India Institute is evaluating a critical thinking curriculum in Tamil Nadu government schools through a randomised controlled trial commissioned by the state government. We were research advisors to the trial's principal investigators, and we worked with the team on the evaluation design, the review of instruments and the measurement choices that will decide what the RCT can and cannot conclude.
We kept coming back to a question the team was somewhat reluctant to ask, which was what exactly the trial measures when it measures critical thinking. The study cannot avoid that question, and we raised it as a methodological point that comes before any analysis. An RCT produces a precise estimate of an intervention's effect on a measured outcome. If that outcome does not map cleanly onto the construct the intervention is trying to change, a rigorously conducted trial can produce a finding that is technically correct and misleads anyone who acts on it.
Critical thinking has a fairly settled philosophical definition. It covers the capacity to analyse claims in a structured way, to recognise logical fallacies, to evaluate evidence, and to hold calibrated epistemic humility about one's own reasoning. There is no universally accepted instrument for measuring it, and especially none for children in non-Anglophone state schools.
Every instrument that exists carries assumptions. The Cornell Critical Thinking Test and the Collegiate Learning Assessment were developed for US undergraduates. The PISA framework for reasoning comes closer to a cross-cultural instrument, but it has had limited validation in Tamil-medium government schools. Locally developed instruments fit the culture better and bring problems of reliability and comparability. None of the options is clean.
Our advice to the team was to triangulate. With a single instrument, the trial produces a headline number whose meaning depends entirely on whether the reader trusts that instrument. Combining test scores with structured classroom observation and measures of teacher practice gives a picture that is harder to summarise in a press release and much harder to dismiss.
The classroom observation component matters more than it may first appear. A critical thinking curriculum changes what students know, and it also changes how classrooms run: the ratio of teacher talk to student talk, whether students are invited to challenge assertions, and whether an error is treated as information or as failure. These changes in practice are often the proximate mechanism through which student outcomes shift. If the trial measures student outcomes and not practice, it cannot tell a well-implemented curriculum from a poorly implemented one that happened to produce a positive test score for unrelated reasons.
Tamil Nadu is a useful site for this kind of study. By Indian standards its government school system has relatively high capacity, and UDISE data consistently place the state among the top states on infrastructure, teacher availability and enrolment ratios. The ASER 2023 report shows Tamil Nadu's foundational literacy and numeracy outcomes are strong relative to national averages, so the study is working above the floor. Gains from a critical thinking curriculum are more likely to be visible in a system where basic literacy is not the binding constraint.
The trial assigns whole schools to treatment and control, and a cluster-randomised design suits a curriculum delivered at classroom level. The power calculations had to account for intra-cluster correlation in Tamil Nadu government schools. We advised the team to take intraclass correlation estimates from ASER and NAS data, because generic assumptions may not hold in these schools.
Whether the trial detects effects depends heavily on implementation fidelity. A curriculum that teachers do not understand well enough to teach is a different treatment from one delivered as designed. Teacher training and ongoing coaching are part of the intervention, and the evaluation needs to document them directly.
UDISE+ state report cards provide school-level baseline data. Tamil Nadu EMIS has more granular district data on enrolment, attendance and teacher deployment.