Undergraduate teachers and students across six Middle Eastern and South Asian universities used participatory discourse analysis to push back against curriculum standardisation and AI-mediated governance, co-creating a 'Discursive Resistance Toolkit' of communicative strategies that bolstered teacher agency and democratic classroom practices.

Main Finding

Participatory Critical Discourse Analysis (PCDA) functions as a practical, collaborative pedagogical intervention that helps undergraduate teachers and students recognize and resist political and algorithmic governance in literacy classrooms. Through co-analysis and classroom-level discursive strategies (re-lexicalisation, collaborative reframing, collective-agency framing, strategic parody), participants sustained dialogic, democratic literacy practices and co‑constructed a "Discursive Resistance Toolkit" that strengthened teacher agency despite curriculum standardisation and embedded AI systems.

Key Points

• Problem framed: curriculum standardisation and AI-assisted educational technologies act together as dual governance regimes that can erode teacher autonomy, critical dialogue, and inclusive pedagogy.
• PCDA extends Critical Discourse Analysis from diagnostic critique to participatory professional learning by making teachers and students co-analysts of classroom talk and materials.
• Recurring micro-discursive strategies identified:
  • Re-lexicalisation: substituting institutional/algorithmic terminology with locally meaningful terms.
  • Collaborative reframing: joint teacher-student reinterpretation of assessment/assignment framing.
  • Collective agency framing: explicitly invoking shared decision-making and local norms.
  • Strategic parody: using ironic or parodic interaction to expose and subvert prescriptive scripts.
• Outcomes: increased participant awareness of how language, policy, and algorithms interact; practical classroom adjustments; and a locally co-developed Discursive Resistance Toolkit to guide practice.
• Contributions claimed: conceptualizing PCDA, integrating Fairclough’s three-dimensional CDA with interactional sociolinguistics, and offering a replicable participatory model tied to SDG goals (quality, equity, institutions).

Data & Methods

• Design: Qualitative multiple-case study within a Research–Practice Partnership (RPP) using PCDA.
• Settings: Six higher education institutions across Pakistan, Oman, and Saudi Arabia.
• Participants: 12 undergraduate English Language Arts (ELA) teachers and ~200 undergraduate students.
• Data sources:
  • Institutional curriculum and policy documents.
  • AI-assisted instructional materials and platforms used in-class.
  • Classroom observations, audio recordings, and transcripts.
  • Structured collaborative reflection sessions where teachers/students co-analyzed transcripts.
• Analytical approach:
  • Integrated Fairclough’s three-dimensional Critical Discourse Analysis with interactional sociolinguistics (turn-taking, footing, pronouns, evaluative practices).
  • Participatory transcript analysis guided pedagogical redesign and toolkit co‑construction.
• Limitations noted by authors: qualitative scope, regional sample (three countries), focus on undergraduate ELA classrooms; findings oriented to process and pedagogy rather than quantitative causal claims.

Implications for AI Economics

• Algorithmic governance as a governance technology: Educational AI systems are not neutral productivity tools — they embed policy priorities and measurable metrics that alter incentives for schools, teachers, and edtech vendors. Economists should model these systems as institutional technologies that change constraints, payoffs, and behaviors, not only as productivity-improving capital.
• Labor complementarities and substitution: The study suggests complementarities between teacher agency and pedagogical quality; over-standardised/automated systems risk substituting away discretionary teacher tasks that generate non-measurable value (critical dialogue, democratic capacities). Economic analyses of edtech adoption should account for lost intangible outputs and potential long-run human capital externalities.
• Measurement and welfare: Algorithm-driven metrics prioritize observable, standardized outcomes. This can create biased investments toward measurable skills and underinvestment in democratic, critical literacies that produce social welfare but are hard to quantify. Cost–benefit assessments of educational AI must include distributional and long-run social-welfare components beyond short-term test-score gains.
• Market design and product incentives: Edtech suppliers face incentives to optimize for institutional KPIs; without regulation or participatory design pressures, product development will favor features that align with accountability metrics. Policymakers and purchasers should demand transparency, multi-dimensional evaluation metrics, and participatory co-design to internalize non-market values.
• Datafication externalities and equity: Automated systems can marginalize linguistic and cultural variation, generating negative distributional effects (worsening SDG 10). Economists should study how data‑driven models propagate biases across student populations and how that affects lifetime earnings, opportunity costs, and social mobility.
• Adoption costs and transaction costs of resistance: The research documents active classroom-level resistance strategies that may raise coordination/transaction costs for both institutions and vendors. These “resistance behaviors” alter realized returns to AI investments and should be incorporated into adoption models (heterogeneous compliance, endogenous institutional response).
• Policy and regulatory implications: Findings support policies that regulate algorithmic transparency, mandate participatory procurement/implementation (e.g., teacher/stakeholder involvement), and fund teacher professional development as complementary capital. Subsidies or evaluation frameworks that reward broader educational outcomes (civic skills, critical thinking) may correct misaligned market incentives.
• Research & evaluation priorities: For AI economics, there is a need for mixed-methods evaluation designs that combine outcome measurement with process tracing of discourse and agency. RPP-style interventions could be treated as policy experiments to estimate treatment effects on non-test outcomes and on long-term labor-market-relevant skills.

Practical recommendations for economists, policymakers, and edtech stakeholders: - Model AI in education as an institutional technology with incentive effects on behavior and long-run human capital formation. - Evaluate edtech using multi-dimensional welfare metrics and distributional analysis, not only short-term test scores. - Require participatory design and implementation practices to align products with contextual pedagogical values and reduce negative externalities. - Invest in teacher capacity-building as a complement to technological adoption; account for these investments in cost-effectiveness analyses.

Additional note: The study is qualitative and context-specific; its implications for market-level generalizations should be tested with complementary quantitative or mixed-methods research across broader settings.