Overview

The Domain Induction Analysis Framework (DIA) handles the case where the user is stepping into a new domain and wants more than orientation — they want a structured induction with a learning plan tailored to their stated familiarity level and induction goal. The framework distinguishes itself from the lighter operations in T14 (Orientation in Unfamiliar Territory): quick-orientation produces a one-minute lay-of-the-land with no learning plan; terrain-mapping produces a five-minute thorough survey with no learning sequence; DIA composes both into the Domain Induction Document with three integrated parts — what is here in the domain (developed concepts with epistemic-status labels and rival schools), what’s connected to what (central nodes, bridge concepts, prerequisite chains), and what to learn next sequenced by genuine dependency rather than analyst convenience or alphabetical order.

The framework runs two component stages and three synthesis stages. Stage 1 (quick-orientation fragment) produces the rapid lay-of-the-land as breadth seed: domain definition, three-to-five major sub-areas spread across the domain, key terms a newcomer encounters, dominant figures, central debates. The fragment serves as orientation seed for Stage 2 — it does not produce the full quick-orientation output (foundational distinctions, entry points, common misconceptions are Stage 2’s work). Stage 2 (terrain-mapping full) produces the thorough concept map: focus question; known territory with concepts classified known/contested/open and epistemic notes; open questions tied to specific concepts; domain structure (organizing framework — hierarchy, hub-and-spoke, network, layered); cross-links to adjacent domains; boundary statement of what’s out of scope. Rival schools are represented; the standard view does not silently elide dissenters.

The three synthesis stages do the integrative work. Synthesis Stage 1 (orientation-and-terrain merge) unifies the lay-of-the-land breadth with the terrain-map detail, producing the structured “what is here” that subsequent synthesis stages operate on. Every Stage 1 sub-area must be developed in Stage 2 (no concatenation gaps). Synthesis Stage 2 (connectivity mapping) identifies central nodes (concepts that other concepts depend on; removing them would unravel multiple downstream concepts), bridge concepts (concepts that link sub-areas to each other), and prerequisite chains (sequences where understanding C_b requires understanding C_a because B’s definition or operation presupposes A). The output is a relational map showing dependencies, not just elements — relation-omission is a named failure mode. Synthesis Stage 3 (structured induction) produces the Domain Induction Document with the three integrated parts. The learning sequence is dependency-ordered (referencing the prerequisite chains from Synthesis 2), respects the user’s stated familiarity level and induction goal, names specific resources (paper, book, course, lecture series, hands-on experience) with rationale per item, and offers alternative learning paths where multiple valid sequences exist.

The framework’s load-bearing intellectual content is the goal-disconnection defense, the dependency-ordered learning sequence requirement, the relation-mapping discipline, and the rival-schools representation rule. The goal-disconnection defense counters the most common DIA failure mode where the framework produces a generic survey when the user has stated a specific induction goal — research-level vs. working-knowledge vs. general-orientation produce materially different sequences and resource selections. The dependency-ordered learning sequence requirement counters the arbitrary-sequencing failure mode where items are listed alphabetically, by familiarity, or by analyst convenience rather than by the prerequisite structure that makes later items possible. Sequencing rationale references the prerequisite chains explicitly.

The relation-mapping discipline counters the relation-omission failure mode where Synthesis Stage 2 produces a list of elements without dependency arrows. Connectivity is shown as relations (central nodes with downstream dependencies; bridge concepts with sub-areas connected; prerequisite chains with reasoning per link), not as elements stacked alongside each other. The rival-schools representation rule counters the failure mode where the standard view of a domain silently elides dissenting traditions — Stage 2 represents rival schools when the domain has them; the user inducting into category theory, for example, sees both the pure-foundationalist tradition (Mac Lane) and the applied/programming tradition (Lawvere/Milewski) rather than only one.

The framework’s epistemic posture honors honest uncertainty. Where dependency claims are well-established in the literature, they are presented as established. Where dependencies are plausible but not pedagogically validated (a new applied direction where prerequisite chains haven’t been worked out by educators), they are flagged as conjectural-mapping rather than presented as established. Resource selection bias — the framework’s recommendations reflect the analyst’s exposure to the domain’s literature, and for domains where exposure is limited the resources may underweight minority traditions or recent developments — is named explicitly in the confidence-per-recommendation per finding rather than hidden.

The framework answers questions like: I want to induct into category theory at working-knowledge level — I’m a software engineer with strong undergraduate math; produce the structured induction. I’m preparing to do original research in cellular agriculture — give me the research-level induction with primary literature and contested-position study. I want general orientation in Tibetan Buddhism’s Madhyamaka tradition — produce the accessible induction without research-depth requirements.

Systemic context

Domain Induction Analysis is the depth-molecular operation in T14 (Orientation in Unfamiliar Territory), sitting above quick-orientation (T14-light, atomic, one-minute lay-of-the-land) and terrain-mapping (T14-thorough, atomic, five-minute concept map) in the depth ladder. It is the heaviest analytical mode in T14. The framework consumes two component operations — quick-orientation as fragment (light-orientation-only) and terrain-mapping as full — and integrates them via three synthesis stages. The framework distinguishes itself from adjacent territories — T20 (passion-exploration) handles generative exploration when the domain doesn’t have inductable structure; T11 (relationship-mapping) handles cases where relational structure rather than orientation is the goal. DIA is invoked from three paths: user invocation directly with a domain name and prior familiarity / induction goal; pipeline-dispatch via the four-stage pre-routing pipeline classifying T14 at depth-molecular position; or handoff from terrain-mapping when terrain-mapping has surfaced that the user wants the learning sequence beyond the survey-level map. The framework draws on Bloom’s taxonomy (optional, when learning sequence requires cognitive-level scaffolding), novice-expert cognition research (optional, when familiarity level is novice), the Novak concept-map tradition (via terrain-mapping), and the Kahneman-Tversky bias catalog (foundational).

Ora’s output

Example question: I want to induct into category theory at working-knowledge level — I’m a software engineer with strong undergraduate math (analysis, linear algebra, basic abstract algebra) but no prior category theory exposure. I want to read modern functional-programming literature with comprehension and use categorical patterns in code. Six months, ~5 hours/week budget.

Ora’s response (running DIA on the named domain):

The user has stated the domain (category theory), prior familiarity (strong undergrad math, no prior exposure), induction goal (working-knowledge for functional-programming literature and applied code), and a time budget. DIA proceeds.

Stage 1 (quick-orientation fragment). Domain definition: “Category theory studies mathematical structures and their relationships at a level of abstraction that unifies algebra, topology, logic, and computer science.” Major sub-areas: (1) basic categorical structures (categories, functors, natural transformations); (2) limits and colimits; (3) adjunctions; (4) monoidal categories and enrichment; (5) applied category theory (Bayesian networks, databases, programming languages). Key terms: object, morphism, functor, natural transformation, limit, colimit, adjunction, monad, comonad, Yoneda lemma, presheaf, Kan extension. Dominant figures: Mac Lane (foundational), Eilenberg (co-founder), Lawvere (functorial semantics, topos theory), Awodey (modern textbook), Bartosz Milewski (programmer-facing). Central debates: foundationalism (set theory vs. category theory as foundations); applied vs. pure; how much category theory is “actually useful” for working programmers.

Stage 2 (terrain-mapping full). Focus question: “What is the structure of category theory and where does the user enter it?” Known territory: C1 Category (object + morphisms + composition + identity) — known. C2 Functor — known. C3 Natural transformation — known. C4 Limit — known. C5 Adjunction — known. C6 Monad (endofunctor with unit and multiplication) — known. C7 Yoneda lemma — known. C8 Topos — known. C9 ∞-category — contested in pedagogy (when to introduce; how much foundational rigor needed). C10 Operad/multicategory — open as research-frontier for applied work. Open questions: foundation question (sets vs. type theory vs. ETCS); applied-vs-decorative boundary; how ∞-category theory changes pedagogical sequencing. Domain structure: layered — basic structures → universal constructions → derived structures → foundational → frontier. Adjacent connections: cross-link to type theory (Curry-Howard-Lambek correspondence); cross-link to algebraic topology (categories arose there); cross-link to functional programming (monads in Haskell). Boundary: not covering homotopy type theory in depth; not covering full applied domain.

Synthesis Stage 1 (orientation-and-terrain merge). All 5 sub-areas from Stage 1 developed with concepts from Stage 2. Cross-sub-area concept: monads (appears in basic structures C6 AND in applied programming). Rival schools: pure category theorists (Mac Lane lineage, foundational-set-theoretic) vs. applied/programming category theorists (Lawvere lineage, structural; Bartosz Milewski’s pedagogy). Domain structure: layered.

Synthesis Stage 2 (connectivity mapping). Central nodes: C1 (category) — every other concept depends on it; C2 (functor) — limits, adjunctions, monads all defined via functors; C5 (adjunction) — connects to monads, Galois connections, free-forgetful structure throughout. Bridge concepts: C6 (monad) bridges pure category theory and functional programming; C2 (functor) bridges category theory and algebraic topology. Prerequisite chains: PC1: C1 → C2 → C3 → C4 → C5 → C6; PC2: C1 → C2 → C7 (Yoneda); PC3: C5 → C8 (topos requires adjunction fluency); PC4 (applied): C1 → C2 → C6 → applied monad use in Haskell.

Synthesis Stage 3 (structured induction). For working-knowledge level with software engineering background:

Foundation (months 1-2). Item 1: Bartosz Milewski, Category Theory for Programmers (book, ~30 hours). Why first: programmer-facing entry point with software engineering examples; covers C1-C3 thoroughly with code analogies. After: comfortable with category, functor, natural transformation in programmer vocabulary. Item 2: Awodey, Category Theory, chapters 1-3 (textbook, ~25 hours). Why second: provides rigorous mathematical foundation for the same concepts; the parallel treatment (Milewski’s intuition + Awodey’s rigor) is high-leverage for working-knowledge level.

Building (months 3-4). Item 3: Awodey chapters 4-6 (limits, colimits, adjunctions — ~20 hours). Why now: requires C2-C3 fluency from items 1-2; introduces C4-C5. Item 4: Milewski follow-up chapters on monads (~10 hours). Why now: PC4 (C1 → C2 → C6) is now traversable; monad understanding via adjunction (Item 3) makes the formalism click rather than feel arbitrary.

Specialization (months 5-6). Item 5: Riehl, Category Theory in Context, selected chapters on Yoneda and Kan extensions (~20 hours). Why now: PC2 chain complete; Yoneda becomes the central theoretical insight. Item 6: Hands-on project applying categorical patterns in Haskell, Scala with cats, or OCaml (~30 hours). Why now: working-knowledge means demonstrated fluency in a working context.

Alternative paths. Theory-first variant: start with Awodey, defer Milewski to later — works for users with stronger pure-math background. Practice-first variant: start with monads in Haskell, derive theory backward — works for users with strong programming background and weaker tolerance for abstract math first.

Confidence: high for sequence ordering (well-established prerequisite chains in the literature); medium for time estimates (varies by individual); high for resource selection at working-knowledge level (Milewski + Awodey is canonical for this user profile).

That is what DIA produces on a domain with a stated goal and stated familiarity. The induction document carries the three integrated parts — what is here (developed concepts, rival schools, adjacent connections, boundary), what’s connected to what (central nodes, bridge concepts, prerequisite chains), what to learn next (six items, dependency-ordered, with rationale and alternative paths). The user can act on the sequence with eyes open about the prerequisite structure, the time budget, and the alternative paths if their cognitive style favors theory-first or practice-first over the recommended balanced sequence.

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 Domain Induction Analysis pattern with any AI of your choice. The composition is single-pass through the two component stages and three synthesis stages.

The prompt:

[Paste the framework specification]

Run DIA on this domain.

Domain: [The field, area, or topic.]

Prior familiarity (optional but high-leverage): [novice / some prior exposure / working knowledge in adjacent domain.]

Induction goal (optional but high-leverage): [research-level / working-knowledge / general-orientation.]

Time budget for learning (optional): [hours/week × duration.]

Prior resources consulted (optional): [books, papers, courses already encountered.]

Why interested (optional): [informs sub-area emphasis.]

The AI runs the two component stages, three synthesis stages, and produces the Domain Induction Document with executive summary plus the four required sections (what is here; what’s connected to what; what to learn next sequenced by dependency; learning dependencies and prerequisites table) plus the confidence map.

For best results:

  1. State the induction goal explicitly. Research-level vs. working-knowledge vs. general-orientation produce materially different sequences. The framework’s most common failure (goal-disconnection) is producing a generic survey when the user has stated a specific goal. Stating the goal up front routes the synthesis stages correctly.
  2. State the prior familiarity honestly. Overstating familiarity produces a sequence that skips foundational items the user actually needs; understating produces redundant items the user could skip. Honest familiarity assessment is the input that lets the dependency-ordered sequence respect the user’s actual starting position.
  3. Push back on relation-omission. If Synthesis Stage 2 produces a list of central concepts without showing dependencies, ask explicitly which concepts does C_n depend on, and which concepts depend on C_n? The relational map is what makes the learning sequence dependency-ordered rather than arbitrary.
  4. Ask for alternative paths if your cognitive style differs. The framework offers theory-first and practice-first alternatives where multiple valid sequences exist. If neither matches your style (e.g., you prefer historical-development sequencing or problem-driven sequencing), ask for a path tailored to that style — the framework can produce it from the prerequisite chains.

The framework is deliberately tool-agnostic. The goal-disconnection defense, the dependency-ordered learning sequence requirement, the relation-mapping discipline, and the rival-schools representation rule are conceptual disciplines that survive the lift to any environment.

Other examples

  • Research-level induction into a field with active debates. A user is preparing to do original research in cellular agriculture and wants the research-level induction. DIA runs Stage 1 (lay-of-the-land of cellular agriculture’s sub-areas: scaffolding, growth media, bioreactor design, regulatory landscape, consumer acceptance), Stage 2 (terrain-mapping with rival schools represented — the engineering-first tradition vs. the cell-biology-first tradition; the contested questions about cost-curve trajectories and scale-up bottlenecks), Synthesis 1 (merged structured “what is here”), Synthesis 2 (central nodes — cell-line selection; bridge concepts — bioreactor design bridging cell biology and chemical engineering; prerequisite chains across the sub-areas), Synthesis 3 (research-level learning sequence with primary literature, contested-position study, lab-attendance recommendations, and conference attendance with rationale per item). Demonstrates DIA at research-level depth where the sequence includes contested-position engagement, not just canonical material.

  • General-orientation induction into a contemplative tradition. A user wants general orientation in Tibetan Buddhism’s Madhyamaka tradition with no research aspirations. DIA runs the stages with the goal shaping Synthesis 3’s resource selection toward accessible secondary literature (Jay Garfield’s Engaging Buddhism, the Stanford Encyclopedia entry on Madhyamaka, Mark Siderits’s introductory work) rather than primary literature in classical Tibetan or the technical contemporary debates. The learning sequence is dependency-ordered (basic Buddhist concepts → emptiness doctrine → Madhyamaka’s specific moves → contemporary scholarly framings) but at general-orientation depth (10-15 hours total rather than research-level’s hundreds). Demonstrates DIA at general-orientation depth where the framework’s discipline against goal-disconnection prevents producing a research-level survey when the user wanted accessible orientation.

  • Handoff from terrain-mapping to DIA. A user invokes terrain-mapping on “post-Keynesian macroeconomics” and the survey-level map surfaces that the user wants to actually learn the field rather than just orient. The framework hooks upward to DIA. The terrain-mapping output becomes Stage 2’s input directly (no re-running terrain-mapping); DIA runs Synthesis 1 onward, producing the structured induction with learning sequence. The user’s induction goal is elicited at the handoff (working-knowledge for academic engagement); the resource selection is shaped accordingly (canonical post-Keynesian texts plus secondary literature plus engagement with the active debates between Marc Lavoie’s tradition and the modern monetary theory tradition). Demonstrates the handoff pattern that lets users start lighter and escalate when the lighter mode surfaces an induction need.

Citations

The Domain Induction Analysis Framework draws on educational-theory and concept-mapping traditions. The dependency-ordered learning sequence discipline draws on the broad cognitive-science literature on prerequisite knowledge and skill acquisition — concepts have genuine dependency structure, and learning sequences that respect the structure produce faster acquisition than sequences that violate it. Bloom’s taxonomy (Bloom et al., 1956; revised by Anderson and Krathwohl, 2001) provides the cognitive-level scaffolding when learning-sequence items need to be tagged for the cognitive operation they require (remember, understand, apply, analyze, evaluate, create).

The terrain-mapping component draws on Novak’s concept-mapping tradition (Novak and Gowin, 1984) — concept maps as relational graphs with labeled links rather than flat lists; the discipline against relation-omission in Synthesis Stage 2 operationalizes the concept-mapping convention that the value is in the relations, not the nodes. The novice-expert cognition research (Chi, Glaser, and Farr, 1988; the broader literature on expertise) informs the framework’s discipline of treating prior familiarity as a load-bearing input that shapes sequence and resource selection — novices and experts approach the same domain through different entry points and require different scaffolding.

The framework consumes the Quick-Orientation and Terrain-Mapping mode specifications as components; the territory framework for those modes is implicit in the mode files (Modes/quick-orientation.md and Modes/terrain-mapping.md). The framework’s character — heaviest analytical mode in T14, depth-molecular composition, three-part Domain Induction Document — is shared with the Wave 4 depth-molecular pattern used by decision-architecture (T3) and argument-audit (T1), where two-or-more component modes are composed with synthesis stages to produce an integrated document the lighter modes individually could not.

The Kahneman-Tversky bias catalog runs as foundational lens substrate — DIA is vulnerable to the dominant-subfield-bias (the framework’s resource selection over-weights the analyst’s familiar sub-area) and the goal-disconnection failure mode (producing a generic survey when the user has stated a specific goal). The framework’s named-failure-modes discipline (dominant-subfield-bias, relation-omission, arbitrary-sequencing, goal-disconnection) operationalizes the defenses against these biases at the structural level.

The framework is single-author and originated 2026-05-01 as part of the analytical territory build-out. The within-territory disambiguation between quick-orientation, terrain-mapping, and DIA operationalizes the depth-axis selection in T14; the handoff from terrain-mapping to DIA ensures users can start lighter and escalate when the lighter mode surfaces an induction need.

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