AFSC Strategy Matrix (Full Matrix + Machine-Readable Scaffold) The AI for Science Strategic Compass (AFSC) is a function-based decision framework that supports strategy-level planning for AI-enabled scientific discovery. AFSC aligns six core AI functions (Represent; Reason & Infer; Optimize & Control; Simulate & Emulate; Generate & Create; Autonomize & Orchestrate) with four cross-domain, recurrent discovery tensions (System Complexity, Experimental Constraint, Data Scarcity, Combinatorial Explosion), yielding a 6×4 Strategy Matrix. In the Strategy Matrix, rows are core AI functions and columns are discovery tensions. Each function–tension cell indicates how the function can mitigate the corresponding tension and provides (i) a concise keyword naming the cell-level mitigation logic, (ii) three strategic pathways that capture mechanism-level mitigation routes, each anchored to a minimal atomic signature, and (iii) illustrative method-family terminology. Method-family examples are non-prescriptive; exemplar citations and expanded pathway definitions are provided in the companion preprint (see below). What’s in this record A high-resolution figure of the full AFSC Strategy Matrix. A machine-readable scaffold JSON (afsc_scaffold_v0.1.0.json) that encodes the stable structure (tensions, functions, dependency chain, function-internal atomic layer/triads, cell keywords, pathway names, and atomic signatures). Note: the JSON intentionally omits long prose definitions of pathways to minimize maintenance overhead; those definitions live in the companion preprint/Supplementary. AI Core Functions The six core AI functions and their atomic triads are defined in more detail in the AI Core Function Ontology: A Two-Level Capability Framework (DOI: 10.5281/zenodo.17664037), which provides the capability backbone for AFSC. F1 Represent — encodes raw inputs into structured or latent states. F2 Reason & Infer — derives constraints, relations, or calibrated beliefs from representations. F3 Optimize & Control — selects actions or designs under objectives and constraints. F4 Simulate & Emulate — forecasts or tests via solvers, surrogates, or hybrids. F5 Generate & Create — synthesizes samples, designs, or hypotheses under specifications. F6 Autonomize & Orchestrate — composes and supervises multi-step processes. Scientific discovery tensions T1 System Complexity — structural intricacy that impedes modeling and explanation. T2 Experimental Constraint — limits on acquiring new empirical evidence. T3 Data Scarcity — insufficiently informative/reliable evidence relative to difficulty. T4 Combinatorial Explosion — growth of search spaces that makes exhaustive exploration infeasible. (Full definitions are in the companion preprint; the JSON includes concise tension/function definitions aligned with the manuscript.) How to use Identify the dominant discovery tension(s) for your problem. Enter the corresponding column and use starred (★) cells as typical high-leverage entry points (non-exclusive). Select pathway(s) whose mechanisms fit your evidence needs and constraints. Use illustrative method families to guide instantiation with context-appropriate implementations. Record the reasoning chain (tension → cell → pathway → instantiation) so the plan remains explicit and revisable. Responsible use note: This Compass is a decision-support framework, not a normative standard or automated decision rule. You should adapt it to your context, values, constraints, and legal/ethical requirements, and remain accountable for downstream choices and impacts. Companion paper and full definitions The full conceptual framing of AFSC, expanded pathway definitions, and exemplar method-family references are available in the companion preprint (DOI: 10.5281/zenodo.17783976 ). This record is intentionally focused on the full matrix figure and a concise, machine-readable scaffold. Version notes v1.2 — Machine-readable scaffold release (no matrix changes). Added a machine-readable scaffold JSON (afsc_scaffold_v0.1.0.json) encoding tensions, functions, dependencies, atomic triads/quadrants, cell keywords, pathway names, and atomic signatures. v1.1 — Standardized spelling to American English; refined a few mitigation keywords for consistency. No structural changes to the Strategy Matrix. v1.0 — Initial release.