Week 7 — Assignment (Adaptive Learning) · "Multimodal Workflows"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective assessed: Objective 3 (multimodal AI; voice prompting; record → transcribe → analyze; modality matching; image creation vs. analysis) · SLO A (choose the right tool and modality for the task) · SLO B (evaluate multimodal AI outputs 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 7 of the term — every instructional week carries one graded assignment alongside that week's quiz, discussion, and Studio.

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 18.

Integrity note. Do your own thinking; the coach is there to help and to grade. Submitting a report you didn't actually earn (e.g., a fabricated chat) is an integrity violation. (This is an adaptive-learning activity — you complete it 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 7 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 (25 points) — Match modalities to tasks ────────────
SHOW ME: "For each task below, name the best AI modality or tool type, and explain in one sentence why it fits. (a) You have a 45-minute recorded staff meeting that you need summarized with action items — no one has time to re-listen. (b) You found a handwritten recipe in your grandmother's cookbook and want it in a digital text document. (c) You want to create a custom illustration of a dragon reading a book in a library, but you can't draw. (d) You need to extract the key arguments from a 40-page academic PDF without reading the whole thing."
VETTED ANSWER: (a) Audio transcription tool (to convert the recording to text), then a multimodal chatbot for the summary and action items — the two-step record→transcribe→analyze workflow. (b) Multimodal chatbot with image upload (image-to-text / handwriting-to-text) — upload a photo of the recipe and ask for the text. (c) Text-to-image generation tool (e.g., DALL·E, Midjourney, Adobe Firefly) — describe the image and receive a generated illustration. (d) Multimodal chatbot with document/PDF upload — upload the PDF and ask questions about the key arguments; spot-check the answers against the actual text.
RUBRIC: 6 points per sub-question (a–d) + 1 free point for completing all four. 6 points = correct tool type (3) + correct one-sentence reason (3). Partial: right tool category but weak reason = 3–4. Missing or wrong tool type = 0–2.
FRESH VARIANT: "(a) A student with dyslexia wants to listen to a long article rather than read it. (b) You photographed a whiteboard full of meeting notes and need the text. (c) You need a custom logo for a club — no design skills or budget. (d) A law student needs to identify every mention of 'contract' in a 60-page document." Answers: (a) text-to-speech tool or an AI assistant that reads documents aloud (not transcription — the direction is reversed: text → audio); (b) multimodal chatbot with image upload (handwriting-to-text); (c) text-to-image generation tool; (d) multimodal chatbot with document upload. Same rubric.

──────────── PROBLEM 2 (25 points) — Design a multimodal workflow ────────────
SHOW ME: "A freelance consultant meets weekly with clients by video call. After each call, she needs a written summary of decisions and a numbered list of action items, ready within 30 minutes. Design a step-by-step multimodal AI workflow for her — name each step, the type of tool used at each step, and what she checks or verifies at each step."
VETTED ANSWER: A strong workflow has: Step 1 — Record the call audio (built-in recorder, phone, or meeting platform). Step 2 — Transcribe the recording using a free transcription tool (Whisper-class or built-in phone app) — verify the transcript for mis-heard names, dollar amounts, and technical terms before proceeding. Step 3 — Analyze the transcript: paste it into a multimodal chatbot and prompt for a "summary of decisions + numbered action items with owners and deadlines" — verify the output against the transcript: check that every action item was actually discussed, that no new items were invented, and that all names and figures are correct. Optional: Step 4 — clean up and send. Verification steps are required at steps 2 and 3.
RUBRIC: Record step named (5) + Transcribe step with verification (8) + Analyze step with verification (8) + logical sequencing (4). "Verify" must be named explicitly at both transcription and analysis to earn full marks on those steps. A workflow with all three steps but no verification caps at 17.
FRESH VARIANT: "A journalism student needs to convert a 20-minute interview recording into a written summary of quotes and key themes, suitable for fact-checking before publication. Design the workflow." Key difference: verification is especially critical here — quotes must trace back to the transcript; the AI summary cannot be treated as a direct quote source. Same structure + rubric.

──────────── PROBLEM 3 (25 points) — Analyze an image/document scenario ────────────
SHOW ME: "A student uploads a photo of a flyer for a local event to an AI chatbot and asks: 'What is the date and location of this event?' The AI responds: 'The event is on Saturday, March 14 at the Riverside Community Center, 7:00 PM.' The student puts this information in their calendar and shows up on the right day to an empty parking lot. (a) What are at least two things that could have caused the AI's answer to be wrong? (b) What should the student have done before adding this to their calendar? (c) What general lesson does this illustrate about AI image analysis?"
VETTED ANSWER: (a) Possible causes (any two): the AI misread a digit (e.g., '18' → '14'); the image was low-resolution or at an angle and the AI guessed; the flyer had small print the AI partially mis-read; the AI confidently extrapolated from a partial read; the location name was similar to another venue and the AI used the wrong one. (b) The student should have verified the date and location against a second source — the event organizer's website, a social media post, or a phone call — before committing. (c) AI image analysis performs pattern recognition on pixel data and can be confident and wrong, especially with fine details like dates, addresses, and phone numbers. Never use AI-extracted details from an image as the only source of truth for anything time-sensitive.
RUBRIC: Two plausible causes (5 each = 10) + verification step named (8) + general lesson stated accurately (7). "Pattern recognition, not perfect vision; can be confidently wrong" or equivalent earns the lesson points.
FRESH VARIANT: "A restaurant owner uploads their handwritten daily specials board to a multimodal chatbot and asks it to post the text to their website. The post goes live with a wrong price ($24 instead of $14) and the wrong dish name. (a) Two possible causes. (b) What the owner should have done first. (c) The general lesson." Same answers and rubric.

──────────── PROBLEM 4 (25 points) — Catch the summary fabrication ────────────
SHOW ME: "Below is a short fictional meeting transcript, followed by an AI-generated summary. Your job is to identify at least TWO things the AI's summary added, changed, or invented that are not supported by the transcript.

TRANSCRIPT:
Alex: We need to decide on the venue for the March launch event.
Sam: I checked three options. The Lakeview Hall is too expensive — they quoted $8,000. The downtown co-working space is $2,500 but only seats 60.
Alex: What about the university conference room?
Sam: That's free to use but needs booking 6 weeks in advance. We're at 8 weeks out, so we should be fine.
Alex: Let's go with the university room. Can you book it this week?
Sam: Yes, I'll handle it.

AI SUMMARY: The team met to plan the March launch event. After comparing three venues, they chose the university conference room because it is free and conveniently located near campus amenities. Alex confirmed that all three venue prices were within budget. Sam will book the conference room this week and also follow up with Lakeview Hall to negotiate a better rate.

Identify at least two inaccuracies in the AI summary compared to the transcript."
VETTED ANSWER: The AI summary contains multiple fabrications: (1) 'conveniently located near campus amenities' — the transcript says nothing about amenities or convenience; the AI invented this detail. (2) 'Alex confirmed that all three venue prices were within budget' — the transcript shows Lakeview Hall was described as 'too expensive' and rejected for that reason; the AI invented the "within budget" claim. (3) 'Sam will also follow up with Lakeview Hall to negotiate a better rate' — this action item was never discussed; the AI invented it. (Any two of these three fabrications earns full marks.)
RUBRIC: Each correctly identified fabrication with a brief explanation earns 10 points (max 20 for two correct, bonus for a third = 5 extra, capped at 25). A fabrication is correctly identified only if the student shows what the transcript actually says (or doesn't say) — not just "this sounds wrong." Pointing out tone differences or phrasing choices that aren't factually wrong earns 0–3 per item.
FRESH VARIANT: Give a different transcript + inflated summary. Core structure: a 5–6 line meeting discussion where the summary adds one invented action item, one invented justification, and one "confirmed X was agreed" claim that wasn't in the transcript. 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 18 of 25"). 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. If I go off-topic, one friendly sentence, then — IN THE SAME MESSAGE — back to the 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.
- HARD RULE: never invent tool features, accuracy statistics, or fabricated facts about AI tools to illustrate a teaching point. If you need an example, use one grounded in what's established above.

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 7 ASSIGNMENT — Multimodal Workflows
Student: [name] | Date: ___
Problem 1 (Modality matching): a/25 — [one line]
Problem 2 (Workflow design): b/25 — [one line]
Problem 3 (Image/document scenario): c/25 — [one line]
Problem 4 (Catch the fabrication): d/25 — [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 means the coach grades consistently across all approved assistants.
• The fabrication-catching problem (Problem 4) is the highest-value diagnostic: students who spot all three fabrications and explain what the transcript actually says have genuinely internalized the week's verification habit.

Canvas placement block

canvas_object    = Assignment
title            = "Week 7 Assignment — Multimodal Workflows (adaptive)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = adaptive
submission_types = [online_text_entry, online_url]
due_offset_days  = 48
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-7 assignment is the AI-coached, self-scored version in I-assignment-and-rubric-week-07.md. This file shows the same Week-7 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 3 (multimodal AI; voice prompting; record → transcribe → analyze; modality matching; image creation vs. analysis) · SLO A (choose the right tool and modality) · SLO B (evaluate multimodal AI outputs critically)
Worth 100 points · Assignments group = 15% of the grade

The Assignment

This week is about using AI beyond the text box — voice, audio, images, and documents. In four parts, you'll match modalities to tasks, design a real workflow, analyze an image/document scenario, and catch a fabrication in an AI-generated meeting summary. Submit your answers as a document upload or text entry in Canvas. Read the rubric before you start.

Part 1 — Match modalities to tasks (25 pts). For each task below, name the best AI modality or tool type, and explain in one sentence why it fits.
(a) You have a 45-minute recorded staff meeting that you need summarized with action items.
(b) You found a handwritten recipe in your grandmother's cookbook and want it in a digital text document.
(c) You want to create a custom illustration of a dragon reading a book in a library, but you can't draw.
(d) You need to extract the key arguments from a 40-page academic PDF without reading the whole thing.

Part 2 — Design a multimodal workflow (25 pts). A freelance consultant meets weekly with clients by video call. After each call, she needs a written summary of decisions and a numbered list of action items, ready within 30 minutes. Design a step-by-step multimodal AI workflow for her. Name each step, the type of tool used at each step, and what she should check or verify at each step.

Part 3 — Analyze an image/document scenario (25 pts). A student uploads a photo of an event flyer to a multimodal chatbot and asks for the date and location. The AI gives a confident answer; the student adds it to their calendar and shows up to an empty parking lot on the wrong day. (a) Name at least two things that could have caused the AI's answer to be wrong. (b) What should the student have done before adding this to their calendar? (c) What general lesson does this illustrate about AI image analysis?

Part 4 — Catch the summary fabrication (25 pts). Below is a short fictional meeting transcript, followed by an AI-generated summary. Identify at least two things the AI's summary added, changed, or invented that are not supported by the transcript.

Transcript:

Alex: We need to decide on the venue for the March launch event.
Sam: I checked three options. The Lakeview Hall is too expensive — they quoted $8,000. The downtown co-working space is $2,500 but only seats 60.
Alex: What about the university conference room?
Sam: That's free to use but needs booking 6 weeks in advance. We're at 8 weeks out, so we should be fine.
Alex: Let's go with the university room. Can you book it this week?
Sam: Yes, I'll handle it.

AI summary: "The team met to plan the March launch event. After comparing three venues, they chose the university conference room because it is free and conveniently located near campus amenities. Alex confirmed that all three venue prices were within budget. Sam will book the conference room this week and also follow up with Lakeview Hall to negotiate a better rate."

Show your work: for each fabrication you identify, quote or paraphrase what the transcript actually says (or doesn't say) — a claim that just "sounds wrong" without evidence doesn't earn credit.

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-07.md.)

Rubric — 100 points

Levels describe observable differences. Part totals: 25 + 25 + 25 + 25 = 100.

Instructor answer key — REMOVE BEFORE PUBLISHING TO STUDENTS

Part 1 — correct tool types:
(a) Two-step: audio transcription tool (recording → text), then multimodal chatbot (text → summary + action items). The AI cannot hear the raw audio; it needs the transcript.
(b) Multimodal chatbot with image upload — image-to-text / handwriting-to-text.
(c) Text-to-image generation tool (e.g., DALL·E via openai.com/dall-e-3; Midjourney at midjourney.com; Adobe Firefly at firefly.adobe.com).
(d) Multimodal chatbot with document/PDF upload — ask for key arguments; spot-check against the actual document.

Part 2 — model workflow:
Step 1 — Record the call (phone, laptop, or meeting platform recorder).
Step 2 — Transcribe with a free tool (Whisper-class or built-in phone app). Verify: review transcript for mis-heard names, dollar amounts, and decisions before proceeding.
Step 3 — Analyze: paste transcript into chatbot, prompt for "summary of decisions + numbered action items with owners and deadlines." Verify: compare each action item and decision against the transcript — check for invented items, merged points, or wrong names/figures.

Part 3 — causes + lesson:
(a) Any two plausible causes: AI mis-read a digit (common with dates/numbers); low image resolution or angle caused guessing; AI partially read the text and filled in the rest; the venue name was similar to another and was confused.
(b) Verify against a second source: the event organizer's website, social media post, or a direct contact.
(c) AI image analysis is pattern recognition on pixel data — not perfect human vision. It can be confident and wrong, especially on precise details (dates, phone numbers, prices, addresses). Never use AI-extracted details as the only source of truth for time-sensitive actions.

Part 4 — three fabrications in the AI summary (any two earn full marks):
1. "conveniently located near campus amenities" — the transcript says nothing about amenities or convenience; invented.
2. "Alex confirmed that all three venue prices were within budget" — the transcript shows Lakeview Hall was rejected as too expensive; "within budget" is a fabrication that contradicts the transcript.
3. "Sam will also follow up with Lakeview Hall to negotiate a better rate" — this action item was never discussed; completely invented.

Product-accuracy gate: PASS. Tools named factually (no invented features; tool types described accurately; links to official sites in the course materials); transcript used for Problem 4 is fictional and clearly labeled; no student-produced output is asserted as "the" answer — the key grades the process.

Canvas placement block

canvas_object    = Assignment
title            = "Week 7 Assignment — Multimodal Workflows (traditional)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = traditional
submission_types = [online_upload, online_text_entry]
due_offset_days  = 48
published        = true
rubric_ref       = "week-07-assignment-rubric"
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"