← Files EduInsightsARCHIVED FILE
skills/research-with-eduinsights/references/question-to-evidence.md
2.68 KB · Oct 5, 2026 · 18:12 UTC
# Question-to-evidence map Choose evidence by the person's meaning, then preserve the source's actual scope. ## Colleges and fields - College identity and current context: IPEDS, College Scorecard, Carnegie Classifications, and dated accreditation coverage where available. - Reported degrees: IPEDS completion records at college, CIP-6 field, credential level, reporting cycle, and release. - Earnings and debt: broader College Scorecard field cohorts. These may be inherited by a narrower CIP-6 field and are not direct named-program outcomes. - Enrollment history: comparable IPEDS collection cycles and accepted releases. - Accreditation: dated DAPIP records. No matched record is not proof of no accreditation. - Current course requirements and catalog program names: outside the serving snapshot. ## Fields and careers - CIP-to-SOC edges describe structural preparation paths. - They do not show where graduates worked, placement rates, probabilities, demand, or a guaranteed transition. - Compare fields at the same taxonomy level and occupations at the same SOC level. ## Careers and work - BLS OEWS supplies employment and pay estimates. Geographic and industry packages can overlap and are not additive. - BLS Employment Projections supplies national outlook over a stated decade. It is not an AI forecast. - O*NET supplies detailed occupations, tasks, skills, ratings, and work context. - State projections use state-specific periods. Do not imply that all states share one period. ## AI-use evidence - Anthropic Economic Index observations are empirical evidence about AI use within named Claude.ai or API populations. - Lead with those findings when they are the strongest available evidence. - State the product channel, observed population, geography, period, metric, and release. - Do not convert AI use, task penetration, augmentation, or automation into unmeasured job loss, productivity, pay, or time savings. - National AI context cannot be attached to a college, program, or occupation as direct evidence. ## Applicability and coverage Translate the evidence fit into plain language: - direct: the evidence matches the entity and level asked about; - inherited: broader field evidence provides context for a narrower field; - fallback: college-level evidence stands in when field evidence is unavailable; - structural: a taxonomy or crosswalk indicates a possible relationship; - empirical calibration: a bounded observed population informs the broader concept; - context: macro or methodological evidence describes the environment only. A zero is an observed number. A blank may mean suppressed, not collected, not applicable, not participating, stale, or unresolved. Use the reason returned by the tool.
SHA-256: d99993ccbe03cf6bf0554be4cef18e6c2892759bdb138270dea3605f4bbb30d5