Conceptual teaching and visual slides

The largest single cluster of praise, 21 segments, was for the “why”-first approach to the material. Students noticed that I spoke through ideas instead of reading slides, and several said the sparse text kept them engaged and made them take notes. A near-equal group of 20 segments praised the slides themselves for being visual and well organized, and tied that directly to studying. Entropy, reduction potentials, and the reasoning behind mechanisms came up by name as things that finally made sense. This is the part of the course I most wanted to get right, and the feedback says it landed.

Access, support, and openness to feedback

“Was available to me” was my highest-rated item at 4.0, and the comments explain the score. Students singled out the per-module discussion board, the number of office hours, and quick answers to questions. Nine separate segments noted that I was open to feedback and adjusted mid-course, and a few asked me not to take their criticism personally because they could tell I cared. I want to keep that read in view, because it is the ground that makes the harder feedback usable rather than adversarial.

Course design that students endorsed

Students backed most of the structural choices even while asking me to fix how I ran them. The tiered question concept (Core, B, Analytical, Synthesis) drew 16 positive segments: students liked knowing the difficulty level of each question and called it a fairer idea than a blanket curve. The volume of practice across recitations and homework (15 segments), the two standalone quizzes as low-stakes memorization checks, and the grade drops and buffers all drew praise, with one student calling the padding rare in STEM and “so, so helpful.” The posted keys and free supplementary materials were valued too. The scaffolding was right. The problem, as the next section shows, was consistency and calibration, not concept.

Make exams consistent and recalibrate the tiers

The most-cited problem, 27 segments, was that the exam format and weighting kept changing. Students reported the format shifting at least twice and asked, reasonably, that the syllabus be set and then held. A related 14 segments said the tier labels did not track difficulty: C-tier questions that were harder than A-tier ones, and heavy weight on the many C-tier items, so that a few slips sank a score. Thirteen more flagged confusing wording, with the Exam 3 turn toward multiple-answer “select all that apply” singled out as testing second-guessing more than biochemistry. Locking the format for the term and re-checking every question’s tier against its difficulty is the highest-value change I can make.

Align practice with the exam and state expectations plainly

The single biggest change theme, 28 segments, was that students did not know what or how deeply to study. The most common request was concrete: study guides, key takeaways, and practice exams. Alongside it, 20 segments said the practice they did have (homework, recitation, slides) did not resemble the exam. That likely explains my lowest numeric item, helpful feedback at 3.06: the issue is less about slow grading than about students not seeing the path from the material to the test. Practice that mirrors the exam format, plus an explicit statement of scope before each exam, answers both.

Slow down and thin the slides

Eleven segments described lectures as rushed, often because a fifty-plus-slide module got compressed into one session and question time got cut when I ran short. Nine more said the slides were too dense and jargon-heavy, with terms left undefined. The fix students proposed is the one I will take: split each module into pre-class and in-class content, define terms as they appear, and protect time for questions instead of sacrificing it to finish the deck.

Decide the grading philosophy up front

Fourteen segments objected to the absence of a curve against low class averages, and 52 percent of students judged my standards “too high” while none judged them too low. This is the item where student authority is weakest, and where my own framework applies most directly: a hard course that produced learning is not the same as an unfair one, and the near-zero link between satisfaction and learning is the reason I built the coding scheme I am using here. Even so, the strength and steadiness of the signal points at something I can fix without lowering the bar. Decide the curve-or-no-curve policy before the term, write it into the syllabus, and stop revisiting it mid-course, which is what turned a defensible grading choice into a running source of distrust.

What I take this to mean

Two things reassure me, and one worries me. The response rate was high, so I am not reading noise off a handful of loud voices. And the criticism, even at its sharpest, stayed on my teaching and my conduct. There were no comments about my identity or appearance, and nothing I would code as hostile; students engaged the work. What worries me sits underneath the average: the pain concentrated in assessment and in a stretch of the term when my exams and expectations kept moving, and a small number of comments describe a classroom that felt unpredictable.

The classroom incident

Several comments refer to a lecture in which I made a rash decision to add previously optional material to an upcoming exam in reaction to something a student said, then retracted it by email within the hour. I want to name it plainly rather than explain it away. Whatever I felt in the moment, and whatever the student meant by a hyperbolic, insensitive joke, using an exam as leverage in response to a comment was wrong. The fast retraction and the apologies I sent were the right immediate move, but they do not undo the fact that students left that room less sure of what the course would do next.

A comment about the Exam 1 curve, and a message I sent afterward, were read as defensive and dismissive. That tells me the lesson is broader: when a class is already under stress, my tone carries more weight than I assumed, and defensiveness can read as contempt even when I do not intend it. I have put guardrails in place. I pause before responding to a comment that lands as a provocation, and I read stress-humor as stress rather than malice. And I have fixed a rule that exam scope and course content follow the learning objectives only, never classroom dynamics.

What the record shows

A few of the sharpest comments claimed the class was broadly failing: that most students were doing poorly, that most would finish below a B, that no one was near the top. My own framework treats a reported fact as a perception until it is checked against the record, so I checked it.

The record does not match the claim. Across the four module exams, the averages sat in the mid-70s (75, 76, 72, and 76 percent, with a pooled median of 76.5 percent). The final grade distribution tells the same story: the median grade was a B-, a quarter of the class finished in the A range, 40 percent at B+ or higher, and two-thirds at B- or higher. Nine students out of seventy-five finished at D or F. One part of the complaint held up: no one earned an A+. That reflects a deliberate ceiling rather than weak work, since exam scores cap at 100 percent and the tier design concentrates difficulty at the top, so an A+ is hard by construction. A quarter of the class reached the A range anyway.

There was no grade inflation, which is the opposite of the problem of what most universities have been fighting. Harvard’s October 2025 report found that A’s had grown to about 60 percent of undergraduate grades, up from roughly a quarter in 2005. (In May 2026 its faculty voted to cap A’s at about 20 percent of students per course starting in fall 2027, calling the grading too compressed and too inflated.) ¹ My distribution did the opposite of compressing at the top: it spread across the full range with its center at a B-. That spread is the grading system doing its job, distinguishing levels of mastery instead of flattening them.

None of this puts the grading beyond criticism. The tier calibration had defects, including C-tier questions that were harder than A-tier ones, and I did not explain the ceiling clearly enough, so students met a distribution that was demanding by design and read it as unfair. Those are calibration and communication failures, and I own them. What the record does not support is the claim that students were failing.

The redesign

Over break, I have rebuilt the tiered system for Computational Biology, and the rebuild is mostly in how I explain it. The pitch now starts from a claim I left implicit in Biochemistry: in science, “knowing” is not binary, and an exam that mixes recall and analysis at random hands back a grade that does not tell a student what they mastered. So every question carries a visible tier. The D and C tiers test definitions and concepts, the B tier tests dynamic reasoning, the A tier tests multi-variable diagnosis, and the two S-tier questions are synthesis, offered as extra credit. I put the ceiling on the table in week one: master the D and C tiers and you land around 73, add the B tier and you land around 90, add the A tier and you reach 100. Because that ceiling sits below a conventional exam’s, I have shifted the letter-grade boundaries down three points so the letters track cognitive levels rather than an arbitrary percentage. The homework mirrors the exam’s tiered format and doubles as the study guide, and I keep the item review that returns points when a poorly worded question trips more than a third of the class. The machinery is the same machinery Biochemistry students met. The difference is that a student can see, from the first week, exactly what each question demands and what a given grade means.

What I am not changing

The feedback also tells me what to protect. The “why”-first lectures, the visual slides, the discussion board and office hours, the volume of practice, and the grade drops all drew steady praise, and several students who were hard on me still said they learned more in this course than in almost any other at Pitt. I am reading these evaluations the way my own framework asks me to: as a formative signal about how the course landed, not as a verdict on my worth as a teacher. The signal is clear enough to act on, and specific enough that I know where to start.