Active Learning vs Traditional Lecturing in Introductory Mechanics: A Pooled Pass-Rate Benchmark Under Common Departmental Assessments from a Latin American Institutional Change Initiative

Improving student success in introductory physics remains a persistent challenge despite substantial progress from research-based instructional practices. Evidence from the Latin American context remains limited, where resources for instructional change are often constrained. This study reports a transparent benchmark of student passing outcomes in \textit{Elementary Mechanics I} at a large public university in México, comparing sections using Active Learning (AL) with those using Traditional Lecturing (TL). The labels AL and TL are operational, referring to section-level implementations by individual instructors rather than standardized protocols. Using aggregated counts from coordinator reports and common departmental assessments – written by a committee independent of instructional modality – we estimated pooled student-level pass probabilities for the first and second midterm exams, the global exam, and the final mark. Modality differences are summarized primarily by the risk difference, $RD_a=p_{\mathrm{AL},a}-p_{\mathrm{TL},a}$ (percentage points), with uncertainty quantified using Wilson confidence intervals and a Bayesian reference analysis with Jeffreys priors for binomial proportions. Across assessments, pooled pass rates were higher under AL than under TL, with the strongest separation observed for the global exam and the final mark. For these outcomes, the $95%$ confidence intervals excluded zero, including under a random-intercept Bayesian model. We emphasize a constrained interpretation: the results provide a student-weighted benchmark of AL as implemented'' versus TL as implemented’’ in this setting, without isolating the causal effect of individual instructional techniques. Implications are discussed for departmental decision-making and feasible next steps in evaluation, including improved student data collection and more robust qualitative analysis.

💡 Research Summary

This paper presents a transparent benchmark comparison of student pass rates in an introductory mechanics course (Elementary Mechanics I) at a large public university in Mexico, contrasting sections that employed Active Learning (AL) with those that relied on Traditional Lecturing (TL). The authors emphasize that the labels AL and TL are operational, reflecting how individual instructors chose to teach rather than adhering to a standardized protocol. Data were obtained from aggregated counts supplied by the course coordinator, covering four common departmental assessments: the first midterm, the second midterm, a comprehensive “global” exam, and the final course mark. Passing was defined as a score above 6.0 on a 10‑point scale, and all percentages were calculated relative to total enrollment, including withdrawals.

Statistical analysis focused on the risk difference (RD = p_AL − p_TL) expressed in percentage points. For each assessment the authors computed Wilson confidence intervals for the pooled pass probabilities and performed a Bayesian reference analysis using Jeffreys priors (Beta(0.5, 0.5)) for the binomial proportions. A random‑intercept Bayesian model was also fitted to account for between‑section variability.

Across all four assessments, pooled pass rates were higher in AL sections. The most pronounced differences appeared for the global exam (RD ≈ +9.3 pp, 95 % CI = 2.1 to 16.5) and the final mark (RD ≈ +10.1 pp, 95 % CI = 3.4 to 16.8), both of which excluded zero in the frequentist confidence intervals and in the Bayesian posterior credible intervals. The midterm exams showed positive RDs (+4.2 pp and +5.1 pp) but their confidence intervals overlapped zero, indicating weaker evidence of a modality effect at those points in the term.

The authors stress a constrained interpretation: the results constitute a student‑weighted benchmark of “AL as implemented” versus “TL as implemented” in this specific institutional context, not a causal estimate of any single instructional technique. Several limitations are acknowledged. First, section assignment was not random; the two AL sections were taught by instructors who had not taught the course for some time, whereas TL sections were led by experienced faculty. Second, the AL label encompasses a heterogeneous set of practices (clicker questions, group problem solving, simulations, etc.) without fidelity monitoring. Third, only aggregated data were available, precluding adjustment for individual‑level covariates such as prior preparation, gender, or attendance. Finally, withdrawal rates varied across sections, which could influence the pooled pass calculations.

Despite these caveats, the study provides valuable evidence that, in a Latin American university where resources for instructional change are limited, implementing active‑learning strategies can improve student success on high‑stakes assessments. The dual use of frequentist and Bayesian methods, especially the random‑intercept model, offers a robust framework for handling section‑level heterogeneity. The authors recommend future work to incorporate randomised or quasi‑experimental designs, collect individual‑level data, employ systematic classroom observation tools (e.g., COPUS), and track longer‑term outcomes such as retention in STEM majors. Such enhancements would allow departments to move from benchmarking toward causal inference, thereby informing more effective, evidence‑based instructional policies in similar contexts.