The short version
Your work is graded across three strands separately: Knowledge & Understanding (K/U), Thinking & Transfer (T/T), Communication (C). Each is scored on an 8-point scale. Your grade is reported as three numbers, like 7-6-5, not one. This tells you (and me) what you're strong at and what you need to grow.

1. Why strand-based grading

Most courses you've taken probably gave you a single grade per assignment: a number like 87, or a letter like B+, that summarizes how well you did overall. That format is convenient. It is also dishonest, because it hides what specifically you did well and what specifically you need to improve.

Consider two students. Student A writes a Position Paper with very strong scientific content, but the argument arrives out of order and the reader has to reconstruct it. Student B writes a Position Paper that is easy to follow from start to finish, but the scientific content is confused. Under a single-number rubric, both might land at 75: "Pretty good, room to grow." The single number tells neither student what to actually work on. Student A needs feedback on structuring an argument for a reader. Student B needs feedback on content depth. The single number erases the difference.

Strand-based grading separates these dimensions. Student A's grade might be K/U 7, T/T 6, C 4: strong content, decent reasoning, developing organization. Student B's grade might be K/U 4, T/T 5, C 7: developing content, decent reasoning, strong organization. Now both students can see exactly where to focus.

2. The three strands

K/U: Knowledge & Understanding

Your grasp of the scientific content: facts, concepts, models, mechanisms, terminology. Did you understand the science you were supposed to learn? Did you apply it accurately? Did you avoid or correct misconceptions?

T/T: Thinking & Transfer

Your ability to do analytic work: making connections, weighing evidence, constructing arguments, evaluating sources critically (OPVL), transferring knowledge from one context to another, designing solutions under constraints, taking perspectives that aren't your default.

T/T is where the course's defining work happens. The eight defenses, the OPVL throughline, the perspective-taking assessments. These are all primarily T/T. If you grew across the year, T/T is usually where you'll see it most clearly.

C: Communication

Your ability to communicate scientific work clearly for the intended audience. Writing structure, logical flow, appropriate scientific language, clarity for the reader who is NOT already inside your head, whether an oral defense holds its audience, visual representation quality where applicable.

What Communication measures, and what it does not. C is about whether your scientific reasoning reaches a reader: is it organized, is the scientific vocabulary used precisely, can someone who is not already inside your head follow the argument? It is not a score for grammatical accuracy, for range of English vocabulary, or for how native your English sounds. An error that does not obscure your meaning does not lower your band.

This is also why the strands are reported separately. A student still building academic English is scored on the science they know and the thinking they do, on their own terms. And because most deliverables can be written, recorded or drawn, nobody is assessed through the channel they find hardest.

3. The 8-point scale

Each strand is scored 1–8, with band descriptors paired (7-8, 5-6, 3-4, 1-2). Precision at the individual-point level is illusory: distinguishing a 6 from a 7 is meaningful; distinguishing a 6.3 from a 6.4 is not.

BandWhat this score means
7–8 ExceedingDemonstrates depth, nuance, and substantive engagement. 7 = solidly exceeding; 8 = exemplary, suitable as a worked example for future students.
5–6 MeetingDemonstrates the required understanding, skills, or quality. 5 = meeting; 6 = meeting with strength. This is the band most students hit on most work; it's not "average," it's "doing the work the course asks for."
3–4 DevelopingDemonstrates partial understanding or skill, with notable gaps. 3 = significant gaps; 4 = closer to meeting but still developing. Useful feedback signal: tells you where to focus.
1–2 BeginningLimited engagement with the criteria. 1 = minimal evidence; 2 = some evidence but substantial work still needed. Honest indicator that the work needs significant revision or that the underlying concepts need more time.

4. Reading your grade as a profile

Three strand scores together tell a richer story than any single number could:

Profile (K/U, T/T, C)What this tells you
7 – 7 – 7Strong across the board. Use the work as a model. Push yourself on the strand you find easiest: coast is the enemy of growth.
7 – 6 – 4Strong content, decent reasoning, developing communication. Focus on Communication: structure, audience awareness, scientific language precision.
4 – 5 – 7Developing content, decent reasoning, strong communication. Focus on K/U. Read more carefully; check facts; apply frameworks rigorously. Your communication is doing more work than your understanding right now.
6 – 4 – 6Decent content, developing reasoning, decent communication. Focus on T/T. Practice making claims supported by reasoning, OPVL on sources, engaging counter-arguments substantively.
6 – 7 – 5Decent content, strong reasoning, developing communication. Focus on Communication. Your thinking is sharper than your structure is letting through.
3 – 4 – 6Developing across content and reasoning; decent communication. The clear presentation is masking that the underlying engagement with the material is not yet there. Time to slow down and rebuild from the K/U side.

No profile is unambiguously "bad" or unambiguously "great." Every profile points to something specific to work on. That's the design intent: the rubric is a feedback tool, not just an evaluation tool.

5. The AI Documentation requirement

AI Documentation completeness
If you used AI at any stage in producing a deliverable AND your AI Documentation Template is missing or insufficient, the work is returned as incomplete and re-enters the reassessment process. It is not scored until the documentation is provided.

No strand is capped. Once the documentation arrives, every strand can score in any band. The requirement is about being able to tell whose thinking is whose, not about penalising AI use.

6. How strands work across assessment types

Different assessments weight strands differently because they ask different things. Each unit's assessment package specifies the strand focus explicitly. A few patterns worth knowing:

AssessmentTypical primary strand focus
Source Dossier (U0)Heavy on K/U (understanding what counts as a source) and T/T (OPVL substantively applied). C lighter: format is mostly templated.
Goldilocks Report (U1)Balanced K/U + T/T + C. K/U for habitability framework. T/T for the comparison reasoning. C for the report + Q&A.
Landscape Reading (U2)Heavy on T/T (reading a landscape as a system across time) + C (the share-out for a planning audience). K/U distributed.
Hazard Profile + Engineering Portfolio (U3)Heavy across all three. K/U for the geology + engineering principles. T/T for design reasoning + trade-off analysis. C for the Defense Panel.
Chennai Climate Brief (U4)Heavy on T/T (policy-relevant reasoning) + C (writing for a decision-maker). K/U baseline-required.
Position Paper + Roundtable (U5)Heavy on T/T (perspective-taking, counter-argument engagement) + C (written argument + oral defense). K/U for water-rights framework.
Sustainable Land Plan (U6)Heavy on T/T (multi-objective optimization, equity reasoning) + K/U (ecosystem services, biodiversity). C in the Planning Board defense.
Energy Mix + Op-Ed (U7)Heavy on C (public-audience writing is unusually demanding) + T/T (lifecycle reasoning, just-transition framing). K/U distributed.

Look at the rubric before you start the work, not after: knowing what's being measured changes how you approach the deliverable.

7. Unit 0 example: per-strand band descriptors

The general bands above are the framework. Per-strand band descriptors get more specific in each assessment package. Here's Unit 0's Source Dossier rubric as an example: every assessment has its own version of this:

K/U: Source Dossier specifics

7–8 Exemplary5–6 Proficient3–4 Developing1–2 Limited
OPVL applied thoroughly to every source. Source types correctly identified (peer-reviewed vs. news vs. corporate vs. AI). Discipline-specific vocabulary (origin, purpose, value, limitations, sycophancy, hallucination, training cutoff) used precisely. AI Documentation Template filled out completely and accurately. All four OPVL letters addressed for each source; minor imprecision in one or two sub-elements. Source types correctly identified. Vocabulary used accurately most of the time. AI Documentation Template substantially complete with minor gaps. Some knowledge of OPVL but with gaps. One or two letters consistently weaker than others (often Limitations is the weakest). Source-type identification sometimes imprecise. Vocabulary inconsistent. AI Documentation Template partially complete or unevenly applied. Little knowledge of OPVL. Multiple letters missing or treated superficially. Source types not distinguished or misidentified. Vocabulary largely absent or misused. AI Documentation Template missing or trivially completed.

T/T: Source Dossier specifics

7–8 Exemplary5–6 Proficient3–4 Developing1–2 Limited
OPVL evaluations identify substantive, specific limitations, not generic ones. The AI source's evaluation names at least one specific thing the AI got wrong, with evidence. Cross-source synthesis identifies non-obvious agreements and disagreements between sources of different types. Critical evaluation, not uncritical reporting. Most evaluations are substantive; one may be more generic than the others. The AI source's evaluation identifies a real limitation but may be less specific. Cross-source synthesis is present and accurate but covers expected rather than non-obvious points. Evaluations skew generic ("this source might be biased") rather than specific ("this source is funded by the manufacturer of the product it studies"). AI source's limitation is vague ("AI can be wrong") rather than specific. Cross-source synthesis is superficial: restating what each source says without comparing. OPVL evaluations are generic or absent. The AI source is treated as if it were just another reliable source, or no real limitation is identified. Cross-source synthesis is missing or merely lists what each source says with no comparison.

C: Source Dossier specifics

7–8 Exemplary5–6 Proficient3–4 Developing1–2 Limited
Extremely clear with a logical, easy-to-follow structure. Ideas fully developed and presented concisely. Scientific and OPVL-specific vocabulary used precisely throughout. Sources cited consistently and integrated effectively (not just listed). Defense well-paced; Q&A responses confident, substantive, and specific. Mainly clear with a mostly logical structure. Ideas mostly developed; concise enough for the audience. Vocabulary precise most of the time. Citations proficient with minor inconsistencies. Defense delivers content clearly; Q&A responses clear but may miss some depth. Somewhat clear; structure attempted but result not always easy to follow. Ideas generally developed but may lack concision or detail. Some vocabulary used correctly but with significant errors. Citations have errors or inconsistencies. Defense delivery uneven; Q&A responses partial. Unclear. Ideas underdeveloped. Beginner-level vocabulary. Citations missing or unreliable. Defense lacks structure; Q&A responses minimal or off-topic.

8. Using your grades to grow

The grade is the beginning of the next round, not the end. Specific moves that work:

  1. Read all three strand scores. Don't focus only on the lowest or only on the highest. The profile is the information.
  2. Find the strand-specific feedback I wrote. "This section confuses cause and effect" is usually T/T; "The atmospheric retention claim is incorrect" is usually K/U; "This paragraph buries the lede for a policy audience" is usually C.
  3. Pick ONE strand to focus on for the next assessment. Trying to improve all three simultaneously usually means improving none.
  4. Ask me about a specific score if you don't understand it. Bring the work + the rubric + the score: "I expected X here; you scored Y. What did I miss?" Sometimes you'll convince me to revise; sometimes I'll convince you the reasoning held. Either is a real conversation.
  5. Re-read your old work after each new assessment. The patterns are visible only across multiple deliverables. "I keep losing points on T/T for not engaging counter-arguments" is a pattern; one missed counter-argument is a moment.

9. Common questions

Why isn't my grade reported as a single number?

Because a single number hides what the three strands tell you separately. If you need a single number for some external purpose (transcript, college application), the strands can be mapped to a percentage, that mapping is documented separately. But your primary feedback is always the three strands.

Does the lowest strand drag down the others?

No. The strands are reported separately. A 4 on Communication does not lower your K/U or T/T scores. They are independent measurements.

What if I think the rubric is unfair to me?

Bring the specific concern. "This rubric expects format X but I produced format Y because Z: should that be a different grading conversation?" is a legitimate question. Sometimes I'll agree the rubric needs revision; sometimes I'll explain why the rubric is the right one. Both are productive.

Does AI use affect my grade beyond the documentation requirement?

Hidden AI use is academic dishonesty, see Class Expectations. Beyond that, disclosed AI use is treated as scholarship. Use AI heavily and document well; you'll be fine. Use AI heavily and document poorly; the K/U gate fires.

10. One last note

Strand-based grading takes a few assessments to get used to. By the third or fourth, you'll start reading the profile like a doctor reads a blood-test panel: these three numbers tell different stories that together describe what's happening.

By the end of the year, you should be able to predict your own profile before you submit. "I think this is K/U 6, T/T 5, C 6, I rushed the counter-argument paragraph." When your self-predicted profile matches my scoring within a point on each strand, you've internalized the rubric, and at that point, you don't really need me to grade you anymore. You can do it yourself.