Peripheral Intelligence in the STEM Classroom: Investigating Generative AI Readiness, Institutional Capacity, and Educational Equity in Secondary Schools of South Punjab, Pakistan

Abstract

After 2020 the number of studies in AI for science, technology, engineering and mathematics (STEM) education has grown very fast, and the evidence base is currently geographically clustered in North America, East Asia and Europe. Peripheral education systems (in which infrastructure is lacking, language is complex, and secondary school completion rates have been low for historical reasons) are seldom considered as places where theory building occurs. To fill this gap, this mixed-methods study explores the readiness of secondary schools for generative AI (GenAI), capacity of institutions and STEM opportunity in Multan, Bahawalpur, and Dera Ghazi Khan, the three divisions of South Punjab, Pakistan. A newly designed Peripheral AI–STEM Capability Framework (PASCF) was developed, based on AI–Technological Pedagogical Content Knowledge, the Unified Theory of Acceptance and Use of Technology 2, and Sen's capability approach and we surveyed 486 STEM teachers and 812 Grade 9–12 science-group students, and conducted 48 semi-structured interviews with teachers, headteachers, and students. Measurement model was confirmed by confirmatory factor analysis. Structural paths revealed that institutional capacity (β = .41), greater than behavioral intention and ethical readiness, uniquely predicted AI–TPACK, which in turn, was the strongest predictor of classroom GenAI–STEM integration (β = .38). While the informal use of GenAI tools was common (71% of teachers said they used them out of class), the formal use of the tools aligned with the curriculum was rare (18%). The lowest Opportunity-to-Learn (OTL) scores were found for females in rural settings. Themes from the interviews uncovered two streams: a rich, informal digital hustle and a formally silent STEM classroom, limited in electricity, devices, language (Urdu/Saraiki/English) relations, assessment practices that discourage inquiry, and the lack of policies on AI integrity at the school level. The study presents a theory of peripheral AI–STEM capability that accounts for the context-specific nature of this skill; an equity-centered measurement model to evaluate the same; and policy recommendations consistent with Punjab's upcoming STEM secondary-education reforms. The findings alert to the dangers of scaling GenAI without making structural access "pedagogical capability" to reduce, but not close, the opportunity gap for the South Punjab STEM.

Keywords: Generative Artificial Intelligence, STEM Education, Teacher Readiness, AI–TPACK, UTAUT2, Capability Approach, Educational Equity, South Punjab, Mixed Methods