STEM Faculty Know Active Learning Works but Don't Adopt It — and Training Won't Fix It

Active learning in STEM courses improves student outcomes across every measured dimension — exam scores, failure rates, conceptual understanding, persistence in the major — yet lecture remains the dominant instructional mode in U.S. undergraduate STEM education. Large-scale classroom observation studies show that only 18% of STEM instruction is student-centered, and lecture still occurs in 50% of time segments even in classrooms categorized as "active." The gap between what is known and what is practiced has persisted for over two decades despite hundreds of millions of dollars in NSF-funded faculty development programs, because the interventions target the wrong level of the system: they train individual faculty, but the barriers are institutional.

STEM attrition is concentrated in introductory courses where lecture dominates — an estimated 40% of students who enter college intending to major in STEM switch to other fields or drop out entirely. The attrition is not uniformly distributed: women and students from underrepresented racial/ethnic groups leave STEM at disproportionately higher rates, and the evidence consistently shows that active learning disproportionately benefits these populations. Every year that the adoption gap persists, it compounds the demographic narrowing of the STEM pipeline. NSF's IUSE: EDU program commits approximately $61 million per year to this problem, yet the program's own framing acknowledges that "there are widespread barriers to the adoption of these practices."

Faculty development workshops are the primary intervention. They raise awareness and can change beliefs, but they do not produce sustained behavior change. A 2023 ecological model study found that faculty instructional decisions are shaped by interacting personal, social, and contextual factors — attitudes, beliefs, student expectations, colleague norms, departmental culture, time constraints, and institutional reward systems. Training addresses only the personal factor. Faculty frequently cite that tenure and promotion criteria reward research output, not teaching quality, creating a rational incentive to minimize time spent on instructional innovation. When faculty do try active learning, they encounter student resistance (students socialized into lecture expectations push back on unfamiliar formats), which faculty interpret as evidence that the method doesn't work, and they revert to lecture. One-time workshops produce short-term enthusiasm but not sustained change; longitudinal studies show that without ongoing coaching and institutional support structures, adoption decays within two semesters. The problem is not that faculty don't know about active learning — it's that the institutional environment makes adoption costly and unrewarded.

Restructuring the incentive architecture of STEM departments rather than retraining individuals within the existing architecture. Tenure and promotion criteria that weight teaching quality alongside research productivity; departmental norms that make peer observation and iterative course improvement standard practice; redesigned classroom spaces that physically prevent defaulting to lecture (fixed-seat lecture halls make group work impractical); and pre-tenure faculty development that integrates teaching innovation with research identity rather than positioning them as competing demands. The NSF IUSE: EDU "Institutional and Community Transformation" track acknowledges this — it funds systems-level change — but remains a fraction of the portfolio relative to individual-faculty-focused grants.

A student team could conduct a structured audit of how a specific STEM department's promotion criteria, space allocation, and scheduling practices interact to enable or prevent active learning adoption, producing a concrete institutional change proposal with estimated costs. Alternatively, a team could design and prototype a "faculty adoption support system" — a combination of peer mentoring structure, observation protocols, and student feedback mechanisms — modeled on medical residency supervision, tested in a single department. Relevant disciplines: organizational behavior, higher education policy, learning sciences, human-computer interaction (for classroom technology design).