Ampere has created a glossary of terms associated with AI for teaching and learning. Teachers, students and parents can share the same language and understanding about key terms as they navigate the world of AI in education. We have developed the glossary of AI terms and concepts with easy to understand definitions through a grade-band approach to ensure accessibility.
The Kindergarten to Grade 2 glossary provides students with a simple foundation for exploring digital tools, recognizing patterns, and building an early understanding of artificial intelligence.
The Grade 3–5 glossary provides students with a foundational reference for exploring, questioning, and using artificial intelligence and digital literacy concepts in their everyday learning.
The Grade 6-8 glossary provides students with a foundational reference for navigating, analyzing, and applying artificial intelligence and digital literacy concepts across their coursework.
The Grade 9 glossary provides students with a foundational reference for navigating, analyzing, and applying artificial intelligence and digital literacy concepts across their coursework.
Kindergarten – Grade 2
This glossary provides Kindergarten to Grade 2 students with a simple foundation for exploring digital tools, recognizing patterns, and building an early understanding of artificial intelligence.
At this band, students are noticing, recognizing, and beginning to to name AI-related concepts in their world. Language should be simple and direct.
Artificial Intelligence (AI)
AI is when a computer seems smart. It can find a face in a photo. It can pick a song for you. But it does not think or feel. People made it and told it what to do.
Think of it like…a fast helper that can do tricks, but it does not know why it does them.
See also: AI tool, Machine, Generative AI
AI chatbot
An AI chatbot is a program you can talk to. You type or speak, and it gives words back. It sounds like a person, but it is not. It does not know what its words mean.
Think of it like…a bird that can copy words. It sounds real, but it does not know what it said.
See also: AI tool, Generative AI, Bot
AI-generated content
This is something a computer made, not a person. It could be text, a picture, or a sound. It can look real. But it might not be true.
Think of it like…a picture drawn by a machine. It looks nice, but no human person chose what it should mean.
See also: Generative AI, AI writing tool
AI tool
An AI tool is an app or program that uses AI to help you. A voice helper on a phone is an AI tool. So is spell check. People made each tool to do one job.
Think of it like…a tool in a toolbox. You pick the right one, like a wrench, for the right job.
See also: Artificial Intelligence, Digital tool
AI writing tool
An AI writing tool is a program that can write for you. You tell it what to write about. It puts words in a row based on what other people wrote. The ideas are not its own.
See also: Generative AI, AI-generated content
Algorithmic bias
This is when a computer is not fair to some people. It happens when the computer did not learn about all kinds of people. It is not mean on purpose. It just did not learn enough.
Think of it like…a cook who only knows one kind of food. That cook may not make meals that work for everybody e.g. like someone who has allergies.
See also: Data bias, Data
Automation
Automation is when a machine does a job on its own. No person has to do it. A dishwasher cleans on its own. That is automation. But if a computer does your work, you miss the chance to learn.
Think of it like…riding an escalator instead of using the stairs. You get to the top, but your legs do not get stronger.
See also: AI tool, Responsible AI use
Bot
A bot is a program that does a task on its own. No one is running it while it works. Some bots help. Some do not, like ones that send junk notes or pretend to be real people.
Think of it like…a wind-up toy. No one is steering it, it just runs.
See also: AI chatbot, AI tool
Data bias
Data bias is when the facts a computer learned from are not fair. Some people may be left out. Then the computer may not work well for all. It needs facts about all kinds of people.
Think of it like…reading only books about one place and then trying to describe the whole world. A lot would be left out.
See also: Algorithmic bias, Data
Data privacy
Data privacy means keeping facts about you safe. When you use apps, they may save facts about you. You should know what is being saved. Make sure your facts stay safe.
Think of it like…closing your curtains. Just because someone could look into your house does not mean they should.
See also: Responsible AI use
Debugging
Debugging means finding what went wrong in a program and fixing it. You try an idea. If it does not work, you try a new one. You keep going until it is fixed.
Think of it like…figuring out why a recipe went wrong by checking each step. Did you skip a step? Or use the wrong ingredient at some point?
Digital tool
A digital tool helps you do things with a screen. A tablet is one. So is a search engine.
Some digital tools use AI and some do not.
Think of it like…tools in a toolbox. Each one does a different job.
See also: AI tool
Generative AI (GenAI)
Generative AI is a kind of AI that makes new things. It can make words, pictures, or music. It looks at things people made first. Then it mixes them up. What it makes looks new. But the parts came from other people.
Think of it like…a collage made by a machine. It puts parts together, but every part came from something someone else made.
See also: AI-generated content, AI writing tool, AI chatbot
Machine
A machine is something built by people to do a job. A computer is a kind of machine. Machines do not think or feel. They follow steps that people gave them.
Think of it like…a bicycle. It helps you go faster, but it does not pick where to go. You do.
See also: AI tool, Digital tool
Prediction
A prediction is a guess about what comes next. AI tools make guesses all the time. They guess what word you will type. They guess what show you might like. These guesses come from patterns, not from knowing.
Think of it like…guessing what a friend will pick for lunch based on what they ate last week. You can spot a pattern but you cannot read their mind.
See also: Artificial Intelligence, Data
Responsible AI use
This means thinking before you use AI. Be honest when AI helps you. Keep your facts safe. Check if what AI tells you is true.
Think of it like…borrowing a friend’s toy. Use it with care. Give it back. Be honest about what happened.
See also: Data privacy, AI tool. Automation
Grades 3-5
This glossary provides Grade 3–5 students with a foundational reference for exploring, questioning, and using artificial intelligence and digital literacy concepts in their everyday learning.
At this band, students are consolidating awareness and beginning to actively apply AI literacy. They evaluate tools, encounter AI-generated content, and begin understanding how AI learns from data. Definitions should add technical depth while remaining concrete and example-driven.
Artificial Intelligence (AI)
A computer or device which is built to do tasks that usually need human thinking, like suggesting what to watch next or like recognizing faces.
Think of it like…a very fast helper that can follow complicated instructions, but cannot understand why it’s doing what it’s doing i.e. the way a calculator can add numbers but doesn’t know what numbers mean
See also: AI tool, Machine, Generative AI
AI chatbot
A computer program that can have a text or voice conversation with you. It is not a person and does not understand what you are saying—when creating its response, it is just predicting what words are most likely to come next.
Think of it like…a parrot that has heard millions of conversations and can repeat phrases that sound right for the moment, but it does not know what the words mean.
See also: AI tool, Generative AI, Bot
AI-generated content
Text, pictures, audio, video etc. which was made by a computer program rather than created directly by a person. AI-generated content can look and sound very real, even when the information in it is wrong e.g. the image shows something that never happened.
Think of it like…a drawing made by a machine that studied millions of other drawings. The result can look like a real drawing but no person imagined it or chose what it should mean.
See also: Generative AI, AI writing tool
AI tool
Any program, app, or device which uses artificial intelligence to help people do something. Voice assistants, spelling suggestions, photo filters are all AI tools. Every AI tool was designed by people who made choices about what the tool does and how it works.
Think of it like…a special kind of tool in a toolbox. It can do a specific job but a person has to choose the right one and decide when to use it.
See also: Artificial Intelligence, Digital tool
AI writing tool
A program that can write sentences, paragraphs, or stories based on instructions you give it. It produces text by predicting which words are most likely to come next, based on patterns from the large amount of text it was trained on. The writing may sound natural but it is not the tool’s own ideas, it is just remixing patterns from things other people have written.
See also: Generative AI, AI-generated content
Algorithmic bias
When a computer program treats some people unfairly because of what it learned from. If an AI tool is built using examples that don’t include all kinds of people, or that reflect unfair patterns from the past, the tool may make decisions that leave some people out or treat them differently.
Think of it like…learning to cook using only one family’s recipes. You’d make great food for that family, but you might not know how to make meals that work for everyone.
See also: Data bias, Data
Automation
When a machine or computer does a task that a person used to do, without needing a person to do it each time. Some automation is helpful, like a dishwasher cleaning dishes. But when a computer does your learning for you, you miss the chance to grow and get better at it yourself.
Think of it like…riding an escalator instead of climbing the stairs. The escalator gets you there, but your legs don’t get any stronger.
See also: AI tool, Responsible AI use
Bot
A computer program that does tasks automatically, without a person controlling it in real time. Some bots are helpful, like a chatbot that answers questions on a website. Other bots can be harmful, like programs that send spam messages or pretend to be real people online.
Think of it like…a wind-up toy that keeps doing the same things on its own once someone starts it.
See also: AI chatbot, AI tool
Data bias
When the information used to teach an AI tool is unfair, incomplete, or doesn’t represent everyone equally. If an AI learns mostly from examples about only one group of people, it may not work well – or may work unfairly – for other groups. Data bias is one of the main reasons AI tools can produce biased results.
Think of it like…reading only books about one country and then trying to describe the whole world (your description would leave a lot of people out!)
See also: Algorithmic bias, Data
Data privacy
Keeping personal information safe and controlling who can see it, collect it, and use it. When you use AI tools and apps, information about you – what you type, where you are, what you click on – may be collected or stored. Protecting data privacy means understanding what information is being collected and making sure it stays safe.
Think of it like…closing the curtains in your room. Just because someone could look in doesn’t mean they should, and you get to decide who you let in.
See also: Responsible AI use
Debugging
Finding and fixing problems in a computer program or system. When something isn’t working the way it should, you can test different ideas to figure out what went wrong and how to fix it.
Think of it like…figuring out why a recipe didn’t turn out right by checking each step (did you forget an ingredient? Was the oven too hot?)
Digital tool
Any device or program that uses a computer to help you do something, like a tablet, a calculator app, a drawing program. Some digital tools use AI and some don’t.
Think of it like…different tools in a toolbox. A hammer and screwdriver are both tools, but they do different things, so choosing the wrong one can make a job harder.
See also: AI tool
Generative AI (GenAI)
A type of AI that produces new content – like text, images, music, video – based on patterns it has learned from large amounts of existing content. The results look and sound new, but they’re not truly original: every output is built by remixing patterns from things other people have created. AI chatbots and AI writing tools are examples of generative AI.
Think of it like…a collage made by a machine which has seen millions of other pictures. It can arrange the pieces into something that looks new, but every piece came from somewhere else.
See also: AI-generated content, AI writing tool, AI chatbot
Machine
A device built by people to do a specific job. Computers are a type of machine. AI tools run on computers and other machines, but the machines themselves do not think, feel, or understand—they just follow instructions created by people.
Think of it like…a bicycle. It helps you go faster, but it doesn’t decide where to go (you do!)
See also: AI tool, Digital tool
Prediction
A guess about what will happen or what something is, based on patterns. AI tools make predictions all the time: predicting what word you’ll type next, what video you might want to watch. AI predictions are based on patterns in data, not on understanding—the AI doesn’t know if its prediction is right or wrong.
Think of it like…guessing what your friend will choose for lunch based on what they’ve chosen every other day this week. You might be right, but you’re just noticing a pattern, you’re not reading their mind.
See also: Artificial Intelligence, Data
Responsible AI use
This involves thinking carefully about when, why, and how you use AI tools, and taking responsibility for the results of your use. It means being honest about when AI helped you, protecting your own and other people’s personal information, checking whether the AI’s output is accurate, and making sure the tool is being used fairly.
Think of it like…borrowing a friend’s bike. You’re responsible for using it carefully, returning it in good shape, and telling the truth about what happened while you had it.
See also: Data privacy, AI tool. Automation
Grades 6-8
This glossary provides Grade 6-8 students with a foundational reference for navigating, analyzing, and applying artificial intelligence and digital literacy concepts across their coursework.
At this band, students are active AI tool users engaging with evaluation, application, and the beginning of creation. They analyze real-world cases, evaluate systemic impacts, and design responses. Definitions are technically precise and engage with the social, ethical, and structural dimensions of AI.
Artificial Intelligence (AI)
The design of computer systems that can perform tasks that normally require human intelligence, like recognizing images, understanding speech, or translating languages.
AI systems are built by people using different approaches: some follow pre-written rules, and others use machine learning to find patterns in large amounts of data. AI systems do not think or understand—they process information in ways that can appear intelligent but are fundamentally different from how humans think.
See also: Machine learning, Algorithm, AI tool
AI chatbot
A computer program designed to have conversations with people using text or speech. Modern AI chatbots are powered by large language models (LLMs) and generate their responses by predicting the most likely next word, over and over, based on patterns in the text they were trained on. They can produce helpful, well-written responses, but they do not understand what they are saying and may generate information that is confident-sounding but wrong.
Think of it like…a very fast autocomplete that writes entire paragraphs instead of single words. It’s guessing what comes next, not thinking about what’s true.
See also: Large Language Model, Generative AI, Hallucination
AI-generated content
Text, images, audio, video, or code produced by an AI system rather than created directly by a person. Because AI generates content by remixing patterns from its training data, the output can look convincing but may contain errors, biases, or fabricated information. Any content produced by generative AI should be checked for accuracy before it is used or shared.
See also: Generative AI, Hallucination, AI writing tool
AI tool
Any software application or device feature that uses artificial intelligence to perform a task. AI tools include search engines, voice assistants, spelling and grammar checkers, recommendation systems, image filters, and AI chatbots.
Each AI tool was designed by people who made choices about what data to use, what tasks to prioritize, and what the tool can and cannot do, and those choices affect how well the tool works and who it works well for.
See also: Artificial Intelligence, Digital tool, Algorithm
AI transparency or disclosure
Being open and clear about how an AI system works, what data it uses, what its limitations are, and when AI is involved in producing content or making a decision.
Transparency applies to the companies that build AI tools (being honest about what their tools do with user data) and to individuals who use AI (being honest about when and how AI contributed to their work).
Think of it like…showing your work on a math test. The answer matters, but so does letting people see how you got there.
See also: Responsible AI use, Ethical AI, ML explainability
AI writing tool
A generative AI program that produces written text – sentences, paragraphs, essays, stories, or code – based on a user’s instructions. It works by predicting the most likely sequence of words based on patterns in its training data. The output may be grammatically correct and well-organized but is not the tool’s own thinking, it is a recombination of patterns from text written by other people.
See also: Generative AI, AI-generated content, Large Language Model
Algorithm
A set of step-by-step instructions that a computer follows to complete a task or solve a problem. Every AI tool runs on algorithms: rules that tell the computer what to do with the data it receives. Some algorithms are simple (like sorting a list alphabetically), while others are enormously complex (like the ones that power search engines or recommendation systems.)
Think of it like…a recipe. The recipe tells you what ingredients to use, what order to do things in, and what the result should be. An algorithm does the same for a computer.
See also: Artificial Intelligence, Machine learning
Algorithmic bias
Happens when an AI system produces results that are systematically unfair to certain groups of people. This usually occurs because the training data the system learned from was incomplete, unrepresentative, or reflected existing social inequalities.
Algorithmic bias is rarely intentional—it is a consequence of the data and design choices that went into building the system. Detecting and correcting algorithmic bias requires deliberately checking whether the system works equally well for all the people it affects.
See also: Data bias, Training data, Ethical AI
Automation
The use of technology to perform tasks without ongoing human involvement. AI-powered automation can be valuable e.g. sorting emails, flagging spam, or assisting with repetitive calculations. But when automation replaces the thinking, practicing, or creating that helps people learn and grow, the person misses the chance to develop skills and understanding, even if the task gets done.
See also: Responsible AI use, Machine learning
Bot
A software program designed to perform tasks automatically, often on the internet. Some bots are useful e.g. search engine bots that index websites, chatbots that answer customer questions. Others are harmful e.g. bots that spread spam, manipulate social media, or impersonate real people.
Knowing that bots exist helps you evaluate whether the content you encounter online was created by a person or generated automatically.
See also: AI chatbot, AI-generated content
Data
Information in a form that computers can store, process, and analyze, including text, numbers, images, audio, and video. AI systems depend on data: they are trained on it, they collect it from users, and they produce new data as output. The quality, completeness, and fairness of the data used to build an AI system directly affects how well and how fairly that system works.
See also: Training data, Data bias, Data privacy
Data bias
Occurs when the data used to train an AI system does not accurately or fairly represent the real world. If training data overrepresents some groups and underrepresents others, or if it reflects historical patterns of discrimination, the AI system will learn and reproduce those imbalances.
Data bias is one of the most common sources of unfair outcomes in AI and is often difficult to detect because the data may appear neutral on the surface.
Think of it like…making a survey of “favourite foods” but only asking kids in one neighbourhood. Your results would tell you about that neighbourhood, not about everyone, but you might mistakenly treat them as if they did.
See also: Algorithmic bias, Training data
Data privacy
The right to control what personal information is collected about you, how it is stored, who can access it, and what it is used for. AI tools and apps routinely collect data from users, including conversation history, location, browsing habits, and device information.
Understanding data privacy means recognizing what data is being collected, evaluating whether that collection is fair and necessary, and taking steps to protect your information.
See also: Informed consent, Responsible AI use
Debugging
The process of finding and fixing errors in a computer program, system, or process. Debugging requires systematic thinking: identifying what went wrong, testing possible causes, and verifying that a fix works. It is a core problem-solving skill in both computing and everyday life.
Ethical AI
The practice of designing, building, and using AI systems in ways that are fair, transparent, accountable, and respectful of human rights and dignity. Ethical AI is not a feature of a specific tool—it is a set of principles applied by the people who create, deploy, and use AI systems.
Questions of ethical AI include: Is this tool fair for all users? Is it transparent about how it works? Who is accountable when it causes harm?
See also: Responsible AI use, AI transparency, Algorithmic bias
Generative AI (GenAI)
A category of AI that produces new content – text, images, audio, video, or code – by identifying and recombining patterns from the large datasets it was trained on. The output may look and sound original, but it is not truly new in the way human creativity is: every element traces back to patterns in existing data. AI chatbots, AI image generators, and AI writing tools are all forms of generative AI.
See also: AI-generated content, Large Language Model, Training data
Hallucination
When a system generates information that sounds confident and plausible but is factually wrong or completely made up. This happens because generative AI predicts the most likely next word or pattern, not the most true one. The AI has no way to check whether what it produces matches reality.
Hallucination is not a rare bug; it is a structural feature of how current AI models work.
Think of it like…someone telling a story and filling in details they don’t actually remember because the details sound right. They’re not lying, they genuinely don’t know the difference between what they remember and what they invented.
See also: AI-generated content, Prediction, Generative AI
Informed consent
Agreeing to something after you genuinely understand what you are agreeing to. In the context of AI and data privacy, informed consent means understanding what data a tool will collect, how it will be used, who will have access to it, and what the consequences might be, before you agree to use the tool.
Many apps and AI tools have terms of service that are long, complex, and difficult to understand, which means users often agree without being truly informed.
See also: Data privacy, Personal information, Ethical AI
Large Language Model (LLM)
A type of AI that has been trained on enormous amounts of text (books, websites, articles, conversations) to predict and generate human-like language.
LLMs power AI chatbots and many AI writing tools. They produce responses by predicting the most likely next word based on patterns in their training data; they do not understand meaning, hold beliefs, or access live information. The word “large” refers to the billions of numerical values (called parameters) the model uses to make its predictions.
Think of it like…someone who has read millions of books and can write sentences that sound like they belong in any of them, but they’ve never actually experienced anything those books describe.
See also: Generative AI, AI chatbot, Training data, Hallucination
Machine Learning (ML)
A way of building AI systems by showing them large amounts of data and letting them find patterns, instead of programming them with step-by-step rules for every situation. The system “learns” by adjusting its internal settings to make better predictions over time.
Machine learning is what makes it possible for AI to recognize faces, recommend songs, filter spam, or translate languages—tasks too complex to solve with written rules alone.
Think of it like…learning to recognize dogs by seeing thousands of photos of dogs, rather than by reading a written description. After enough examples, you can spot a dog you’ve never seen before, not because someone told you the rules, but because you’ve absorbed the pattern.
See also: Artificial Intelligence, Training data, ML model, Algorithm
ML explainability
Refers to how well people can understand why a machine learning model made a particular prediction or decision. Some models are relatively transparent, so you can trace their reasoning. Others, especially large and complex ones, operate as “black boxes” where the internal process is difficult or impossible for humans to interpret.
Explainability matters because when an AI system makes a decision that affects people, those people deserve to understand how and why that decision was made.
Think of it like…a math teacher asking you to “show your work.” If you can only give the answer but can’t explain how you got there, it’s harder for anyone to trust that the answer is right, or to fix it if it’s wrong.
See also: AI transparency, ML model, Ethical AI
ML model
The result of training: it is a mathematical structure that has learned patterns from data and can use those patterns to make predictions about new information.
When people say “the AI” – e.g. “the AI recommended this song” – they are usually referring to an ML model making a prediction based on the patterns it learned during training.
Different ML models are trained for different tasks: recognizing images, translating text, generating music, or detecting spam.
See also: Machine Learning, Training data, Prediction
Prediction
A prediction in AI is the system’s output: its best guess about what something is, what will happen, or what a user wants, based on patterns it has learned from data.
Every time an AI recommends a video, finishes your sentence, or identifies a face in a photo, it is making a prediction. AI predictions can be useful but are not guaranteed to be correct, as they reflect patterns in training data, not truth.
See also: Machine Learning, ML model, Hallucination
Responsible AI use
This involves choosing AI tools thoughtfully, using them with awareness of their limitations, being honest about when and how AI contributed to your work, protecting your own and others’ privacy, verifying the accuracy of AI outputs, and considering whether the tool treats all people fairly.
Responsible use is not just about following rules, it is about developing the habit of asking: is this the right tool, for this purpose, at this moment?
See also: Ethical AI, AI transparency, Data privacy
Supervised learning
A way of training a machine learning model using data that has been labelled by people. Each example in the training data includes both the input (e.g. a photo) and the correct answer (e.g. “this is a cat”). The model learns by comparing its predictions to the correct answers and adjusting until it gets better at predicting labels for new, unseen examples.
Think of it like…a teacher showing you flashcards with the answers on the back. Each time you guess, you check the answer and learn from your mistakes. Over time, you get better at guessing correctly, even for cards you haven’t seen before.
See also: Machine Learning, Training data, Unsupervised learning
Training data
The collection of examples – text, images, numbers, audio, or video – that a machine learning model studies during training to learn patterns. Training data determines what the model knows and how it behaves: if the data is biased, incomplete, or inaccurate, the model’s outputs will reflect those problems. This is often summarized as “garbage in, garbage out.”
Think of it like…the textbook a student studies from. If the textbook has errors or only covers one point of view, the student will learn those errors and limitations—not because the student is flawed, but because the source material was.
See also: Data, Data bias, Machine Learning, ML model
Unsupervised learning
A way of training a machine learning model using data that has not been labelled by people. Instead of learning from correct answers, the model looks for hidden patterns, groupings, or structures in the data on its own.
Unsupervised learning is useful for discovering patterns that humans might not have noticed or looked for.
Think of it like…sorting a jar of mixed buttons without anyone telling you the categories. You might group them by colour, size, shape, or number of holes—the patterns you find depend on what you notice, not on instructions someone gave you.
See also: Machine Learning, Supervised learning, Training data
Grade 9
This glossary provides Grade 9 students with a foundational reference for navigating, analyzing, and applying artificial intelligence and digital literacy concepts across their coursework.
At this band, students are designing, analyzing, creating, and leading. They are active and regular AI users. Definitions are technically precise and include structural, systemic, and governance dimensions.
Artificial Intelligence (AI)
Computer systems designed to perform tasks that typically require human cognitive abilities e.g. decision-making, content-generation, pattern recognition, language processing. AI systems can perform specific tasks at superhuman speed and scale, but they operate without understanding, consciousness, or intent. The distinction between what AI can do (process data, generate text.) and what it cannot do (understand meaning, feel consequences) is fundamental to using it responsibly.
See also: Machine Learning, Algorithm, Generative AI, Responsible AI use
AI chatbot
Conversational interface powered by a large language model which generates text responses to user input. It works by predicting the statistically most likely sequence of words given the conversation context and its training data. AI chatbots can produce fluent, well-structured, and persuasive text on virtually any topic, including topics they have no reliable information about.
See also: Large Language Model, Hallucination, Prompt
AI-generated content
Any output – text, images, audio, video, code – produced by a generative AI system. AI-generated content can be difficult or impossible to distinguish from human-created content by appearance alone. Evaluating AI-generated content requires the same verification skills applied to any unverified source.
See also: Hallucination, Deepfake / Synthetic media
AI tool
Any application or system feature that uses artificial intelligence to perform or assist with a task. AI tools range from everyday features (autocomplete, spam filters) to specialized systems (medical imaging analysis, code generation.)
Every AI tool reflects the design choices of its creators—what data it was trained on, what it was optimized to do, what it was not designed to handle, and whose needs it prioritizes.
See also: Artificial Intelligence, Algorithm, Algorithmic bias
AI transparency
The principle that AI systems and their use should be open to scrutiny: how a system works, what data it uses, what decisions it makes, what its limitations are, and when AI is involved in producing content.
Transparency applies at multiple levels—companies disclosing how their tools work, institutions explaining their AI policies, and individuals being honest about how AI contributed to their work.
See also: ML explainability, Ethical AI, Responsible AI use
Algorithm
Defined sequence of instructions or rules that a computer follows to process data and produce and output. In AI, algorithms govern everything from how a model is trained to how it generates predictions.
See also: Machine Learning, Recommendation algorithm
Algorithmic bias
Systematic unfairness in the outputs of an AI system, resulting from biased training data, flawed design assumptions, or optimization choices that advantage some groups over others.
Algorithmic bias is rarely intentional but can have significant real-world consequences e.g. in hiring, healthcare, education, criminal justice. It is detected through structured analysis and addressed through diverse data, inclusive design. and ongoing auditing.
See also: Data bias, Training data, Ethical AI
Automation
The use of technology to perform tasks with minimal or no human involvement. AI-powered automation is expanding into areas that once required human judgement e.g. writing, analysis, creative work, and decision-making.
The question for responsible AI use is not whether automation is efficient (it usually is) but whether it serves or undermines human development.
See also: Responsible AI use, Machine Learning
Computer Vision
Field of AI focused on enabling computers to process and interpret information from images and video. Applications include facial recognition, medical image analysis, self-driving cars, and photo filters.
Computer vision systems are trained on large datasets of images and learn to detect patterns; however, they can fail when encountering images that differ from their training data (different lighting, skin tones, angles, or contexts.) These failures are often systematic, not random.
See also: Machine Learning, Training data, Algorithmic bias
Conversational AI
The broader category of AI systems designed to engage in dialogue with humans, through text, voice, or both. It includes AI chatbots, voice assistants, and automated customer service agents.
Conversational AI systems are powered by LLMs and natural language processing, enabling them to parse user input, generate relevant responses, and sustain multi-turn conversations.
They do not understand the content of conversations or form genuine intentions.
See also: AI chatbot, Large Language Model (LLM), Prompt
Data
Information in a form that can be stored, processed, and analyzed by computers. In AI contexts, data includes training data (used to build models), user data (generation through interactions with tools), and output data (produced by AI systems.)
Data is not neutral—what is included, excluded, and how it is categorized reflects the choices of the people who collected and organized it.
See also: Training data, Data bias, Data privacy, Data sovereignty
Data bias
The presence of systematic distortions in a dataset that cause an AI system trained on it to produce unfair or inaccurate results.
Sources of data bias include underrepresentation (some groups are missing or undersampled), historical bias (the data reflects past discrimination), and measurement bias (data was collected using methods that systematically favour some outcomes over others.)
See also: Algorithmic bias, Training data, Ethical AI
Data privacy
The right of individuals to control the collection, storage, use, and sharing of their personal information.
In AI contexts, data privacy involves understanding that AI tools collect data through every interaction—not just what you explicitly share, but metadata like location, timing, device information, and usage patterns.
Effective data privacy requires individual awareness alongside legal protections, transparent data practices by companies, and institutional policies which safeguard the people whose data is being used.
See also: Informed consent, Data sovereignty
Data sovereignty
The principle that data is subject to the laws and governance structures of the community or nation where it is collected. In the context of Indigenous data sovereignty, the idea is for communities to have the right to own, control, access, and possess information about their own people.
Data sovereignty recognizes that data is not just a technical resource, it is also an expression of identity, culture, and self-determination.
See also: Data privacy, Data, Ethical AI
Deepfake / Synthetic media
An AI-generated video, audio, or imagery that realistically depicts a person saying or doing something they never actually said or did, Synthetic media is the broader term for any media content that has been substantially created or layered using AI.
The existence of deepfakes means that seeing or hearing something is no longer sufficient evidence that it actually happened.
See also: AI-generated content, Generative AI, AI transparency
Ethical AI
The practice of designing, building, deploying, and using AI systems in accordance with principles of fairness, transparency, accountability, privacy, and respect for human dignity.
It is not a technical feature but a continuous process of evaluation: who is affected by this system, how are they affected, and what safeguards are in place?
See also: Responsible AI use, AI transparency, Algorithmic bias, Data sovereignty
Generative AI (GenAI)
AI systems that produce new content – text, images, audio, video, code – by identifying and recombining patterns learned from massive training datasets. The outputs are not original in the way human creative work is: they are statistical reconstructions of patterns the model has seen.
Generative AI’s rapid adoption raises significant questions about authorship, accuracy, intellectual property, and the relationship between human effort and machine output.
See also: Large Language Model, AI-generated content, Training data, Hallucination
Hallucination
Refers to outputs that are factually incorrect, fabricated, or internally inconsistent. Hallucination occurs because generative AI models predict statistically probable outputs, not truthful ones; the model has no mechanism for checking its predictions against reality.
Research has demonstrated that some level of hallucination is mathematically unavoidable in current large language model architectures. This means that any output from a generative AI system must be independently verified before it is relied upon.
See also: AI-generated content, Large Language Model, Prediction
Informed consent
In AI and data contexts, this means that a person has genuinely understood what data will be collected, how it will be used, who will access it, and what the potential consequences are, before agreeing to the data collection.
See also: Data privacy, Ethical AI
Large Language Model (LLM)
An AI system trained on vast quantities of text data to predict and generate human language. LLMs power AI chatbots, writing assistance, translation tools, and coding aids.
Despite their fluency, LLMs have no understanding of meaning, and no ability to verify whether their outputs are true. The “intelligence” of an LLM is the appearance of understanding produced by extremely sophisticated pattern-matching.
See also: Generative AI, AI chatbot, Hallucination, Training data
Machine Learning (ML)
Method of building AI systems in which algorithms learn patterns from data rather than following manually written rules. During training, an ML system adjusts its internal parameters to improve its ability to make accurate predictions on new data.
Machine learning enables AI applications that would be impossible to build with explicit rules, like recognizing speech, translation languages, or identifying objects in images.
See also: Artificial Intelligence, Training data, ML model, Supervised learning, Unsupervised learning
ML explainability
The degree to which the internal workings and decision-making processes of a machine learning model can be understood by humans. Some models, like decision trees, are relatively transparent; others, like neural networks, are black boxes whose reasoning is extremely difficult to interpret.
Explainability is critical in high-stakes applications – e.g. healthcare, criminal justice, hiring – where people affected by AI decisions have a right to understand how those decisions were made.
See also: AI transparency, ML model, Ethical AI
ML model
A machine learning model is a mathematical structure that has been trained on data to recognize patterns and make predictions. The model is the product of the training process: it encodes the patterns the system has learned and applies them to new inputs.
Different models are trained for different tasks – e.g. image recognition vs. fraud detection – and their behaviour is shaped by the training data, the algorithm used, and the design choices made by developers.
See also: Machine Learning, Training data, Prediction
Prediction
An AI system’s estimate of the most likely answer, category, or next step based on patterns learned during training. An AI prediction is the system’s output for a given input.
Predictions can range from useful (recommending a song you’ll probably enjoy) to consequential (flagging a medical scan for further review) to harmful (systematically scoring one group of applicants lower than another.)
Understanding that all AI outputs are predictions – not facts – is foundational to using AI responsibly.
See also: ML model, Machine Learning, Hallucination
Prompt
The input – text, question, or instruction – that a user gives to a generative AI system. The quality and specificity of the prompt significantly affects the quality of the output.
Understanding how prompts work also means understanding that the AI is responding to the order of words in the prompt, not to the user’s intention i.e. it cannot infer what you meant if it isn’t explicitly stated.
See also: Generative AI, AI chatbot, Large Language Model
Recommendation algorithm
An AI system that predicts what content, products, or information a user is most likely to engage with, based on their past behaviour and the behaviour of similar users. Recommendation algorithms power the feeds on social media, music streaming services, video platforms, and eCommerce sites.
They are designed to maximize engagement, which does not always align with what is accurate, healthy, or in the user’s best interest.
See also: Algorithm, Data prediction, AI tool
Responsible AI use
The practice of selecting, using, and evaluating AI tools with deliberate attention to their accuracy, fairness, privacy implications, and impact on human learning and development. It involves verifying AI outputs, disclosing AI use, protecting the privacy of yourself and others, evaluating whether a tool is appropriate for a given purpose, and considering who might be harmed or excluded by the tool’s design.
See also: Ethical AI, AI transparency, Data privacy, Algorithmic bias
Supervised learning
A machine learning approach in which a model is trained on labelled data (data where each example is paired with its correct answer.) The model learns by comparing its prediction to the known labels and adjusting to reduce errors.
Supervised learning is widely used for classification tasks (e.g. spam or not spam, cat or dog) and prediction tasks (e.g. forecasting temperature, estimating house prices.) Its effectiveness depends on having a large, accurately labelled, representative dataset.
See also: Machine Learning, Training data, Unsupervised learning
Training data
The dataset used to teach a machine learning model to recognize patterns and make predictions. The content of the training data – what’s included, what’s excluded, how it’s labelled, where it came from – fundamentally shapes the model’s behaviour, capabilities, and limitations.
Training data that reflects historical biases will produce a model that reproduces those biases. Training data that lack diversity will produce a model that fails when encountering the diversity of the real world.
See also: Data, Data bias, Machine Learning, ML model
Unsupervised learning
A machine learning approach in which a model is trained on unlabelled data and must discover patterns, groupings, or structures on its own. Unlike supervised learning, there are no correct answers provided during training—the model identifies clusters, associations, or anomalies in the data.
Unsupervised learning is useful for exploratory tasks where the categories aren’t known in advance, such as identifying customer segments or detecting unusual network activity.
See also: Machine Learning, Supervised learning, Training data