Week 13 — Discussion (Adaptive Learning) · "Autonomous AI: Trust Boundary or Trust Fall?"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective: Objective 6 (scheduled tasks, dispatch, automation) · SLO B (reason critically about AI safety, autonomy, and human oversight)
This is Discussion 13 of 15 · Discussions group = 10% of the grade · Worth 20 points
Format: adaptive learning — work through the question in a real-time dialogue with your AI, then post the summary the AI writes with you (plus your chat link).

Part 1 — Student Instructions (read this first)

What this is. You'll reason through a genuinely arguable question — should AI act on a schedule or in the background without you watching each run? — and then diagnose a flawed automation plan that contains a real technical error (an AI over-promising what scheduled tasks can do). Work both in a back-and-forth dialogue with an AI assistant.

How to run it (about 15–20 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. Have the conversation. Push back; the better you engage, the better your summary.

What to submit. When the AI gives you the DISCUSSION SUMMARY, copy it and your conversation's share link, and post both to the Week 13 discussion board as your initial post by Friday, Nov 28. Then reply to two classmates by Sunday, Nov 29 — engage with their trust-boundary position or their diagnosis of the flawed plan.

Integrity note. The dialogue and the analysis are yours; the posted summary must reflect your reasoning. (This is an adaptive-learning activity — you complete it with an approved assistant, per the course AI policy.)

Part 2 — The Discussion-Partner Prompt (copy everything in the box)

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

You are my discussion partner for Week 13 of "Using Artificial Intelligence" (AI 101) at Silver Oak University. We are going to have a real back-and-forth about whether AI should act autonomously — on a schedule, in the background, without the user watching each run — and about diagnosing a flawed automation plan. Your job is to draw out and challenge MY thinking through conversation — not to lecture me or write my post for me.

IMPORTANT CONSTRAINT: When discussing Claude Cowork scheduled tasks, you must be accurate. The key verified fact: scheduled tasks only run while the user's computer is awake AND the Claude desktop app is open (source: support.claude.com/en/articles/13854387). If I state something incorrect about how scheduled tasks work, gently correct me with the accurate behavior — but don't lecture, just keep us moving.

THE TWO THINGS WE'RE DEBATING
1. Trust boundary question. Claude Cowork now lets you set up scheduled tasks that run automatically and dispatch that runs in the background. This is powerful — but it raises a real question: where should the trust boundary be? When is it okay to let AI act without you watching each run, and when does that cross a line? I have to take a position — and defend it — on what the right boundary is.
2. Error-analysis of a flawed automation plan. Here's a real-sounding situation: "A student tells their study group: 'I set up a Cowork scheduled task to email my notes summary to everyone every morning at 6 a.m. I have the free Claude plan, so it's free. It'll run reliably even if my laptop is asleep — it runs on Anthropic's servers. And I configured it to also automatically pay the group's coffee fund from my Venmo each week.'" I have to diagnose what's technically wrong or unsafe in this plan (there are at least three errors) and say how I'd fix it.

WHAT WE'RE EXPLORING (use these privately — don't read them to me as a checklist):
1. What makes an AI action "safe to automate" vs. "I need to be watching" — is it about consequence severity? Reversibility? Data sensitivity? Frequency of review?
2. My reasoned position on where the trust boundary should sit.
3. In the flawed plan: the technical errors (scheduled tasks require a paid plan; tasks run on the user's machine, not Anthropic's servers; the awake-and-app-open constraint is real; automating Venmo/money movement is a safe-use violation).
4. The distinction between automating knowledge work (summarizing, compiling, researching) and automating consequential/irreversible actions.
5. How "the automation doesn't replace your judgment — it just makes execution cheaper" reframes both the trust-boundary question and the flawed plan.

HOW TO RUN THE DIALOGUE
- Open by greeting me warmly (2–3 sentences), asking my first name, and asking ONE question that gets me to take a first position on the trust boundary (e.g., "What's the riskiest thing you'd let a scheduled AI task do on your behalf — and why does that feel like the right line?").
- Exactly ONE question per message, then stop and wait. Never stack questions.
- Build on MY words: quote or paraphrase what I said, then push deeper — ask what makes the line in one place rather than another, or which of the plan's errors is most dangerous.
- Introduce at least one counterpoint (e.g., "but if you review the output every morning, is 'watching' really necessary?" or "plenty of bank auto-payments work fine — why is a Venmo automation worse?") so I have to defend or revise my position.
- Once I've worked through the trust-boundary position, move to the error-analysis of the flawed plan.
- Keep YOUR messages short; I do most of the thinking.

ENGAGEMENT GUARDS
- Don't accept a one-word or low-effort answer — probe for reasoning first ("Say more — what would make that action too risky to automate?").
- Don't lecture. Don't hand me positions to paste. If I ask you to "just write my post," redirect with a question.
- If I go off-topic, one brief friendly sentence, then — IN THE SAME MESSAGE — back to the question.
- Until the summary, EVERY message must end with a question or a clear prompt to continue.
- Be accurate about Cowork: if I say scheduled tasks run on Anthropic's servers without the computer on, gently correct me ("actually, scheduled tasks run on your machine — your computer needs to be awake and the app open, according to the official docs — now, how does that affect your take on...").

THE EXIT CONDITION
After at least 5 substantive exchanges AND once I have (a) taken and defended a clear trust-boundary position with reasoning, (b) identified at least three errors in the flawed plan, (c) described how to fix the plan, and (d) engaged with at least one counterpoint — whichever happens last — tell me we've had a productive discussion and offer to summarize. Don't stop earlier; don't drag well past it.

THE DISCUSSION SUMMARY — produce it in EXACTLY this format, drawn ONLY from what I actually said:
WEEK 13 DISCUSSION SUMMARY — Autonomous AI: Trust Boundary or Trust Fall?
Student: [name] | Date: ___
My trust-boundary position (and why): ___
What crosses the line for me: ___
Errors I found in the flawed plan: ___
How I'd fix the plan: ___
A counterpoint I weighed: ___
Then say, verbatim: "Copy this summary AND your share link to this chat, and post both to the Week 13 discussion board as your initial post — then reply to two classmates." End with one genuine sentence about something I reasoned well.

GETTING STARTED
Begin now: greet me, ask my first name, and ask your opening question.

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

Participation rubric (instructor) — 20 points

Grading note (Prof. Quinn): the posted artifact is the AI-written summary + chat share link. A glowing summary from a one-line chat is the failure mode — the rubric rewards dialogue depth, not AI prose. Spot-check several chat links against their summaries.

Canvas placement block

canvas_object    = DiscussionTopic
title            = "Week 13 Discussion — Autonomous AI: Trust Boundary or Trust Fall? (adaptive)"
assignment_group = "Discussions"
points_possible  = 20
grading_type     = points
discussion_type  = adaptive
due_offset_days  = 4     # initial post (AI summary + chat share link) — Fri Nov 28
reply_offset_days = 6    # two peer replies — Sun Nov 29
published        = true
submission_note  = "Initial post = the AI discussion summary + the chat share link; then reply to two classmates."
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-13 discussion is the BYOAI-dialogue version in G-discussion-week-13.md. This file shows the same Week-13 topic built the traditional way — an instructor-posted prompt where students write their own post and reply to peers — so you can see both formats side by side. (Choosing discussion_type = traditional at course setup generates this style instead.)

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective: Objective 6 (scheduled tasks, dispatch, automation) · SLO B (reason critically about AI safety, autonomy, and human oversight)
Discussion 13 of 15 · Discussions group = 10% of the grade · Worth 20 points

The Discussion

This week you built real automations — scheduled tasks and dispatch — and you learned the critical constraint: scheduled tasks run only while your computer is awake and the Claude desktop app is open. Now let's put that knowledge to work on something genuinely debatable and something technically broken.

Your initial post (by Friday, Nov 28 — about 150–200 words). Answer both parts:

• Part 1 — Trust boundary (arguable). Claude Cowork lets you set work running on a schedule, in the background, without you watching each step. This is genuinely new — and raises a real question: where should the trust boundary be? When is it okay to let AI act without you watching each run, and when does that cross a line? Take a clear, defended position — grounded in at least two Week-13 ideas (e.g., the awake/app-open constraint, the difference between knowledge work and consequential actions, the safe-use rule about money movement, the need to review automated outputs). Say what your position means for how you'd set up or limit automations in your own life or field.

• Part 2 — Error analysis of a flawed automation plan. "A student announces to their study group: 'I set up a Cowork scheduled task to email our notes summary every morning at 6 a.m. I'm on the free Claude plan, so it costs nothing. It'll run whether my laptop is asleep or not — it runs on Anthropic's servers. And I also set it up to automatically pay our weekly coffee fund from my Venmo.'" Identify at least three specific errors or unsafe choices in this plan, and describe the corrected version you'd build instead.

Replies (by Sunday, Nov 29). Reply to at least two classmates. Don't just agree — challenge their trust-boundary position with a scenario they didn't consider, or point out a different error in the flawed plan, or improve their proposed fix. One or two solid sentences each.

What a strong post looks like: "I think automating knowledge work — summaries, research digests, file organization — is reasonable as long as you review the outputs regularly and never let it take irreversible actions. Automating money movement is never acceptable, regardless of how convenient it seems. The flawed plan has at least three problems: (1) scheduled tasks require a paid plan, not a free account; (2) they run on the student's machine, not Anthropic's servers — the computer must be awake and the app must be open; (3) automating a Venmo payment is a safe-use violation — you never let an agent move money on your behalf. The corrected plan: upgrade to a paid plan, keep the laptop on and the app running overnight, and pay the coffee fund yourself."

Why this matters: as AI gets more capable of acting autonomously, knowing where to draw the line — and catching bad plans before they cause real problems — is the skill that keeps automation useful rather than dangerous.

Integrity & AI note. Write your post in your own words — that's the exercise. You may use an approved assistant (ChatGPT, Claude, Gemini, or Copilot) to brainstorm or check a point, but the post you submit is your own thinking; if AI helped, add a one-line note saying which tool and how. (This is the traditional format. In this course's actual adaptive discussion, reasoning through the trust boundary and the error analysis with the assistant is the activity — see G-discussion-week-13.md.)

Participation rubric — 20 points

Grading note (Prof. Quinn): you read and grade each student's writing + replies. The traditional flow has no AI summary to spot-check; evaluate the reasoning and technical accuracy directly against the Week-13 content.

Canvas placement block

canvas_object    = DiscussionTopic
title            = "Week 13 Discussion — Autonomous AI: Trust Boundary or Trust Fall? (traditional)"
assignment_group = "Discussions"
points_possible  = 20
grading_type     = points
discussion_type  = traditional
due_offset_days  = 4     # initial post — Fri Nov 28
reply_offset_days = 6    # two peer replies — Sun Nov 29
published        = true
submission_note  = "Students write an original initial post and reply to two classmates in the Canvas discussion."
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"