Data_Extraction_Workbook_Final - Public (Including ICC)

Résumé fourni par la source

Bangladesh published one of South Asia’s earliest national AI strategies in 2020, yet by 2025 ranks behind India and only marginally ahead of Pakistan and Sri Lanka — both considerably later adopters — on the Government AI Readiness Index. This paper asks why. Combining a six-year trajectory analysis across six South Asian countries, a formalised policy-document crosswalk (the Institutional Continuity Crosswalk, or ICC), a panel regression, and triangulation against the UN E-Government Development Index, we find Bangladesh’s readiness gains are real but concentrated in domains the state controls directly — public service delivery — and largely absent where readiness depends on coordinated private-sector action, such as industrial and financial AI diffusion. The ICC traces this split to Bangladesh’s own policy record: priorities aimed at government operations were sustained across two strategy documents six years apart, while private-sector-facing priorities were dropped or diffused into vaguer commitments. A parallel crosswalk of India’s 2018–2024 strategy revision shows the opposite pattern — broad continuity with nothing dropped — corresponding to a more evenly strong readiness profile. Theoretically, the paper extends Bromley and Powell’s (2012) distinction between policy-practice and means-ends decoupling by locating decoupling within a single strategy rather than treating it as a property of the strategy as a whole, and proposes its locus is conditioned by state capacity and directness of state control. Methodologically, the ICC is offered as a replicable protocol for tracing policy-priority continuity against measured outcomes, applicable beyond this case. The panel regression finds a positive but statistically insignificant association between strategy adoption and readiness gains, reported descriptively rather than causally given the small comparator sample.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.