Skip to content

Grading, and What Happens When You Actually Teach

The last stretch of KTCP covers grading philosophy, then puts everything from the previous workshops to the test: write a real syllabus, teach a real (if short) lesson to peers, and see what survives contact with an actual classroom. This closing post covers both — the argument against conventional grading, and what I learned from watching my own teaching not go the way I'd planned.

Grading is a technology, not a measurement

The most useful reframing in this whole program came from Schinske & Tanner's history of the A–F grading scale: it's an administrative invention from the 1940s, built for record-keeping, not a pedagogical tool anyone designed on purpose. Treated as a scientific measurement, it falls apart under scrutiny — inter-rater agreement on written work is famously poor, grades measurably dampen intrinsic motivation, and once a grade is attached to a piece of feedback, most students stop reading the feedback at all.

McClymer & Knoles push the critique further with a phrase I keep coming back to: ersatz learning. Students can pass a course by pattern-matching the specific, contrived shape of an exam question, without that skill having any real connection to how a practitioner in the field actually thinks. The fix isn't a better test of the same kind — it's an authentic task, one that mirrors what someone in the discipline actually does with the material.

Nilson's proposed alternative, specifications grading, was the most concretely adoptable idea in the whole reading list:

Traditional (points/curve) grading Specifications grading
Unit of grading Partial credit, averaged across assignments Pass/fail against an explicit, rigorous bar
What a grade means An average of unrelated skills, hard to interpret Which specific outcomes were actually demonstrated
Recovering from a bad week Averaged in permanently Tokens for a late pass or a revision
Final course grade Sum of points Bundles of outcomes met
Ersatz-learning risk High — pattern-matching to the curve is rewarded Lower — the bar is the same regardless of the curve

I reinvented a piece of this before I knew its name

My own note from that workshop, written before I'd gotten to the Nilson reading, was simply: let students redo and resubmit graded work for a make-up score. It's the same underlying mechanic — grading as something a student can recover from, rather than a single irreversible measurement — arrived at independently because the problem (a system that punishes a bad week more than it measures actual understanding) was obvious enough to reinvent the solution.

Designing the syllabus, then living it imperfectly

The capstone folded everything together into one artifact: a full syllabus, evaluated against a rubric whose criteria read like a checklist of this whole series:

  • [ ] Course-level learning outcomes use explicit, action verbs (not "understand")
  • [ ] Content is organized around real questions/themes, not a flat topic list
  • [ ] Each summative assessment has a stated rationale and is authentic to the discipline
  • [ ] Assessments are paced across the term, not backloaded onto one final
  • [ ] Tone is inclusive and growth-oriented (personal pronouns, positively-framed policies)
  • [ ] High expectations are paired with visible support (office hours, revision paths)
  • [ ] Policies are explained as pedagogy — why they help, not just what happens if you break them
  • [ ] The syllabus itself explains the backward-design logic behind the unit order

Every item on that list traces back to a post earlier in this series — the rubric isn't a new idea, it's the same design habit applied to one final document.

Mine was for a course called Earth System Modeling, built in exactly the shape the first post in this series describes: four units, each adding one subsystem to a computational model of the Earth's climate, building incrementally toward a capstone project where a student proposes a scenario, runs it, validates the result, and presents it.

Then came microteaching — two short practice lessons in front of peers, with feedback in between. The gap between designing well and teaching well turned out to be its own lesson.

flowchart LR
    A["Microteaching 1
    ✅ relatable hook
    ✅ logical scaffolding"] -->|"peer feedback:
    color-code equations,
    fix table order,
    tighten pacing"| B["All 3 fixes applied"]
    B --> C["Microteaching 2
    ⚠️ denser slides
    ⚠️ verbal-only computation
    ⚠️ lower energy that day"]
    C --> D["Net result: worse overall,
    despite every fix landing"]

My first session got specific, actionable critique: color-code the equation terms to match the diagram, tighten the pacing on a derivation, keep a results table in a consistent order. I fixed all of it for the second session. And the second session was, by my own and my peers' assessment, worse overall — the slides were denser, I explained computation steps verbally instead of showing them, and the room felt less engaged.

The honest diagnosis wasn't a design flaw at all — it was that I was tired that morning, and it showed in my delivery fluency in a way no amount of slide preparation could compensate for. That was worth sitting with: teaching quality isn't additive. You can fix every item on a feedback list and still have a worse session, because delivery is holistic in a way a checklist can't fully capture. The forward-looking fix wasn't "prepare more content" — it was structural: chunk a lesson into roughly fifteen-minute blocks, decide in advance what's safe to cut if energy or time runs short, and keep leaning on the scaffolding and active-recall techniques from earlier in this series rather than trying to carry a session on delivery alone.

Looking back at the whole arc

Across these four posts, the throughline is really one habit applied at different scales: define the outcome you actually want, build the smallest honest way to check whether you got there, and only then decide what to do — for a course, for a unit, for a single piece of feedback, for a grading policy. It's the same discipline I now reach for outside a classroom entirely, any time I'm tempted to start with the interesting technique instead of the question it's supposed to answer.