Skip to main content
Tokenline

Resources

Keep learning beyond the line.

Trusted places to go deeper, a glossary of every term from our lessons, and a checklist to use before you use AI on schoolwork.

Before you use AI on an assignment

Six questions that sum up everything in Lines 2 and 3. If any answer is “no,” pause and rethink.

  1. 1Does my teacher or school allow AI for this assignment? If unsure, ask first.
  2. 2Am I using AI to help me think — not to do the thinking for me?
  3. 3Have I kept personal and private information out of my prompt?
  4. 4Did I check every fact, number and source the AI gave me?
  5. 5Can I explain the final work in my own words without the AI?
  6. 6Did I disclose or cite my AI use the way my class requires?

A–Z

Glossary

Every key term from every lesson, in one searchable list.

Academic integrity
Being honest and responsible in schoolwork: doing your own work, giving credit to sources, and following the rules for each assignment. From: Academic Integrity: Using AI Without Cheating Yourself
Accountability
Being responsible for a decision and its effects — including explaining it and fixing mistakes. It belongs to people and organizations, not to software. From: Human Oversight: Knowing When Not to Use AI
Active recall
A study strategy where you practice pulling information out of memory, such as answering questions, instead of just rereading notes. From: AI for Studying & Brainstorming
Algorithmic bias
Systematic errors in an AI or computer system that create unfair outcomes, such as consistently favoring or disadvantaging certain groups of people. From: Bias & Fairness: Whose Patterns Does AI Learn?
Artificial intelligence (AI)
The broad field of building computer systems that perform tasks usually associated with human intelligence. From: What Is Artificial Intelligence?
Artificial neuron
A small unit that multiplies its inputs by weights, adds them up, and produces an output based on that total. From: Neural Networks, Conceptually
Brittleness
When an AI system fails suddenly on inputs that are only a little different from what it was trained on. From: Strengths & Limits of AI
Checkpoint
A planned moment to review work before moving on, so mistakes are caught early instead of piling up. From: Breaking Down Complex Tasks
Claim
A single statement that can be checked as true or false, such as a date, a number, a quote or a cause-and-effect statement. From: Verifying AI Output
Context window
The maximum amount of text (measured in tokens) a model can consider at once, including the conversation so far and its own reply. From: How AI Assistants Work
Data retention
How long a service keeps the information you give it, and what it is allowed to do with that information during that time. From: Privacy: What Not to Tell a Chatbot
Dataset
An organized collection of data, such as a table of songs with their features and labels. From: Training Data: You Are What You Learn From
Debugging
The process of finding and fixing errors in computer code. From: AI for Research & Coding
Deep learning
A type of machine learning that uses large neural networks with many layers to find complex patterns. From: What Is Artificial Intelligence?
Deepfake
Synthetic media that realistically shows a real person doing or saying something they did not actually do or say. From: Deepfakes, Copyright & Misinformation
Disclosure
Clearly stating how you used an AI tool in your work — for example, "I used a chatbot to generate practice questions and to check my grammar." From: Academic Integrity: Using AI Without Cheating Yourself
Edge case
An unusual or extreme input, like an empty list or a negative number, that can reveal bugs normal testing misses. From: AI for Research & Coding
Fabricated citation
A reference to a book, article, study or web page that looks real but does not exist, or that does not say what the AI claims it says. From: Hallucinations: When AI Sounds Sure and Is Wrong
Feature
A measurable piece of information the model uses as input, like a song's tempo or an email's number of links. From: Machine Learning Basics
Few-shot prompting
Including a few examples of the input and output you want in your prompt so the AI can follow the pattern. From: Prompt Design: Goal, Context, Format, Limits
Garbage in, garbage out
The idea that a model trained on flawed data will produce flawed results. From: Training Data: You Are What You Learn From
Generative AI
AI that creates new content — text, images, audio, code — based on patterns learned from training data. From: What Is Artificial Intelligence?
Hallucination
When an AI model produces information that sounds plausible but is false or made up. From: Strengths & Limits of AI
High-stakes decision
A decision with serious or hard-to-reverse effects on someone's life, such as medical treatment, legal outcomes, school discipline, or hiring. From: Human Oversight: Knowing When Not to Use AI
Human in the loop
A setup where a person reviews, approves or can override an AI system's output before it's acted on. From: Human Oversight: Knowing When Not to Use AI
Inference
Using an already-trained model to make a prediction on new input. From: Machine Learning Basics
Iteration
Improving a result through repeated rounds of feedback and revision, such as refining a prompt based on the last response. From: Prompt Design: Goal, Context, Format, Limits
Knowledge cutoff
The point in time after which a model's training data stops, so it may not know about later events unless it can look them up. From: How AI Assistants Work
Label
The correct answer attached to a training example, such as “spam” or “not spam.” From: Machine Learning Basics
Large language model (LLM)
A very large neural network trained on huge amounts of text to predict likely next tokens, which lets it generate language. From: Generative AI & Large Language Models
Lateral reading
Checking information by opening new tabs to see what other reliable sources say, instead of only reading the original page top to bottom. From: Verifying AI Output
Layer
A group of neurons at the same stage of a network. Outputs from one layer become inputs to the next. From: Neural Networks, Conceptually
Machine learning (ML)
A way of building AI where the computer learns patterns from examples (data) instead of following only hand-written rules. From: What Is Artificial Intelligence?
Misinformation
False or misleading information shared without necessarily intending harm. When it's spread on purpose to deceive, it's often called disinformation. From: Deepfakes, Copyright & Misinformation
Model
The result of training: a set of learned patterns (stored as numbers) that turns an input into a prediction. From: Machine Learning Basics
Narrow AI
AI designed for a specific task or range of tasks. All AI in use today is narrow AI. From: What Is Artificial Intelligence?
Neural network
A machine learning model made of many connected artificial neurons arranged in layers. From: Neural Networks, Conceptually
Next-token prediction
The core task of an LLM: given the text so far, estimate which token is most likely to come next. From: Generative AI & Large Language Models
Personally identifiable information (PII)
Any information that can identify a specific person, such as a full name combined with an address, a phone number, a student ID, or a government ID number. From: Privacy: What Not to Tell a Chatbot
Primary source
An original record or first-hand account, such as a historical document, a dataset or a research study, rather than someone's summary of it. From: AI for Research & Coding
Privacy policy
The document where a company explains what data it collects, how it uses and shares that data, and what choices you have. From: Privacy: What Not to Tell a Chatbot
Prompt
The input text you give a generative AI model to tell it what you want. From: Generative AI & Large Language Models
Proxy variable
A piece of data that isn't about a sensitive trait directly but closely tracks it — for example, a zip code that strongly relates to income or race. From: Bias & Fairness: Whose Patterns Does AI Learn?
Representative data
Data that reflects the full variety of cases a model will actually be used on. From: Training Data: You Are What You Learn From
SIFT
A four-step fact-checking method — Stop, Investigate the source, Find better coverage, Trace claims — developed by digital literacy researcher Mike Caulfield. From: Verifying AI Output
Socratic tutoring
A teaching style that guides you with questions so you reason your way to an answer instead of being told it. From: AI for Studying & Brainstorming
Supervised learning
Machine learning from examples that come with labels, so the model can compare its guesses to the right answers. From: Machine Learning Basics
Synthetic media
Images, audio, video or text created or heavily altered by AI rather than captured directly from the real world. From: Deepfakes, Copyright & Misinformation
System instructions
Hidden directions set by the company or developer that shape how an assistant behaves, such as its tone and what it should refuse to do. From: How AI Assistants Work
Task decomposition
Breaking a large, complex goal into smaller steps that can each be completed and checked on their own. From: Breaking Down Complex Tasks
Token
A chunk of text — a whole word, part of a word, or punctuation — that a language model reads and writes. From: Generative AI & Large Language Models
Training
The process of adjusting a model using example data so its predictions get better. From: Machine Learning Basics
Training cutoff
The point in time after which a model has no training data, so it may not know about more recent events. From: Strengths & Limits of AI
Training data
The collection of examples a machine learning model learns from. From: Training Data: You Are What You Learn From
Weight
A number that controls how much a particular input influences a neuron's output. Learning means adjusting weights. From: Neural Networks, Conceptually