Glossary

Key terms for understanding NotebookLM, Gemini Gems, and AI-assisted writing.

NotebookLM Terms

Notebook: A workspace in NotebookLM containing sources and notes on a specific topic. Each notebook is separate — sources in one notebook aren't available in others. Create a new notebook for each peer response project.
Source: A document, URL, or text you add to a notebook. Sources are the "ground truth" that NotebookLM uses to answer questions. When you ask questions, NotebookLM only draws from sources you've added — it won't make things up from its training data.
Note: Your own writing saved within NotebookLM. Notes can capture your thinking, useful AI responses, or working drafts. Notes can be converted to Sources if you want them searchable within your notebook.
Discover Sources: A NotebookLM feature that searches the web for sources related to your topic. You can review the results and add credible ones to your notebook. The sources it finds are real, but you should still evaluate their quality before adding them.
Audio Overview: A podcast-style audio summary of your sources that NotebookLM can generate. Useful for getting oriented to your research, but don't cite the audio itself — cite the underlying sources.
Citation Numbers: When NotebookLM answers a question, it includes numbered citations (like [1], [2]) that link to specific passages in your sources. Click the numbers to see exactly where the information comes from.

Gemini Gems Terms

Gem: A custom AI assistant in Gemini with standing instructions you define. Unlike regular chats that start fresh, a Gem remembers its instructions across conversations. Create Gems for repeatable workflows like peer response drafting.
Standing Instructions: The rules and process you give a Gem when you create it. These instructions persist — every time you chat with that Gem, it follows the same instructions. You can include tone preferences, step-by-step processes, and output formats.
Gem Manager: The section in Gemini where you create, edit, and organize your Gems. Access it from the left sidebar in the Gemini interface.
Attached Notebook: When you attach a NotebookLM notebook to a Gemini conversation, the AI can access your curated sources while chatting. This combines NotebookLM's source grounding with Gemini's conversational flexibility. (Requires Google One AI Premium.)

AI and Writing Terms

Grounding: When AI answers are tied to specific sources rather than generated from training data. NotebookLM is "grounded" because it only answers from your sources. Regular chatbots are "ungrounded" — they generate from everything they've seen during training, which can include errors.
Hallucination: When AI generates false information that sounds confident and plausible. This includes fake sources, incorrect statistics, and fabricated quotes. Hallucinations happen because AI generates "likely" text, not "true" text.
Fluency: How smooth and readable AI-generated text sounds. High fluency makes text seem trustworthy even when it's inaccurate. "Fluency illusion" is the mistake of assuming fluent text is factually correct.
Fluency Illusion: The tendency to trust information because it's presented confidently and reads well. AI is extremely fluent, which makes it easy to accept its outputs without verification. Also discussed in your course as "grasping a coherent mechanism" that may not reflect reality.
Steelman: To present the strongest possible version of an argument, especially one you disagree with. Steelmanning is the opposite of strawmanning. Good peer response requires steelmanning — representing your peer's argument as charitably as possible before critiquing it.
Strawman: A weak or distorted version of someone's argument that's easier to attack. Bad faith peer response often strawmans opponents. If your summary of a peer's argument is something they'd disagree with, you might be strawmanning.

Course-Specific Terms

Peer Response: A blog post that responds to another writer in the network. In ENGL 170, every post should engage with at least one peer's argument — summarizing it fairly and then extending, challenging, or building on it.
Network: The interconnected set of student blogs in ENGL 170. Posts link to each other, creating a web of arguments. The dashboard shows all posts in the network. Writing for the network means being aware that others may respond to you.
Dashboard: The class website that aggregates all student blog posts. Use it to find peer posts to respond to, track who's posting, and see how conversations develop across the network.
Consumption Paradox: A concept from Gabriel Bell's posts: if AI replaces workers, who buys the products that AI helps produce? The argument that AI abundance might create scarcity by dismantling wage-based purchasing power.
Black Swan: An unpredictable event that defeats statistical prediction. Term from Nassim Taleb. Used by Zay Amaro to argue that prediction markets can't account for completely random, out-of-nowhere moments that change everything.
Adaptive Capacity: The ability to adjust to changing circumstances. Used in the context of labor displacement: some workers have the skills and resources to adapt to AI disruption, while others lack "adaptive capacity."

Technical Terms

LLM (Large Language Model): The type of AI that powers NotebookLM, Gemini, ChatGPT, and Claude. LLMs generate text by predicting what words come next based on patterns in their training data. They're powerful but not reasoning engines — they generate plausible text, not verified truth.
Training Data: The text an LLM learned from during its development. This includes books, websites, articles, and more. When an AI generates ungrounded text, it's drawing on patterns from training data — which may include errors, outdated information, or biases.
Prompt: The text you send to an AI to get a response. Better prompts get better results. In NotebookLM, prompts are questions about your sources. In Gems, standing instructions are a kind of persistent prompt that shapes every response.
Context Window: How much text an AI can consider at once. NotebookLM has a large context window, which lets it analyze multiple sources together. If you add too many sources, older ones might be given less attention.