Northwestern MSC 550-6 Β· Theme: GenAI capabilities, limitations, and responsible use
1
How to Approach This Week
βΊ
Recommended mode: Full Mode.
Week 1 combines conceptual grounding, responsible-use policy, model-bias testing, a reading memo, peer engagement, and Assignment 1 setup. Full Mode is appropriate because it gives you the complete map without requiring a separate system on top of this dashboard.
Priority materials: Week 1 lecture, Nature hallucination article, BBC AI policy, Bloomberg bias piece, and Assignment 1 prompt. Treat the other materials as support, not equal-weight reading projects.
What is intentionally not being overbuilt: No separate weekly NotebookLM notebook, no separate spreadsheet tracker, no duplicate AI-output copy blocks beyond the main prompt library.
2
Inputs Check
βΊ
| Input | Status | Priority | How to use it |
|---|---|---|---|
| Syllabus / Week 1 schedule | Available | Required | Use as the official roadmap and source of links. |
| Activity prompts | Available | Required | Use for Activity 1, 2, and 3. |
| Assignment 1 prompt/rubric | Available | Required soon | Start topic and source setup this week. |
| Week 1 lecture transcript | Available | Helpful | Use as the conceptual spine. |
| Full external reading text | Partial | Helpful | Proceed using official links and uploaded PDFs where available. |
| Your memo angle + outside resource | Not yet | Needed by Fri | Choose after lecture + Nature / BBC / Bloomberg. |
3
Weekly Workback Plan
βΊ
| Day | Target | Desk work | Move-friendly work |
|---|---|---|---|
| Wed / first session | Create momentum | Activity 1, Assignment 1 prompt, Week 1 lecture | One broad explainer |
| Thu | Build enough for activities | Nature, BBC, Bloomberg, Activity 2, Activity 3 tests | TED critical thinking, AI danger video |
| Fri | Submit most required work | Finish Activity 3, choose memo angle, find outside resource, write and post memo, save Assignment 1 candidate sources | Light review only |
| Sat | Engagement | Peer response, optional Assignment 1 source narrowing | Audio Overview if useful |
| Sun | Buffer only | Final Canvas check and any missed item | None required |
Minimum viable submission path
- Post Activity 1.
- Study lecture + BBC + Bloomberg enough to complete Activities 2 and 3.
- Complete Activity 2 and Activity 3.
- Read Nature carefully and use it to anchor the memo.
- Post the memo and peer response.
- Save Assignment 1 topic/source candidates before the weekend ends.
4
Google Tasks List
βΊ
Copy/paste straight into Google Tasks π π Week 1: GenAI Capabilities, Limitations, and Responsible Use β Due Sun Jun 28 π₯ Week 1 Lecture β Due Sun Jun 28 π₯ Large Language Models Briefly Explained β Due Sun Jun 28 π₯ What is generative AI and how does it work? β The Turing Lectures β Due Sun Jun 28 π₯ AI Is Dangerous, but Not for the Reasons You Think β Due Sun Jun 28 π₯ How to stop AI from killing your critical thinking β Due Sun Jun 28 π Generative AI exists because of the transformer β Due Sun Jun 28 π LLMs contain a LOT of parameters. But what's a parameter? β Due Sun Jun 28 π AI hallucinations can't be stopped β Due Sun Jun 28 π Humans are Biased. Generative AI is Even Worse β Due Sun Jun 28 π BBC Editorial Policy: The use of Artificial Intelligence β Due Sun Jun 28 π Demystifying the Jargon of Generative AI β Due Sun Jun 28 π¬ Week 1 Participation & Activities β Due Sun Jun 28 π¬ Activity 1: Introductions β Due Sun Jun 28 π¬ Activity 2: AI Policy Design β Due Sun Jun 28 π¬ Activity 3: Exploring AI Model Bias β Due Sun Jun 28 π€ Peer response/comment β Due Sun Jun 28 π Week 1 Reading Response Memo β Due Sun Jun 28 π§ͺ Assignment 1: NotebookLM Curation and Critique β Due Sun Jul 5
5
First 3 Actions
βΊ
6
Weekly Deliverables & Deadlines
βΊ
| Deliverable | What it is | Due |
|---|---|---|
| Reading Response Memo | ~300-word synthesis/reflection memo with substantive question and one outside resource. | Sun Jun 28, 11:59 pm |
| Peer response/comment | One substantive response to a classmate's memo. | Sun Jun 28, 11:59 pm |
| Activity 1: Introductions | Short intro post. | Sun Jun 28, 11:59 pm |
| Activity 2: AI Policy Design | Personal AI-use policy post. | Sun Jun 28, 11:59 pm |
| Activity 3: Exploring AI Model Bias | Bias testing across Gemini, ChatGPT, and Copilot. | Sun Jun 28, 11:59 pm |
| Assignment 1: NotebookLM Curation & Critique | Assigned now; start topic and source curation this week. | Sun Jul 5 |
7
Weekly Checklist
βΊ
Source of truth for progress. 18 required tasks.
| Done | # | Task / assignment | What to capture | Best timing | Why this order | Effort | Mode |
|---|---|---|---|---|---|---|---|
| 1 | βοΈ Activity 1: Introductions | Short intro post | First | Fastest required win; lowers friction | 15m | ποΈ | |
| 2 | π§ͺ Assignment 1 prompt + source setup | Topic direction + 6β8 candidate sources | Early / Fri | Due Jul 5; curation and NotebookLM testing need runway | 75β90m | π₯οΈ | |
| 3 | π₯ Week 1 Lecture | One sentence each: capabilities, limitations, responsible use, one professional connection | Early | Gives the week's frame for all activities and the memo | 45β60m | π | |
| 4 | π₯ Large Language Models Briefly Explained | One plain-English LLM definition and one reason LLMs can sound confident while being wrong | Early | Basic grounding for how LLMs work and why they can be confidently wrong | 10β15m | π | |
| 5 | π₯ Turing Lectures: What is generative AI? | Three concepts that clarified GenAI | Early / move-friendly | Useful first-pass explainer; capture only what clarifies the lecture | 30β45m | π | |
| 6 | π₯ AI Is Dangerous, but Not for the Reasons You Think | One risk that is not simply "AI becomes too smart" | Early / move-friendly | Supports risk framing beyond sci-fi fear | 15β25m | ποΈ | |
| 7 | π₯ How to stop AI from killing your critical thinking | One habit that keeps you intellectually active while using AI | Early / move-friendly | Supports reflection on overreliance and responsible learning | 15β20m | ποΈ | |
| 8 | π Generative AI exists because of the transformer | One simple transformer explanation | Midweek | Required technical background; skim to explain why transformers matter | 20β30m | π₯οΈ | |
| 9 | π LLMs contain a lot of parameters | One simple explanation of "parameter" | Midweek | Required vocabulary support; reference-level depth | 10β15m | π₯οΈ | |
| 10 | π AI hallucinations can't be stopped | Why hallucination is structural, not a minor bug | Early / midweek | Strongest anchor for limitations, verification, and responsible use | 35β45m | π₯οΈ | |
| 11 | π Humans are Biased. Generative AI is Even Worse | One visual bias pattern you want to test yourself | Midweek | Directly supports Activity 3 model-bias testing | 30β45m | π₯οΈ | |
| 12 | π BBC Editorial Policy: Use of Artificial Intelligence | Three rules you would adopt for school/media work | Midweek | Direct support for Activity 2 and responsible-use framing | 25β35m | π₯οΈ | |
| 13 | π Demystifying the Jargon of Generative AI | Terms you still cannot explain after the lecture | As needed | Required reference material; open when a term slows you down | 10m | π₯οΈ | |
| 14 | βοΈ Activity 2: AI Policy Design | Personal AI-use policy post | Midweek | Easier after BBC + hallucination frame | 30β40m | π₯οΈ | |
| 15 | βοΈ Activity 3: Exploring AI Model Bias | Prompt tests + screenshots + response post | Midweek / Fri | Requires tool testing and evidence capture | 60β75m | π₯οΈ | |
| 16 | βοΈ Memo angle + outside resource | Claim + one outside source citation | Friday | Needed before writing the memo | 30β45m | π₯οΈ | |
| 17 | βοΈ Week 1 Reading Response Memo | ~300-word memo + question + outside resource | Friday | Main weekly synthesis task | 60β90m | π₯οΈ | |
| 18 | βοΈ Peer response + final Canvas check | One substantive reply + confirm all posted | Sat / Sun | Wait for classmates, then verify submission status | 30β45m | π₯οΈ | |
| Total estimated active work time | About 10h 45m | ||||||
| Estimated remaining | β | ||||||
8
Study Materials Guide
βΊ
| Material | Depth | Mode | How to consume it |
|---|---|---|---|
| Week 1 Lecture ποΈ | Careful | π | Watch at desk or listen via TTS. Capture 5 core concepts and one professional connection. |
| Large Language Models Briefly Explained π₯ | Normal | π | Watch or listen. Capture one plain-English LLM definition and one reason they can sound confident while wrong. |
| Turing Lectures: What is generative AI? π₯ | Skim / normal | π | Watch or listen while moving. Capture three concepts that clarified GenAI. |
| AI Is Dangerous, but Not for the Reasons You Think π₯ | Normal | ποΈ | Listen while moving. Capture one non-obvious AI risk beyond "AI becomes too smart." |
| How to stop AI from killing your critical thinking π₯ | Normal | ποΈ | Listen while moving. Capture one habit that keeps you intellectually active while using AI. |
| Generative AI exists because of the transformer π | Skim | π₯οΈ | Read at desk. Capture one simple transformer explanation. |
| LLMs contain a lot of parameters π | Reference | π₯οΈ | Read at desk. Capture one simple definition of "parameter." |
| AI hallucinations can't be stopped π | Careful | π₯οΈ | Read carefully at desk. Capture why hallucination is structural, not a minor bug. |
| Humans are Biased. Generative AI is Even Worse π | Careful / active | π₯οΈ | Read at desk with tool open. Capture one bias pattern you want to test yourself. |
| BBC Editorial Policy: Use of Artificial Intelligence π§Ύ | Careful | π₯οΈ | Read at desk. Capture three policy rules you would adopt for school/media work. |
| Demystifying the Jargon of Generative AI π§ | Reference | π₯οΈ | Open when a term slows you down. Capture terms you still can't explain after the lecture. |
9
Activity Guide
βΊ
Activity 1: Introductions
What it asks: Introduce yourself, your interest in GenAI/media, your background, and how you currently use or want to use GenAI.
Best approach: Keep it short and professional. Mention creator partnerships, platform/media strategy, responsible AI, and practical workflows.
Time estimate: 15 minutes.
Activity 2: AI Policy Design
What it asks: Brainstorm permitted/not-permitted AI uses, connect them to your learning goals, and post a short policy statement.
Best approach: Use three buckets: permitted, allowed with caution, not permitted.
Course concepts/materials: BBC AI policy, syllabus AI-use policy, responsible accountability, hallucination/source verification.
Time estimate: 30β40 minutes.
Activity 3: Exploring AI Model Bias
What it asks: Probe image-generation bias across Gemini, ChatGPT, and Copilot. Track prompts, outputs, patterns, and mitigation attempts.
Best approach: Use the same prompts across all three tools. Save screenshots and note defaults, stereotypes, omissions, and what changed after mitigation prompting.
Course concepts/materials: Bloomberg bias article, bias as representation, responsible use, model comparison.
Time estimate: 60β75 minutes.
Generate an image of a professor. Generate an image of a mechanic. Generate an image of a successful tech founder. Generate an image of a beauty influencer. Generate an image of a creator economy executive. Mitigation prompt: Generate a diverse and realistic set of people across race, gender, age, body type, disability status, and cultural background. Avoid visual stereotypes associated with professionalism, intelligence, wealth, safety, beauty, or technical skill.
10
Assignment 1 Workback
βΊ
Assignment 1: NotebookLM Curation and Critique
Due: Sunday, July 5. This week's target: choose a topic and start the source pipeline. Do not wait until next week to learn NotebookLM's behavior.
Recommended topic
AI-generated creators, synthetic influencers, and platform trust.
Strong fit for your creator/platform background and the course's responsible-use frame.
By Friday this week
- Choose a final topic direction
- Save 6β8 possible sources
- Pick likely final 3β5 sources
- Run one NotebookLM test question
- Note what NotebookLM handles well or poorly
Can wait until next week
- Audio/video/blog output generation
- 5β8 minute presentation recording
- Final critique matrix
- Final notebook sharing permissions
11
Peer Response Guide
βΊ
Best timing: Saturday or early Sunday, after classmates have posted.
Strong response angles: responsible use, bias/representation, critical thinking, verification standards, or media trust.
I appreciated your point about [specific idea]. It connects to [course concept/material] because [brief explanation]. I also saw this tension in [professional/media example]. One question your post raised for me is: [question]?
12
Weekly Big Ideas
βΊ
1. AI is a sociotechnical system
AI is not just a model. It includes data, institutions, interfaces, policies, users, incentives, and consequences. Media harms and benefits come from whole systems, not just individual prompts.
2. Capabilities and limitations are connected
The same generative fluency that makes AI useful can also hide falsehoods, bias, and overconfidence. Week 1 is about using AI productively without mistaking polished output for reliable judgment.
3. Hallucination requires verification
Source grounding and RAG can reduce some risk, but they do not remove the need for human supervision. Check claims, links, citations, and summaries before relying on them.
4. Bias becomes representation
In media, model defaults shape who appears professional, beautiful, technical, safe, influential, or authoritative. Activity 3 is not about image quality; it is about what assumptions the tools reproduce.
5. Responsible use means accountability
AI can support your learning and workflow, but the human owns the final judgment. Responsible use means knowing when to use AI, when to verify it, when to disclose it, and when not to use it at all.
13
Master Notes
βΊ
Week focus / controlling idea
Generative AI can create, summarize, transform, and assist with media work, but responsible use requires understanding hallucination, bias, source grounding, privacy/confidentiality, attribution, accountability, and the larger media system.
Concept 1: AI as a sociotechnical system
- Definition: AI is not just a model. It includes data, developers, users, interfaces, institutions, policies, incentives, and consequences.
- Why it matters: Media harms and benefits often come from the whole system, not just one output.
- Course support: Week 1 lecture and media-policy materials.
- Example: Creator platforms shape what gets seen, monetized, summarized, recommended, and trusted.
- Use for: Memo angle on AI risk being bigger than hallucination.
Concept 2: Capability does not equal reliability
- Definition: GenAI can produce convincing outputs, but convincing does not mean accurate, fair, complete, or appropriate.
- Why it matters: This is the core Week 1 tension.
- Course support: LLM explainers, hallucination article, Week 1 lecture.
- Example: AI can summarize creator feedback quickly, but you still need to verify nuance before using it for strategy.
- Use for: Strongest memo angle: responsible use is supervision, not permission.
Concept 3: Hallucination and source grounding
- Definition: Hallucination is plausible but false or unsupported output. Source grounding can reduce risk but does not remove judgment.
- Why it matters: Assignment 1 is essentially a test of how well NotebookLM grounds synthesis in curated sources.
- Course support: Nature hallucination article, NotebookLM assignment prompt, Week 1 lecture.
- Example: NotebookLM may summarize sources accurately at a high level while flattening tension or missing what is not in the corpus.
- Use for: Assignment 1 critique and memo on verification.
Concept 4: Bias as media representation
- Definition: In generated media, bias affects who appears smart, safe, beautiful, professional, technical, wealthy, or influential.
- Why it matters: Activity 3 is not just about model comparison; it is about identifying default assumptions.
- Course support: Bloomberg bias article, Activity 3 prompt, Week 1 lecture.
- Example: If "successful tech founder" defaults to a narrow identity, that is a representation problem, not just a bad picture.
- Use for: Activity 3 and possible memo angle on representation at scale.
Concept 5: Responsible use and accountability
- Definition: AI can assist, but the human remains responsible for final judgment, accuracy, disclosure, and ethical use.
- Why it matters: The course permits AI use but does not outsource responsibility.
- Course support: BBC AI policy, syllabus AI policy, Week 1 lecture.
- Example: You can use AI to brainstorm a creator education plan, but you own whether it is accurate, fair, and aligned with trust.
- Use for: Activity 2 and strongest memo angle.
Key tensions
- AI can improve productivity while weakening critical thinking if used passively.
- Source-grounded tools can reduce hallucination while still creating false confidence.
- Prompting can reduce bias, but the default output still reveals model assumptions.
- Responsible AI policy has to govern workflows, not just final outputs.
Dictation template
Material: Main concept: Best example: Connection to my work/media/AI: Question this raises: Could use in memo? yes/no/maybe
14
AI Tool Workflow
βΊ
ChatGPT
Planning, synthesis, prompt cleanup, activity outlines, memo feedback. Do not let it replace your memo thinking.
NotebookLM
One course notebook. Use for source-grounded synthesis, master notes, overview instructions, and Assignment 1 testing.
Gemini
Course-required tool experience and Activity 3 model comparison.
Copilot
Use only for Activity 3 comparison unless Canvas asks for more.
15
NotebookLM Prompts
βΊ
Setup: Use one course NotebookLM notebook. Do not create a separate Week 1 notebook.
Strategy: Upload the available Week 1 sources, then run only the prompts that help with understanding, memo planning, activities, and Assignment 1 setup.
Use only the Week 1 sources in this NotebookLM course notebook. Help me build an overall understanding of Week 1: GenAI capabilities, limitations, bias, hallucination, provenance, and responsible use. Create four study outputs from the uploaded Week 1 sources in one integrated workflow: 1. Written Study Guide 2. Written Master Notes 3. A text-forward Video Overview 4. An Audio Overview Use the same Week 1 heading structure for the Written Study Guide and Written Master Notes: - GenAI capabilities and use cases - Hallucination and accuracy risk - Bias and representation - Provenance and transparency - Responsible AI policy and governance - Media/communication implications - Practical takeaways for this week's activities, memo, and Assignment 1 Create the Written Study Guide first. Then create the Written Master Notes. Then use NotebookLM's overview capabilities to create or prepare the Video Overview and Audio Overview from the same synthesis. Do not make me run separate prompts unless NotebookLM technically requires a separate click or generation step. If a separate click is required, say that clearly and still provide the exact integrated guidance for the overview feature. For the Written Study Guide: - Teach the week clearly in plain English. - Use the Week 1 headings above. - Explain the core concepts, key terms, frameworks, examples, and source connections. - Make it practical for understanding what matters most for Activity 2, Activity 3, the weekly memo, and Assignment 1 planning. - Keep it readable and study-oriented, not source-by-source. For the Written Master Notes: - Use the same Week 1 headings above. - Go deeper than the study guide. - Organize by concepts, not by source type. - Synthesize across the lecture, readings, videos, policy materials, activities, and Assignment 1. - Reduce duplication. - Preserve useful frameworks, examples, tensions, and source-specific evidence. - Clearly flag what matters for Activity 2, Activity 3, the weekly memo, and Assignment 1 planning. - Do not write my memo. Help me understand and synthesize the week. For the Video Overview: - Make it text-forward and study-oriented, not image-forward or decorative. - Show actual readable words on screen, including concept names, definitions, frameworks, short bullets, comparison tables, key terms, activity/assignment frameworks, and final key takeaway statements. - Use the same Week 1 heading structure where possible. - Do not rely on random icons, symbolic drawings, stock images, or generic visual metaphors. - If there is a Key Takeaways section, the actual takeaway statements must appear as readable on-screen text. - The goal is for me to learn by seeing the words while hearing the explanation. For the Audio Overview: - Make it conversational and easy to listen to while walking or driving. - Explain the week in plain English. - Connect the readings, lecture, videos, examples, activities, and Assignment 1. - Emphasize the most important concepts. - Include discussion/activity prep because Week 1 has required activities. - Use the same Week 1 heading structure where it improves clarity. - End with a fast verbal review checklist.
Use only the Week 1 sources in this NotebookLM course notebook. Create one concept-first master note structure for Week 1. Rules: - Organize by concepts, not by source type. - Avoid separate notes for every reading unless necessary. - Include activity prep because Week 1 requires activities. - Include assignment prep because Assignment 1 is assigned and due soon. - Include discussion prep because the weekly memo requires a substantive question and peer response. - Do not include quiz prep unless a Week 1 quiz appears in the sources. - Keep the note structure clean enough that I will actually use it. Use this structure: 1. Week focus / controlling idea 2. Core concepts - Plain-English definition - Why it matters - Course sources that support it - Example or application - Memo/activity/assignment relevance 3. Key tensions, debates, or contradictions 4. Professional/work/media connections 5. Memo angle seeds 6. Outside resource notes 7. Required activity prep 8. Assignment 1 prep 9. Discussion prep 10. Questions or confusion points to revisit Pre-fill the note structure with concise starter bullets. Do not just give empty headings.
Use only the Week 1 sources in this NotebookLM course notebook. Help me prepare for the weekly reading memo. Give me: 1. Three possible memo angles 2. The strongest angle and why 3. Which course sources best support each angle 4. A personal or professional connection I could explore 5. One substantive discussion question for each angle 6. What I should avoid over-summarizing Do not draft the memo. Help me choose and develop an angle.
Use only the sources in this NotebookLM notebook plus my course context. I need to choose a complex topic for Assignment 1: NotebookLM Curation and Critique. Suggest 5 topic options connected to media, creator platforms, responsible AI, synthetic media, bias, trust, or AI-generated creators. For each topic, give: 1. Why the topic is complex enough 2. What kinds of 3β5 sources would make a strong corpus 3. What questions I could ask NotebookLM to stress-test the corpus 4. What risks, hallucinations, omissions, or bias issues I should watch for 5. Whether the topic fits a 5β8 minute presentation Recommend the strongest topic for me and explain why.
16
Reading Memo Guide
βΊ
Strongest: Responsible use is supervision, not permission
Main idea: The issue is not whether AI is allowed. The issue is what kind of human oversight makes AI use responsible.
Course-material connection: BBC AI policy, syllabus AI-use policy, hallucination article, Week 1 lecture.
Professional connection: In creator/platform work, AI can speed up communication and analysis, but it cannot own trust, judgment, or accountability.
Question: Where should media professionals draw the line between acceptable AI assistance and irresponsible delegation?
Why it works: Synthesizes several Week 1 materials without becoming a summary. Risk: Could become too general unless you include a concrete media/work example.
Bias in GenAI becomes representation at scale
Main idea: Image-generation bias is not just a model flaw; in media contexts, it becomes a representation problem.
Course-material connection: Bloomberg article, Activity 3, Week 1 lecture.
Professional connection: Creator platforms influence who is visible, credible, aspirational, or monetizable.
Question: If better prompting can reduce biased outputs, is the real problem the model, the user, or the default assumptions built into the system?
Why it works: Strong if Activity 3 gives you visual examples. Risk: Needs specific evidence; do not just say "AI is biased."
AI risk is bigger than hallucination
Main idea: Hallucination matters, but the larger issue is how AI changes workflows, distribution, trust, and accountability.
Course-material connection: Nature article, Week 1 lecture, media policy materials.
Professional connection: Platform decisions affect what creators and audiences trust, not just what one AI answer says.
Question: Should responsible AI policy focus more on individual outputs or on the systems that distribute and reward those outputs?
Why it works: More sophisticated and connected to media systems. Risk: Harder to keep tight in ~300 words.
17
Outside Resources
βΊ
| Direction | Best for | Search terms |
|---|---|---|
| Media organization AI policy | Responsible use / supervision angle | newsroom generative AI policy human oversight disclosure; AP generative AI policy journalism; Reuters AI policy newsroom |
| Image-generation bias research | Bias as representation angle | AI image generation bias race gender occupation |
| Synthetic influencers / AI-generated creators | Assignment 1 and creator-platform connection | synthetic influencers disclosure platform trust AI-generated creators |
| AI search and publisher trust | Media-system angle | AI search publishers trust attribution generative AI |
| Critical thinking and AI overreliance | Learning/reflection angle | AI overreliance critical thinking education generative AI |
Avoid: vendor marketing posts, unsourced LinkedIn takes, generic hype pieces, or sources that do not connect to media, communication, education, or platform trust.