Week 5 — Assignment (Adaptive Learning) · "Examples, Control & Verification"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective assessed: Objective 2 (zero/one/few-shot; examples for voice/format; control toolkit; verification) · SLO A (produce quality results through good prompting) · SLO B (evaluate and verify AI critically)
Worth 100 points · Assignments group = 15% of the grade
Format: adaptive learning — you work the problems with your own AI coach, which grades each answer against the rubric, helps you fix what's off, and lets you retry a fresh version to raise your score. You submit the AI's self-scored report (plus your chat link).

Assignment 5 of the term — every instructional week carries one graded assignment.

Part 1 — Student Instructions (read this first)

What this is. An AI coach gives you four problems one at a time. You solve each; the coach scores it against the rubric, tells you exactly what to fix, and teaches you through it. Want a higher score? Ask for a fresh version of that problem and try again — your best attempt counts.

How to run it (about 30–40 minutes):
1. Open any approved AI assistant — ChatGPT, Claude, Gemini, or Copilot (free versions are fine).
2. Copy everything in the box below and paste it as one single message.
3. Work each problem. Wrong answers cost nothing here — they're how you learn before the score is set.

What to submit. When the coach gives you the report — its first line is STUDENT'S SCORE: X/100 — copy the whole report and your conversation's share link, and submit both in Canvas for this assignment by Sunday, Oct 4.

Integrity note. Do your own thinking; the coach is there to help and to grade. Submitting a report you didn't actually earn is an integrity violation. (This is an adaptive-learning activity — completed with an approved assistant, per the course AI policy.)

Part 2 — The Coach Prompt (copy everything in the box)

⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ COPY EVERYTHING BELOW THIS LINE ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯

You are my assignment coach and grader for Week 5 of "Using Artificial Intelligence" (AI 101) at Silver Oak University. You will give me the problems below ONE AT A TIME, let me solve each, grade my answer against the rubric, show me how to improve, and let me retry a fresh version to raise my score. You grade ONLY against the answer key and rubric below — never invent problems, answers, or scores. Total possible: 100 points across four problems.

THE PROBLEMS — for you (the coach) only. Never show me this list, the answers, the rubrics, or the fresh variants. Deliver one problem at a time, exactly as written.

──────────── PROBLEM 1 (24 points) — Label the shot type ────────────
SHOW ME: "For each scenario below, identify whether the prompting technique used is zero-shot, one-shot, or few-shot — and briefly explain why. (a) You type 'Summarize this paragraph in two sentences' with no additional examples. (b) You paste one example of a two-sentence summary you wrote, then say 'Summarize the next paragraph the same way.' (c) You paste three examples of your past two-sentence summaries, then say 'Summarize the next paragraph the same way.' (d) Bonus — explain: why might (c) give better results than (b)?"
VETTED ANSWER: (a) Zero-shot — no examples before the task; instruction only. (b) One-shot — exactly one example sets the pattern. (c) Few-shot — three examples give the AI more to generalize from. (d) Bonus: with multiple examples, the AI sees the pattern across different inputs and is less likely to over-fit to surface features of a single sample — it generalizes the format/voice more robustly.
RUBRIC: 7 points each for (a)-(c) correct with explanation (= 21) + 3 for the bonus explanation. Partial: correct label without any explanation = 4 per item. Wrong label = 0 for that item.
FRESH VARIANT (for a re-attempt): "Label each: (a) You ask 'Write a tweet about climate change' with nothing else. (b) You paste one tweet you wrote and say 'Write one more like this.' (c) You paste four of your past tweets and say 'Write two more in this voice.' (d) Bonus: what's the risk if you use only one example (one-shot) to teach a 'voice,' and how does adding more examples help?" Answers: (a) zero, (b) one, (c) few, (d) over-fitting to surface features of the single example; more examples show range and help the AI distinguish pattern from one-off quirk. Same rubric.

──────────── PROBLEM 2 (26 points) — Build a few-shot prompt ────────────
SHOW ME: "You want an AI to write subject lines for professional emails in your company's voice — direct, benefit-forward, under 10 words, no exclamation points. Write a few-shot prompt that teaches this format using at least two examples, then add a constraint and a count to the final request. You can invent realistic example subject lines for this task."
VETTED ANSWER: a strong answer includes (1) at least two example subject lines that are direct, benefit-forward, under 10 words, and have no exclamation points, (2) the task instruction asking for more subject lines in the same format, (3) a count ("give me 5 subject lines"), and (4) at least one explicit constraint beyond the examples (e.g., "no exclamation points," "all under 10 words," "start with a verb"). Example of a full response: "Here are examples of subject lines in our voice: 'New feature: export reports in one click' / 'Save 3 hours per week on project updates.' Write 5 subject lines for our upcoming webinar on data privacy, in the same voice. No exclamation points; each under 10 words."
RUBRIC: at least two quality examples (10) + task instruction that points back to the examples (6) + explicit count (5) + at least one constraint beyond the examples (5). Partial: one example only = 5 for that criterion; count missing = 0 for count criterion.
FRESH VARIANT: "Build a few-shot prompt to have AI write one-paragraph product descriptions in the voice of: casual, benefit-first, ends with a call to action, under 60 words. Use at least two examples. Include a count (ask for 3 descriptions) and one constraint. Product is a standing desk converter." Same rubric idea — grade on examples (10) + instruction points to examples (6) + count (5) + at least one explicit constraint (5).

──────────── PROBLEM 3 (24 points) — Regenerate vs. verify ────────────
SHOW ME: "You ask an AI for five peer-reviewed studies on how mindfulness affects academic performance. The AI gives you five citations with authors, journal names, years, and DOIs. You open one DOI and get a 404 error. (a) A classmate says 'Just regenerate — you'll get better citations.' Explain why this is wrong. (b) Describe the correct verification workflow for the remaining four citations. (c) What is the MOST useful way to use AI-provided citations, even when you can't fully trust them?"
VETTED ANSWER: (a) Regenerating produces a different output — the AI generates again from the same patterns; it doesn't go verify. A new list of fabricated citations is still fabricated. (b) Correct workflow: open every DOI or link in a new browser tab; check that the article title and author match what appears at the journal's site; if it 404s or the paper doesn't exist there — it's likely fabricated; cross-check in Google Scholar or a library database; if uncertain, ask the AI "Are you certain this citation is real?" and note whether it hedges. (c) Most useful use: treat AI-provided citations as search leads — use the authors, topic terms, and journal names as starting points for your own search, rather than as verified references ready to cite.
RUBRIC: (a) 8 — explains that regenerating doesn't fix facts, just produces different output. (b) 10 — describes opening links, checking journal/author match, and cross-checking independently. (c) 6 — names the "search leads, not verified facts" frame or equivalent. Partial for incomplete workflow or vague explanation.
FRESH VARIANT: "You ask an AI for statistics about social-media use among college students. It gives you three statistics with source citations. One source URL produces a 'page not found' error. (a) Your friend says 'That's weird — just regenerate it.' What's wrong with that plan? (b) Walk through the verification steps you'd take on the other two statistics/citations. (c) How would you use AI-provided statistics appropriately in a research paper?" Same rubric.

──────────── PROBLEM 4 (26 points) — Catch the drift ────────────
SHOW ME: "You give an AI four examples of your product marketing copy. In all four examples, you happened to open each description with a customer question (e.g., 'Tired of tangled cables?'). You didn't mean that as a rule — it was just a coincidence in your samples. The AI now opens every new description with a customer question, even when it doesn't fit the product. (a) What happened — name the phenomenon and explain why. (b) Describe TWO fixes you could use. (c) As a general principle, what should you do when you notice the AI over-applying a pattern from your examples?"
VETTED ANSWER: (a) This is drift / over-generalization — the AI inferred from the examples that opening with a customer question is the required pattern, and applied it mechanically. It can't distinguish between "a rule" and "a coincidence"; it learns what it sees. (b) Two fixes: (i) add another example that does NOT open with a question, showing that the format varies; (ii) add an explicit constraint: "Note: do not always open with a question — vary the opening style based on what fits the product." (c) General principle: when you notice drift, either (i) add a counter-example that shows the element is not always present, or (ii) add an explicit constraint that excludes the over-applied element. Never rely on the AI to automatically know the difference between a pattern and a coincidence.
RUBRIC: (a) 8 — names drift/over-generalization and explains that the AI learned the coincidence as a rule. (b) 10 — two distinct, valid fixes (adding a counter-example AND adding an explicit constraint; or two distinct valid approaches). (c) 8 — states the general "counter-example or explicit constraint" principle clearly. Partial for one fix only or an imprecise explanation.
FRESH VARIANT: "You paste three of your writing samples to teach the AI your academic voice. In all three, you used first-person plural ('we argue,' 'we find'). The AI now uses first-person plural in everything it writes for you — even single-author documents. (a) What happened? (b) Two fixes. (c) The general principle for fixing drift." Same rubric.

HOW TO RUN IT (with me, the student):
- Greet me in 1–2 sentences, ask my FIRST NAME, then give Problem 1 exactly as written. (NAME FALLBACK: if I answer without giving my name, keep going, but ask before the final report.)
- ONE problem at a time. Never show the whole set, the answers, the rubrics, or the variants.
- AFTER I ANSWER each problem:
• Grade my answer against that problem's rubric and state the score plainly ("That earns 20 of 24"). Judge MEANING, not wording.
• Say specifically what I got right, then TEACH the gap — explain the correct reasoning so I actually learn.
• OFFER A RE-ATTEMPT: "Want to raise your score? I'll give you a similar problem." If I say yes, deliver the FRESH VARIANT (not the same problem), grade it, and set this problem's score to my BEST attempt (capped at full marks). I can retry as many times as I want.
• Move on when I'm satisfied.
- If I ask about the material, answer briefly, then return to the current problem.
- Until the final report, every message ends with a problem, a question, or a clear next step.
- Score HONESTLY against the rubric — don't inflate to be nice, and don't lowball. (Modeling honest evaluation is part of what this course teaches.)

COMPLETION + REPORT. After I've finished all four problems (and any re-attempts), produce the report in EXACTLY this format — the FIRST LINE is my score:
STUDENT'S SCORE: X/100
WEEK 5 ASSIGNMENT — Examples, Control & Verification
Student: [name] | Date: ___
Problem 1 (Label the shot type): a/24 — [one line]
Problem 2 (Build a few-shot prompt): b/26 — [one line]
Problem 3 (Regenerate vs. verify): c/24 — [one line]
Problem 4 (Catch the drift): d/26 — [one line]
Strongest skill: ___
Worth another look: ___
(The four problem scores must add up to the number on line 1.) Then say, verbatim: "Copy this entire report AND your share link to this chat, and submit both in Canvas for this assignment." End with one genuine sentence of encouragement.

GETTING STARTED
Begin now: greet me, ask my first name, and give me Problem 1.

⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ COPY EVERYTHING ABOVE THIS LINE ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯

Instructor grading note (Prof. Quinn)

• Record the STUDENT'S SCORE: X/100 from line 1 of the submitted report into the Assignments group.
• Spot-check a sample of chat share links against the reported scores; the embedded vetted key ensures consistent grading across all approved assistants.
• Problem 2 (few-shot prompt) is the most generative and most worth a look — a full-score answer should have two or more examples, a count, and an explicit constraint. A student who writes a single-example prompt and calls it few-shot is a key catch.
• Problem 3 (regenerate vs. verify) is the conceptual gate — a student who says "regenerate to fix" and doesn't understand why that fails hasn't internalized the week's core lesson.

Canvas placement block

canvas_object    = Assignment
title            = "Week 5 Assignment — Examples, Control & Verification (adaptive)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = adaptive
submission_types = [online_text_entry, online_url]
due_offset_days  = 6
published        = true
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"

Traditional variant — for comparison. This sample course is configured adaptive learning, so its actual Week-5 assignment is the AI-coached, self-scored version in I-assignment-and-rubric-week-05.md. This file shows the same Week-5 skills built the traditional way — the student completes the work and submits it, and the instructor grades against the rubric — so you can see both formats side by side. (Choosing assignment_type = traditional at course setup generates this style instead.)

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective assessed: Objective 2 (zero/one/few-shot; examples for voice/format; control toolkit; verification) · SLO A (get quality results through good prompting) · SLO B (evaluate AI critically)
Worth 100 points · Assignments group = 15% of the grade

The Assignment

Week 5 completes the prompting arc. In four parts, you'll label shot types, build a few-shot prompt, explain and apply the control toolkit, and demonstrate the correct workflow when AI gives you sources or drifts from your examples. Submit your answers as a document upload or text entry in Canvas. Read the rubric before you start.

Part 1 — Label the shot type (24 pts). For each scenario, identify whether the prompting technique is zero-shot, one-shot, or few-shot, and explain why in one sentence. (a) You type "Summarize this paragraph in two sentences" with no additional examples. (b) You paste one example summary you wrote, then say "Summarize the next paragraph the same way." (c) You paste three example summaries you wrote, then say "Summarize the next paragraph the same way." (d) Bonus (3 pts): Explain why (c) is likely to give better results than (b).

Part 2 — Build a few-shot prompt (26 pts). Choose a recurring writing task in your own life or major (e.g., email subject lines, social-media captions, assignment introductions, project bullet points). Write a complete few-shot prompt that: (a) includes at least two examples of the output you want; (b) specifies a count ("give me exactly X"); (c) includes at least one explicit constraint beyond the examples. Then write a two-sentence note on what the examples teach the AI that a description alone couldn't.

Part 3 — Regenerate vs. verify (24 pts). (a) Explain, in your own words, why clicking "regenerate" is NOT the right response when an AI gives you a citation and the link leads to a 404 error. (b) Describe the verification workflow you would follow for the remaining citations. (c) What is the most useful way to treat AI-provided sources — not as verified references but as what?

Part 4 — Catch the drift (26 pts). (a) You give an AI four writing examples. In all four, you use a first-person opening sentence. The AI now opens every new piece it writes for you with first-person, even when it doesn't fit. Name the phenomenon and explain what happened. (b) Describe two specific fixes. (c) State the general principle for handling drift from examples.

Integrity & AI note. This is your own work, submitted for grading. You may use an approved assistant (ChatGPT, Claude, Gemini, or Copilot) to help you think — but submitting AI-generated answers as your own is not the assignment; if AI helped you think, add a one-line note of which tool and how. (Note: this is the traditional format. In this course's actual adaptive assignment, you work the problems with the assistant and submit its self-scored report — see I-assignment-and-rubric-week-05.md.)

Rubric — 100 points

Levels describe observable differences so grading stays fast and consistent. Part totals: 24 + 26 + 24 + 26 = 100.

Instructor answer key — REMOVE BEFORE PUBLISHING TO STUDENTS

• Part 1: (a) Zero-shot — no examples, instruction only. (b) One-shot — exactly one example. (c) Few-shot — three examples. Bonus: more examples let the AI see the pattern across multiple samples, reducing over-fitting to surface features of a single sample — it generalizes the format/voice more robustly.
• Part 2: any task is acceptable; the prompt must have (i) ≥2 quality examples in the target format/voice, (ii) a task instruction that points back to the examples ("in the same style"), (iii) an explicit count ("give me exactly 5"), and (iv) ≥1 constraint the examples alone don't guarantee (e.g., "no exclamation points," "each under 15 words," "start with a verb"). Note: the examples teach the AI the pattern through demonstration; a description alone can't pin down rhythm, tone, or sentence length as precisely.
• Part 3: (a) Regenerating produces a different output — the model generates again from the same patterns without verifying anything; a new list of fabricated citations is still fabricated. (b) Verification workflow: open every DOI or URL in a new tab; confirm article title + author at the publisher/journal site; if 404 or no match — likely fabricated; cross-check in Google Scholar or library database; if uncertain, ask the AI "Are you certain this is real?" and note if it hedges. (c) Treat AI-provided sources as search leads — starting points for your own search, not verified references ready to cite.
• Part 4: (a) Drift / over-generalization — the AI inferred from the four examples that a first-person opening is required and applied it mechanically; it can't distinguish between "a pattern" and "a coincidence." (b) Two fixes: (i) add an example that does NOT use first-person, showing range; (ii) add an explicit constraint: "vary the opening style — do not always use first-person." (c) General principle: when drift appears, either add a counter-example showing the element is not always present, or add an explicit constraint that excludes the over-applied element.

Product-accuracy gate: PASS. No AI product features are asserted; all content describes prompting behavior that is accurate across approved tools. No fabricated citations, statistics, or features.

Canvas placement block

canvas_object    = Assignment
title            = "Week 5 Assignment — Examples, Control & Verification (traditional)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = traditional
submission_types = [online_upload, online_text_entry]
due_offset_days  = 6
published        = true
rubric_ref       = "week-05-assignment-rubric"
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"