Do Now

Terminology and Concepts to look out for

claim · evidence · limit

Relevance to Me

Anybody can say which one they liked. Today you write the frame around your body from yesterday — the introduction that opens the report, and the conclusion that survives somebody disagreeing with you.

Board Question

Based on your observations last week, and your work yesterday, what is one claim you feel confident making about the performances of the different models? For example, "There really was no difference in generated responses of the models when it came to __________..." Or, "Surprisingly, the Mistral ____ model actually performed better / as well / worse than the others on..." Write it in your notebook and commit to it.

The rubric is on the course site and on the board.

An Introduction Sets Up the Question; A Conclusion Answers It

Yesterday you wrote the middle. Today you write the ends.

An introduction tells a reader what the lab was trying to find out and how it was set up, before you tell them what happened. It answers three questions in plain sentences: what were we investigating, what were the three models, and what was kept the same each day so we could see what changed. Someone who was not in this room for those four days should be able to read your intro and know what they are about to read a body about.

A conclusion has three parts. Most people write one of them and stop.

A claim is what you say is true. The largest model was the most reliable of the three. That is a claim. It could be wrong, which is what makes it worth writing.

Evidence is what holds it up, and it has to be specific enough that a reader could go check it. Pane 3 got three out of three on Tuesday and four out of four on Wednesday. Not pane 3 did well. Numbers, days, counts.

A limit is what your evidence cannot reach. This is the part that gets skipped, and it is the part that separates a report from an opinion. I cannot say pane 3 is more reliable in general, because we ran four prompts on four days and all four came from the same class.

Now the thing you need to hear before you write a word.

A result that fell apart scores full marks if you report it honestly.

Some of you ran Day 4 and watched Wednesday's ranking collapse. A different pane came first each time. It felt like the lab broke. It did not. You found out that four criteria and one run each was not enough to separate three models, and that is a real finding about your measurement. Writing that down plainly earns you the same score as a ranking that held.

What loses points is deciding your ranking held when your own numbers say otherwise. Your counts are in your body from yesterday. A reader can check them against what you claim. If those two things disagree, the report fails on the row that matters most.

The order for today. Three tabs on the right: Intro, Body, Conclusion. The Intro form first, because the body you finished yesterday needs a frame before the conclusion can point at it. Then the Conclusion form, which holds the reflection. The Body tab is there so your conclusion can point at real numbers. You are writing a paper. Complete sentences. One last question sits in the Conclusion form and it is the one worth thinking about: across four days, which changed more, the three models, or what you meant by the word better?

The models never changed. Not once. They were frozen before you met them.

I Do · pick an answer, then Check
Question 1notebook cue: claim
"The largest model was the most reliable of the three" is
Question 2notebook cue: evidence
Which of these is evidence for that claim?
Question 3notebook cue: limit
A limit is
Question 4worked aloud
Your Day 4 runs came back with a different pane first each time. What happens to your grade if you write that down plainly?
100%
Lab 1 Close-out, Day 2 — Thursday, September 10 · How to tell which one is better, and how to say why.

Lab Report 1 — rubric

Six rows, zero to four each, twenty four points. This is what a 4 asks for.

  1. The record. Four days submitted, each pointing at the right run, with station data present. Response text is pulled from the server and is not the student's responsibility.
    All four days submitted. Station and run number correct on all four, so all four resolve to real runs.
  2. Description without judgment. Day 1. Statements another reader could check against the response.
    Length, structure, specificity, and hedge counts all present with examples. No ranking language anywhere in Day 1.
  3. Verification. Day 2. Every correctness claim carries the source it was checked against.
    All three questions carry a named source and the answer found there. Counts are raw and match the responses recorded.
  4. Consistency and stability. Days 3 and 4 together. The same criterion applied the same way, and the ranking tested against it.
    All four criteria applied to all three panes with the deciding sentence quoted. Day 4 recopies Wednesday's ranking, reports both new runs, and states plainly whether it held.
  5. Setup. Introduction. A reader who was not here can follow the lab.
    Names the question, the three models and how they differ, what was held, and what changed each day. Includes a prediction made before the lab.
  6. Argument. Reflection and conclusion. A claim, the evidence under it, and an honest limit.
    Claim stated. A specific day and specific numbers cited to support it. One thing named that the lab cannot show, with the evidence that would be needed. The change in the meaning of better is traced across the four days.

A negative result scores full marks.

A student whose ranking collapsed on Day 4 and who says so plainly earns a 4 on rows 4 and 6. Students will assume the opposite. Say this out loud on Wednesday, before they start writing, or half of them will quietly fabricate stability.