” framings are binding. The deferred-mode flag notes a dedicated signaling mode would handle this more directly. Demonstrates deferred-mode flag handling.

  • A repeated-game cooperation analysis on a partnership — Two partner companies on a multi-quarter joint product, each with incentives to under-invest. Classified as repeated, incomplete-information, positive-sum. Equilibrium concept: repeated-game cooperation (Axelrod) — identify discount factor at which cooperation is self-enforcing; identify the implicit trigger strategy. Recommendations: build observability into the partnership; align milestone-and-payment cycles; pre-agree on defection responses. Demonstrates repeated-game analysis with static-frame trap avoided.

Citations

The Strategic Interaction Framework draws on the foundational and extended game-theory literature. Von Neumann and Morgenstern’s Theory of Games and Economic Behavior (1944) is the foundational text. Nash’s “Equilibrium Points in N-person Games” (1950) introduced the Nash equilibrium concept. Schelling’s The Strategy of Conflict (1960) and Micromotives and Macrobehavior (1980) provide the credibility-audit framework — commitment devices, focal points, sunk costs, future shadow. Axelrod’s The Evolution of Cooperation (1984) provides the iterated-prisoner’s-dilemma results that ground repeated-game analysis. Fudenberg and Tirole’s Game Theory (1991) is the standard graduate-level reference for the full equilibrium-concept toolbox (Nash, subgame perfect, Perfect Bayesian, evolutionary stable). Myerson’s Game Theory: Analysis of Conflict (1991) provides the mechanism-design foundations for the resident mechanism-design mode (Mechanism and Incentive Analysis, added 2026-06-01), alongside Akerlof’s “The Market for ‘Lemons’” (1970) on adverse selection, Spence’s “Job Market Signaling” (1973), and Holmström’s “Moral Hazard and Observability” (1979).

The bounded-rationality foundation draws on Simon’s “A Behavioral Model of Rational Choice” (Quarterly Journal of Economics, 1955) and The Sciences of the Artificial (1996). The hyperrationality-trap is the failure mode bounded rationality guards against — equilibrium analyses that assume perfect rationality without bounded-rationality assessment misrepresent how real agents will play. The credibility audit and the alternative-structure test partly address this; explicit bounded-rationality flagging is the additional discipline.

The framework is single-author and originated 2026-05-01 from the T18 territory consolidation. The current population is two resident modes — strategic-interaction (the founder) and mechanism-design (Mechanism and Incentive Analysis, added 2026-06-01, un-deferring the CR-6 expansion candidate) — with the signaling game variant still deferred per CR-6. No architecture-level open debates currently apply; the hyperrationality-vs-bounded-rationality choice is treated within strategic-interaction, and the structural-vs-formal mechanism-design boundary within mechanism-design.

Downloads

  • Framework specification (PDF) — link to ora-ai.org canonical artifact when published
  • Framework specification (plain text) — link to ora-ai.org canonical artifact when published
  • Full white paper (PDF) — link when published