Overview

The Interest and Power Analysis Framework (T2 in Ora’s territory map) handles questions that are not about whether an argument holds but about who is served by it — who benefits, who pays, who has power to shape it, whose voices are absent, and how interest structures explain what is observed. It is the territory that activates when the user’s instinct says something is being framed as natural or inevitable when it is actually serving someone’s interests, or when the policy that sounds objective seems to have predictable beneficiaries, or when the question is whose perspective is being treated as the default.

The framework runs in four modes. Cui-bono is the descriptive mode — a read on who benefits from a situation, decision, or institutional position. The mode names the institutional author of the position (whose office, agency, or coalition produced it), states the rationale on offer, traces concrete distributional impact (money flows, power-position changes, time-and-attention captures, narrative-control gains), constructs the alternative design that would emerge from the opposite constituency’s interests with equal technical sophistication, applies FGL (Fear, Greed, Laziness) symmetrically across constituencies, and separates legitimate value from distributional overlay. Boundary-critique is the critical-stance audit using Ulrich’s twelve boundary categories in is/ought form — what the artifact takes as given vs. what it would take as given if those affected (especially affected-but-not-involved parties) had standing. Wicked-problems is the molecular composition for tangled problems where every solution makes things worse somewhere else. Decision-clarity is the molecular composition that produces a decision-clarity document for a third-party decision-maker.

The framework’s load-bearing intellectual content is the symmetric FGL discipline, the counterfactual-design construction, the involved-vs-affected distinction (Ulrich’s core asymmetry), and the legitimate-value separation. Symmetric FGL says Fear/Greed/Laziness must be applied across all constituencies, not selectively against the disfavoured side. Counterfactual-design construction says the alternative design from the opposite constituency’s interests must be built with equal technical sophistication; a strawman alternative reduces the analysis to confirmation. The involved-vs-affected distinction is Ulrich’s central move: those involved in the system’s design and benefit are not the same set as those affected by it; the asymmetry is the locus of boundary critique. Legitimate-value separation says distributional analysis must distinguish the legitimate value the position serves from the distributional overlay it carries.

The framework is honest about what it is not. It is not argument-soundness analysis (T1 — does the argument itself hold). It is not active negotiation guidance (T13 — how to bargain). It is not multi-explanation evidence adjudication (T5). It is not causal investigation (T4). When the question is “who benefits” rather than “is this true” or “what should I do,” the framework activates; otherwise, cross-territory dispatch routes elsewhere.

The framework includes failure modes specific to interest analysis — the conspiracy trap (over-attribution of coordinated intent where uncoordinated incentive alignment is sufficient), the cynicism trap (treating all stated rationales as cover for distributional self-interest), the mirror trap (projecting the analyst’s own motivational structure onto all constituencies), the frame-bounded blindness (missing parties not visible from the artifact’s own framing), and the asymmetric FGL failure (applying the cynical motivational analysis selectively). Naming the failure modes makes them addressable in the specific mode-runs that surface them.

The framework answers questions like: This policy sounds objective but I suspect it isn’t — who actually benefits? I think the framing of this leaves people out — whose voices are absent? Every solution we try makes it worse somewhere else — what are we missing? I need a decision document for a board, not an exploratory analysis — can you produce something decision-shaped?

Systemic context

Interest and Power Analysis is the interest-structure territory of Ora’s territory map (T2, super-cluster A — Argument and Reasoning). It sits adjacent to T1 (Argumentative Artifact), T8 (Stakeholder Conflict), T13 (Negotiation), and T18 (Strategic Interaction). Cross-territory dispatch routes to T1 first when both interest-pattern and argument-soundness analyses are needed, to T8 when focus shifts to conflict-structure mapping, to T13 when active negotiation guidance is the next step, and to the cross-territory stakeholder-mapping mode (which lives in T8) when interest analysis requires multi-party complexity that exceeds Cui Bono. The decision-clarity molecular has a paired execution framework (Framework — Decision Clarity Analysis, restructured per Decision H) — the molecular spec lives in T2’s framework; the decision-maker-output-shaped restructure lives in the paired framework.

Ora’s output

Example question: A new “ethics in AI” certification program is being proposed by an industry consortium. The certification would require AI products to undergo a $30,000 audit by an approved third-party auditor. The proposal is being marketed as protecting consumers and raising the standard of AI ethics. Run cui-bono on this.

Ora’s response (running Interest and Power Analysis in cui-bono mode):

The question is who benefits from a proposal that is being marketed as objectively pro-consumer; the user’s framing already signals suspicion about the marketing/distribution alignment. I’ll work the case through the seven required sections of cui-bono.

Section 1 — Institutional authorship. Industry consortium of established AI companies. The named consortium is the proximate author; the load-bearing fact is which companies are members and which are not. If the consortium is composed of well-resourced incumbents and excludes early-stage startups and academic groups, that fact alone shifts the cui-bono read significantly.

Section 2 — Stated rationale on offer. “Protecting consumers and raising the standard of AI ethics.” The rationale is consumer-protective and ethics-quality-improving; the distributional structure of the rationale (who pays, who benefits) is not addressed in the rationale itself. The absence is itself a data point — the rationale presents the proposal as if it had no distributional structure.

Section 3 — Distributional impact.

Money flows. $30K per product flows from AI companies to approved third-party auditors. Cost is a fixed fee, not scaled to revenue.

Market-position changes. $30K is trivial for a $100M-revenue company with three products (0.09% of revenue), potentially fatal for a 5-person startup at $200K with one product (15% of revenue). The cost is regressive; the certification functions as a barrier to entry against startups and individual developers.

Power-position changes. The consortium becomes the certifying body; the approved-auditor list is a consortium-controlled chokepoint; the standard is the consortium’s standard. Companies refusing the certification are positioned as either ethically suspect or unable to afford ethics — both narratively unfavourable.

Narrative control. “Certified ethical” and “uncertified” become category markers; the consortium controls the boundary; public discourse about AI ethics becomes shaped by the consortium’s frame.

Section 4 — Alternative design from opposite constituency. What design would emerge if smaller players, academics, and consumer-advocacy groups had equal authoring standing? A scaled-fee structure (e.g., 0.5% of product revenue, capped at $30K for large products, reduced to $500 for revenue under $1M); an independent certifying body with cross-constituency governance; an open standard the consortium and its competitors both reference; a self-attestation pathway for early-stage products with audit triggered by user complaint rather than by launch. The alternative produces the same consumer-protection outcome without the regressive-distribution structure and the consortium-control structure. The fact that this alternative is technically straightforward but not on the table is itself diagnostic.

Section 5 — FGL motivational analysis (symmetric).

Consortium members. Greed: revenue-flow to approved auditors; market-position consolidation. Fear: regulatory action imposing a less-favourable standard. Laziness: certification is the easiest “we did something on AI ethics” path that doesn’t require restructuring practices.

Smaller players. Fear: being positioned as ethically suspect; being priced out. Greed: avoiding the cost. Laziness: not engaging with the political process.

Approved auditors. Greed: $30K per audit. Fear: losing approved status. Laziness: relying on the consortium’s standard rather than independent expertise.

Consumer-advocacy groups. Greed: alignment for access. Fear: marginalization from standard-setting. Laziness: accepting the certification frame.

Consumers. Greed: ethical-product signal. Fear: products priced out. Laziness: outsourcing the ethics judgment.

Symmetric FGL reveals no constituency is operating from purely altruistic motives; the question is which motivational structure is shaped by which interest position.

Section 6 — Legitimate value separated from distributional overlay.

The legitimate value: a discoverable signal of ethical-practice commitment is genuinely useful for consumers who want to make informed choices. The auditor function (someone independent verifying claims) is genuinely useful as a check against unverified claims.

The distributional overlay: the specific design (fixed $30K fee; consortium-controlled standard; consortium-approved auditors) loads the legitimate value onto a structure that produces concentration of power and market exclusion. The legitimate value could be served by the alternative design (scaled fee, independent certifying body) without the distributional overlay. The presence of the overlay is not necessary to the legitimate value.

Section 7 — Confidence per finding.

Distributional impact (Section 3) — high confidence; the math is straightforward. Alternative design (Section 4) — high confidence; the alternative is technically obvious and not on the table. FGL (Section 5) — medium confidence on motivational attribution, high confidence on the symmetric application. Institutional authorship (Section 1) — confidence depends on consortium composition, which I did not check; flag for verification before publication.

Boundary-critique handoff note. The analysis surfaces affected-but-not-involved parties — small developers excluded from the consortium, academic AI ethics researchers not represented in standard-setting, consumers in regions where $30K audit cost prices out the products that would serve them. If a deeper boundary critique is wanted, hand off to boundary-critique with the Ulrich twelve-category audit.

That is what cui-bono produces. The institutional author, the rationale on offer, the concrete distributional impact, the technically-rigorous alternative design from the opposite constituency, the symmetric FGL analysis, the legitimate-value separation, and confidence per finding. The user can act on the analysis directly, or hand off to boundary-critique for the deeper Ulrich audit.

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 Interest and Power Analysis pattern with any AI of your choice. The composition is single-pass for any of the four modes.

The prompt:

[Paste the framework specification]

Run Interest and Power Analysis on this.

Mode (optional): [Cui-bono / boundary-critique / wicked-problems / decision-clarity. If not specified, the framework infers from question shape.]

Situation, decision, claim, or text under analysis: [The artifact.]

Stance toward the artifact (optional): [If you have a stance, declare it up front so the framework can apply FGL symmetrically rather than treating your stance as a hidden bias.]

The AI runs the within-territory disambiguation first if the mode wasn’t specified — Q1 (complexity) for “this one situation” / “landscape of parties” / “tangled / wicked,” Q2 (stance) for boundary-critique cross-cuts, Q3 (output) for decision-clarity-document framing — and routes accordingly. The output is mode-shaped: a seven-section cui-bono read; a nine-section boundary-critique audit; a six-section wicked-problems integrated synthesis; a ten-section decision-clarity document.

For best results: apply FGL symmetrically (if the analysis only catches FGL in the side you’re already suspicious of, it’s degraded into advocacy); build the counterfactual design seriously (a strawman alternative reduces the analysis to confirmation of the original); don’t suppress the legitimate-value separation (the seventh section exists to keep both distributional impact and legitimate value visible); for boundary-critique, complete all four category-clusters. The framework is deliberately tool-agnostic — the cui-bono structure, Ulrich’s twelve categories, symmetric FGL, and legitimate-value separation survive the lift to any environment.

Other examples

  • boundary-critique on a research-grant scoring rubric. A funding body’s rubric scores proposals on impact, feasibility, and alignment with funder priorities. Boundary critique runs the twelve Ulrich categories in is/ought: who is and ought to be the client; whose worldview determines the design (technocratic-managerial vs. community-rooted); who is and ought to be witness for the affected. The audit surfaces that the rubric’s is answers consistently favour funder-side standing while the ought answers would extend standing to affected communities. Demonstrates running all four category-clusters rather than selectively confirming initial suspicion.
  • wicked-problems on a regional housing crisis. Six components run (competing-hypotheses fragment, cui-bono, steelman-construction fragment, systems-dynamics-causal, scenario-planning, red-team-assessment fragment); three synthesis stages. Meadows-leverage reading locates candidate interventions on the hierarchy; senge-archetypes reading identifies the dynamic as “fixes that fail.” Residual tensions named explicitly rather than collapsed. Demonstrates the molecular composition’s discipline against silo-aggregation.
  • decision-clarity for a board on a strategic acquisition decision. Four-component composition (cui-bono full, stakeholder-mapping full, scenario-planning fragment, red-team-assessment fragment) produces a ten-section document — decision at hand, decision-maker context, stakeholder map, scenario range, leading intervention recommendation, stress-test findings, residual risks, recommendations within decision-maker scope, confidence map. Decision-shaped rather than analysis-shaped. Demonstrates the molecular composition for a third-party decision-maker.

Citations

The framework draws on three traditions. The interest-and-power tradition contributes cui-bono — Cicero’s Roman juridical question is the historical antecedent; Buchanan and Tullock’s The Calculus of Consent (1962) provides the public-choice substrate; Rumelt’s Good Strategy / Bad Strategy (2011) provides the alternative-design counterfactual discipline. The symmetric-FGL commitment draws on the cynicism-trap and mirror-trap failure modes documented in the analyst literature as predictable degradations of cui-bono analysis run loosely.

The critical-systems-heuristics tradition contributes boundary-critique — Ulrich’s Critical Heuristics of Social Planning (1983) is the substrate for the twelve boundary categories grouped into four sources of design judgment, each run in is/ought form. The involved-vs-affected distinction is Ulrich’s core asymmetry; the four-cluster checklist discipline defends against selective-categories failure.

The wicked-problems tradition contributes wicked-problems and decision-clarity — Rittel and Webber’s 1973 founding articulation; Conklin’s Dialogue Mapping (2006); Meadows’s twelve leverage points (2008); Senge’s system archetypes (1990). Debate D3 (wicked problems sui generis or extreme cases of complex problems) is carried by the mode without adjudication — applies the Rittel-Webber characteristics while remaining agnostic on whether wickedness is a category or a degree. The framework was compiled 2026-05-01; the decision-clarity molecular has a paired execution framework (Framework — Decision Clarity Analysis, restructured per Decision H).

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