The Feedback Paradox: Why Course Evaluations Aren't Your Real Teaching Feedback

You already know student evaluations are imperfect. But the research from 2022 to 2025 goes much further than "imperfect," and the findings should change how every institution thinks about teaching feedback.

You Already Know This Part

By now, most faculty have heard some version of the argument: student course evaluations are imperfect. Maybe you read a blog post about it, or a colleague mentioned it at a conference. The basic finding (that evaluations don't correlate well with actual learning) has been circulating in faculty development circles for years. Uttl, White, and Gonzalez published a meta-analysis in 2017 covering 51 multi-section studies and found near-zero correlation between student ratings and student learning.¹ That's settled science at this point.

So let's skip the part where we relitigate whether evaluations are flawed. They are. You know it. Your colleagues know it. Your teaching center probably knows it.

The more interesting question is: what has the research revealed since then? Because the findings from 2022 to 2025 go well beyond "imperfect." They're genuinely surprising, and they change the conversation from "evals are a bit noisy" to something much more uncomfortable.

Here's what you probably don't know yet.

1. Evaluations don't just fail to measure learning. They're negatively correlated with it.

This is the finding that should stop every dean mid-sentence. In 2025, researchers at Ross University School of Veterinary Medicine published a study in BMC Medical Education that tracked 160 courses across 12 student cohorts. They used an independent, external measure of learning (the Veterinary Educational Assessment) to compare what students actually retained against how they rated their instructors. The result: student evaluations showed a significant negative correlation with learning outcomes.² Students in courses with higher evaluation scores actually performed worse on independent assessments of what they'd learned. Not zero correlation. Negative correlation. The courses students liked most were, on average, the courses where they learned least.

This isn't just "evals are noisy." This is: evals may systematically reward the wrong thing.

2. Gender bias in evaluations isn't just about individual prejudice. It's structural.

You've probably heard that women receive lower teaching evaluations than men. But a 2023 study in Proceedings of the National Academy of Sciences by Aragón, Pietri, and Powell revealed the mechanism, and it's more insidious than simple bias. Analyzing over 100,000 evaluations at Clemson University, they found that gender discrimination in evaluations is driven by department gender composition.³ When a woman teaches in a male-dominated department, students have implicit expectations about who "belongs" teaching upper-level courses. Violating those expectations triggers evaluation penalties that have nothing to do with teaching quality. Both men and women can face this kind of role-expectation penalty, but women are disproportionately affected because they're more likely to be in the minority in their department. The bias isn't just in the students' heads; it's baked into the institutional structure.

3. Biased evaluations cause measurable mental health damage.

In 2024, Daskalopoulou published an intersectional study in Studies in Higher Education that documented what many faculty have felt but rarely seen validated in research: biased evaluations cause real psychological harm.⁴ Through in-depth interviews with academics in UK business schools, the study found that evaluation bias (compounding across gender, race, age, disability, and other identity markers) leads to measurable mental health impacts and directly harms career progression. The institutional pressure to maintain high satisfaction scores forces faculty to modify their teaching behavior in ways that compromise their wellbeing. This isn't just a data quality problem. It's a health problem.

4. Students themselves don't believe evaluations matter.

Here's an irony worth sitting with. Sullivan, Lakeman, and colleagues published a scoping review in Assessment & Evaluation in Higher Education (2024) examining what actually motivates students to complete course evaluations.⁵ Across 21 qualitative studies, the pattern was consistent: students say they'll participate only if they believe a genuine feedback loop exists, if confidentiality is assured, and if they see evidence that their feedback leads to actual changes. Most students don't believe any of those conditions are met. The result is that response rates are low and the respondents who do participate skew toward the extremes (very satisfied or very dissatisfied), creating a fundamentally compromised sample. The people whose opinions are supposed to validate your teaching? Many of them think the entire exercise is pointless.

5. Grade inflation is the single best predictor of evaluation scores.

A 2022 study in Assessment & Evaluation in Higher Education asked a straightforward question: what actually predicts teaching evaluation scores?⁶ Using multiple regression analysis, the researchers found that GPA (a proxy for grading leniency) was the dominant predictor, not teaching quality, not course design, not pedagogical innovation. Faculty can mechanically increase their evaluation scores by inflating grades. This creates a race to the bottom where rigorous instructors are systematically punished, and the path to high evaluations runs through easier grading, not better teaching.

6. When COVID forced institutions to pause evaluations, nothing collapsed.

This might be the most revealing finding of all, not from a single study but from a natural experiment that played out across hundreds of institutions. During 2020 and 2021, universities including DePaul, UC Irvine, and the University of Georgia either suspended course evaluations entirely or exempted them from promotion and tenure decisions. The result? Departments continued to function. Teaching quality didn't visibly decline. Assessment committees found other ways to evaluate faculty. When given permission to stop relying on evaluations, institutions discovered they could. Most resumed the practice anyway, not because evaluations proved their worth during the pause, but because of administrative inertia and the absence of a ready replacement. Evaluations persist not because they're valid, but because nothing has formally taken their place.

These findings land differently depending on where you sit in the academy. That variation matters, because "evals are biased" is an abstract statement until you see how it intersects with your specific position, identity, and job security.

For Pre-Tenure Faculty

The negative correlation between evaluations and learning creates a genuinely perverse incentive. You're building a tenure file, and evaluations are in that file. Senior colleagues are explicit: "Don't let teaching be a weakness in your case." But the 2022 grade inflation study tells you the most reliable way to raise those scores is easier grading, not better pedagogy. The 2025 BMC study tells you the courses students rate highest may be the courses where they learn least. Pre-tenure faculty are caught between the evidence and the institution, forced to optimize for a metric that may reward the opposite of good teaching.

For Tenured Faculty

The immediate career stakes are lower, but Daskalopoulou's 2024 research on mental health suggests the psychological damage doesn't disappear with tenure. A tenured professor might know intellectually that evaluations are noisy, but that doesn't stop the sting of a comment calling her "disorganized" or "unapproachable." Tenure protects the job; it doesn't eliminate the internalized narrative. And tenured faculty who sit on promotion committees often weight evaluations in exactly the way the research says they shouldn't, perpetuating the cycle.

For Women and Faculty of Color

The Aragón et al. PNAS study (2023) reveals that this isn't just individual student prejudice; it's structural, driven by department gender composition. A woman teaching in a male-dominated department faces role-expectation penalties that have nothing to do with her teaching. Faculty of color receive evaluations that question their belonging in ways their white colleagues don't. Daskalopoulou's intersectional analysis shows these biases compound across identity markers (gender, race, age, disability) and the cumulative effect is measurable career harm and mental health impact. The evaluation form becomes a place where institutional bias finds a quantitative mechanism.

For Contingent Faculty

The stakes are sharpest here. A visiting instructor or adjunct is often rehired based substantially on evaluations. Low evaluations can cost a contract. Combined with the grade inflation finding, this creates an impossible bind: maintain rigor and risk your job, or inflate grades and keep the appointment. The COVID pause demonstrated that institutions can function without evaluation-based rehiring decisions. But most returned to the practice, and contingent faculty bear the consequences.

Across Disciplines

The negative correlation between evaluations and learning likely varies by field. A statistics course with challenging problem sets may show a sharper negative correlation than a descriptive survey course. A philosophy professor's Socratic method might receive lower evaluations than a lecture-based intro biology course while producing deeper learning outcomes. The 2025 BMC study was conducted in veterinary education, but the underlying mechanism (students prefer courses that feel easier) applies broadly.

The point is not to excuse bad teaching or suggest all feedback is worthless. It's to say: your evaluation results are embedded in a system with documented structural biases, documented incentive problems, and a now-documented negative relationship with learning. Your job is to interpret your own results with clear eyes about what they actually measure.

The old framing was: "evaluations are imperfect, so take them with a grain of salt." The new framing, given what we now know, is more fundamental: evaluations are measuring the wrong thing, and the thing they're measuring (student satisfaction, which correlates with grading leniency) may be negatively related to the thing you actually care about (learning).

That's a different kind of shift. It's not "interpret your evals carefully." It's "stop treating your evaluation score as information about your teaching quality at all." The aggregate number is worse than noise; it may be systematically misleading.

This doesn't mean ignoring everything on the evaluation form. Specific, repeated feedback about something concrete still matters. If 80% of your students say they don't understand the grading criteria, that's actionable regardless of what the overall score says. If students consistently report they can't access materials, that's a real problem. But the aggregate "overall teaching effectiveness" score? The 2025 BMC study suggests that score is more likely to reflect how easy the course felt than how much students learned.

The more productive shift is to ask: if evaluations aren't telling me whether I'm a good teacher, what would? What signals actually correlate with learning?

Peer observation. When a colleague sits in your classroom and watches you teach (not evaluating you for personnel purposes, but observing with a collaborative eye) they can tell you things evaluations never will. They can tell you whether your transitions were clear, whether you engaged the quieter students, whether you addressed misconceptions or let them pass. The POD Network maintains practical guides for collaborative classroom observation that focus on improvement rather than evaluation.

Student work analysis. If you have access to actual assignments, exams, or projects from your course, you can ask: are students acquiring the competencies? Are their papers getting better across the semester? Can they apply concepts to new problems? These questions are answerable by looking at the work itself, not by asking students to rate their satisfaction.

Teaching portfolios and reflection. When you review the materials from a course, the assignments you gave, the feedback you provided, and you reflect on what worked and what didn't, you're engaging in the kind of iterative improvement that actually leads to better teaching. The SoTL Commons offers templates and examples of how faculty document and reflect on their teaching practice.

Mid-semester feedback conversations. Instead of waiting until the end of the semester when evaluations come in, you can ask students partway through: what's working? What's confusing? What would help? This gives you information you can actually act on, while there's still time to adjust.

All of these alternatives have something in common: they require real engagement, not just passive reception of evaluation scores. They're more work. But they're also more likely to actually improve your teaching, because they're based on signals that correlate with learning rather than signals that correlate with grading leniency.

There's no prescribed set of steps here. Your situation is genuinely yours, shaped by career stage, identity, discipline, and institutional context. But there are some invitations worth extending, informed by what the recent research makes clear.

One is to stop internalizing the aggregate score. The next time you catch yourself thinking "I'm not a strong teacher" because your overall rating is a 3.8, remember the 2025 BMC study. A 3.8 may mean your course was challenging in ways that produced real learning. A 4.7 down the hall may mean the opposite. The number doesn't carry the information you're looking for.

Another is to experiment with gathering signals that actually correlate with learning. Pick one alternative this semester: peer observation, mid-semester feedback conversations, student work analysis, or a teaching portfolio. The POD Network maintains practical guides for collaborative classroom observation that focus on improvement rather than evaluation. The SoTL Commons offers templates for documenting and reflecting on teaching practice. COVID showed us institutions can function without evaluation-based decisions. Building alternatives is not hypothetical; it's been tested.

A third is community. Sullivan et al.'s 2024 scoping review found that even students don't believe evaluations create genuine feedback loops. If both sides of the evaluation form question its value, the real feedback happens elsewhere: in teaching circles, peer observation networks, departmental assessment groups, and the kind of honest conversations that happen when the person with institutional power isn't in the room.

That's ultimately what this is about. Not giving up on getting better at teaching. Not deciding that feedback doesn't matter. It's about recognizing that evaluations don't just fail to measure teaching quality; recent evidence suggests they may actively point in the wrong direction. The real feedback, the kind that actually improves your teaching, comes from peers, from student work, from reflection, from the messy collaborative process of trying to help students learn something genuinely hard.

Your students need you to be a good teacher. The measure of whether you're succeeding is whether they're learning what matters. That question has never been answerable by a satisfaction survey. Now we have the evidence to prove it.