The vocabulary of AI-native learning. Every entry has a plain-English definition, the Australian classroom context, and the related ideas worth knowing, written for students, parents and schools building genuine capability rather than quiet dependence.
76 terms · Definition · Australian context · Related ideas
A structured, multi-step process in which AI helps plan, execute, check and refine work, with defined points where a human reviews before anything is finalised.
In Australian classrooms. A good agent workflow is a teaching tool: it makes the thinking visible. Edison has students design the checkpoints themselves, so they understand where judgement has to enter the loop.
RelatedAgentic AI · Orchestration · Human-in-the-loop · AI agent
The class of AI systems that plan multi-step work, use tools, observe the result and re-plan. Less about producing text, more about doing the job across several steps.
In Australian classrooms. Agentic systems are powerful and easy to over-trust. For students the lesson is the same as the Edison Method's Evaluate stage: every step an agent takes is a claim to be checked, not an answer to be accepted.
RelatedAI agent · Orchestration · Multi-agent system · Agent workflow
A software system built on a large language model that takes a goal, decides what to do next, uses tools such as search or documents, and reports back. Unlike a chatbot, an agent acts rather than only answers.
In Australian classrooms. Agents are where capable students go beyond chat: a research agent that gathers and checks sources, a study agent that builds and marks practice questions. Edison teaches students to keep a human-in-the-loop so the agent assists thinking rather than replacing it.
RelatedAgentic AI · Tool use · Guardrails · Human-in-the-loop
A higher-order capability than literacy: the conscious ability to think, create, solve and communicate with AI effectively, efficiently and ethically. Knowing AI is the floor; directing it well is fluency.
In Australian classrooms. Anthropic's AI Fluency framework names four habits, Delegation, Description, Discernment and Diligence, that track closely to how Edison teaches students to command AI rather than comply with it. Fluency, not prompting tricks, is the durable skill.
RelatedAI literacy · Prompt engineering · Strategic judgement · Human-AI collaboration
The baseline ability to understand what AI is, how it works, where it fails, and how to use it responsibly. It is broader than operating a chatbot: it is the judgement that sits around the tool.
In Australian classrooms. AI literacy builds directly on the Digital Literacy general capability in the Australian Curriculum (Version 9.0) and maps to UNESCO's AI Competency Framework for Students, which sequences learning across Understand, Apply and Create. At Edison it is the first rung of the Edison Method.
RelatedAI fluency · Digital agency · Source evaluation · Responsible innovation
The ability to operate naturally in a world shaped by AI: to fold it into research, writing, analysis and building without losing your own reasoning or voice.
In Australian classrooms. AI-native capability is the outcome Edison's programs are built toward, from Foundations to Innovators. It is what separates a student who can wield AI as a lever from one who leans on it as a crutch.
RelatedAI fluency · Applied intelligence · Future readiness · Builder mindset
The challenge of making AI systems reliably act in line with human goals and values, including when a goal is stated imprecisely or the system becomes more capable than its designers expected.
In Australian classrooms. Alignment is the serious end of responsible AI, and a genuinely interesting idea for ambitious students. Edison treats it as a way in to ethics and systems thinking, not a doom story.
RelatedThe Paperclip Maximiser · Human-centred AI · Guardrails · Responsible innovation
Knowledge turned into practical output: the move from knowing something to making something with it.
In Australian classrooms. Applied intelligence is the through-line of project-based learning. At Edison a concept is not finished when it is understood; it is finished when a student has built a real artefact that demonstrates it.
RelatedProject-based learning · Prototype thinking · Portfolio learning · Builder mindset
The principle that AI should enhance human capability rather than replace human judgement, a framing that keeps the person in charge of the decision.
In Australian classrooms. Augmented intelligence is the value beneath the Edison Method. The aim is a student who is sharper for having used AI, not one who has quietly outsourced the part that was the learning.
RelatedHuman-AI collaboration · Cognitive offloading · Human-centred AI · AI-native capability
Testing AI systems against standard tasks to compare capability, on coding, reasoning, language understanding or multi-step work. Public benchmarks include MMLU, GPQA and SWE-bench.
In Australian classrooms. Benchmarks matter less for a student than one honest question: does the model actually help me think and build well? Edison teaches students to evaluate a tool against their own real work, not a leaderboard.
RelatedEvaluation (evals) · Output evaluation · Failure modes · Large language model (LLM)
Systematic skew in an AI model's outputs caused by skew in its training data, evaluation or prompting. It ranges from harmless stylistic drift to unfair, harmful outcomes.
In Australian classrooms. The most common bias an Australian student meets is recency and US-centric skew: American spelling, US examples and assumptions unless told otherwise. Spotting it is a Digital Literacy and Ethical Understanding skill in the Australian Curriculum.
RelatedTraining data · Source evaluation · Failure modes · Guardrails
Richard Sutton's 2019 observation that, across seventy years of AI research, general methods that scale with raw computation have consistently outperformed clever, hand-crafted human rules. It is 'bitter' because expertise mattered less than we hoped.
In Australian classrooms. A great idea for senior students debating where progress comes from. Edison uses it to discuss what stays valuable when machines scale, which is judgement, taste and the questions a person chooses to ask.
RelatedScaling laws · Foundation model · Emergence · Frontier model
When an AI system produces a useful answer but humans cannot easily see how it reached it. The model works; its internal reasoning is opaque.
In Australian classrooms. The black box is exactly why Edison insists students can explain and defend their own work without the tool open. If only the machine knows why, the student has not yet learned it.
RelatedExplainability · Interpretability · Hallucination · Verification
Seeing problems as things to design, test and improve, rather than topics to passively absorb. A bias toward making something real.
In Australian classrooms. The builder mindset is the centre of gravity at Edison. Students move from consumer to creator, shipping prototypes and portfolio artefacts they can stand behind.
RelatedPrototype thinking · Founder mindset · Project-based learning · Creative confidence
A structured journey from beginner to advanced, where each stage builds deliberately on the last rather than jumping straight to the tool.
In Australian classrooms. Edison's pathway runs Understand, Use, Evaluate, Build, Lead, age-appropriate and none skipped. It mirrors the way UNESCO's framework sequences competence from Understand to Create.
RelatedAI-native capability · Scaffolding · Future readiness · Project-based learning
Prompting a model to reason step by step before giving a final answer. It improves accuracy on multi-step problems, though modern systems often keep that reasoning hidden.
In Australian classrooms. Useful for any task where the trail of reasoning needs to be visible and checkable, exactly the kind of process-visible work that protects academic integrity. Seeing the steps lets a student catch where the logic breaks.
RelatedPrompt engineering · Verification · Output evaluation · Tool use
John Searle's thought experiment asking whether a machine that produces fluent language genuinely understands it, or merely manipulates symbols by rule without comprehension.
In Australian classrooms. A favourite for classroom debate: it sharpens the difference between sounding right and being right. Edison uses it to ground why verification matters even when an answer reads convincingly.
RelatedStochastic parrot · Hallucination · The Turing Test · Epistemic vigilance
A proprietary AI model whose weights and inner workings are controlled by a company and not published, accessed through an app or API.
In Australian classrooms. Most tools students meet, including the major chat assistants, are closed models. Knowing the difference from open-source AI is part of understanding who controls a tool and where data goes.
RelatedOpen-source AI · Foundation model · Frontier model · Large language model (LLM)
A way of learning in which students observe expert thinking, practise with guidance, and gradually take ownership, the reasoning is made visible, not just the answer.
In Australian classrooms. It is how the Edison Method is taught: a tutor models how to interrogate an AI's output, the student practises with support, then runs the loop alone. Apprenticeship in judgement, not in tools.
RelatedScaffolding · Metacognition · Project-based learning · Human-AI collaboration
Using a tool to reduce mental effort, for drafting, organising, summarising or remembering. Useful in moderation; corrosive when it offloads the thinking that was the point of the task.
In Australian classrooms. Peer-reviewed work links heavy, passive AI use to weaker critical thinking, with the effect strongest in younger users. Edison's answer is the comprehend-first habit: form a view before opening the tool.
RelatedMetacognition · Augmented intelligence · Verification · Epistemic vigilance
Breaking a problem into parts, spotting patterns, defining clear rules, and designing repeatable processes. A foundational thinking skill, with or without a computer.
In Australian classrooms. Computational thinking sits in the Australian Curriculum's Digital Technologies and is the scaffolding beneath good prompting and agent design. At Edison it is taught before tools, not after.
RelatedSystems thinking · Structured thinking · Prompt engineering · Agent workflow
The maximum amount of text, measured in tokens, a model can read at once, including the instructions, the conversation so far, attached documents and the question.
In Australian classrooms. Large context windows let a student feed in a whole reading, a full draft or a unit of work in one pass. Knowing the limit explains why a model sometimes seems to 'forget' earlier parts of a long chat.
RelatedToken · Large language model (LLM) · Memory · RAG (Retrieval-Augmented Generation)
The belief that you can make things, test ideas and solve problems, the willingness to start before you feel fully ready.
In Australian classrooms. Creative confidence is a deliberate outcome at Edison, built through small wins on real projects. AI lowers the friction of making, so students get to the confidence faster, if they stay in command of the work.
RelatedBuilder mindset · Prototype thinking · Machine-assisted creativity · Applied intelligence
A student's ability to use technology intentionally and on their own terms, rather than passively consuming whatever it serves up.
In Australian classrooms. Digital agency is the disposition the Australian Curriculum's Digital Literacy capability is reaching for, and a direct answer to the 2025 NAP-ICT results showing many students using AI heavily but with thin underlying skills.
RelatedAI literacy · Epistemic vigilance · Responsible innovation · Strategic judgement
The ability to use, create and evaluate digital tools and information safely and effectively. In the Australian Curriculum it replaced the older ICT Capability, broadening the focus from how tools work to why and when to use them.
In Australian classrooms. Digital literacy is the foundation AI literacy is built on. ACARA's 2025 national results, the lowest since testing began in 2005, are a reminder that the foundation cannot be assumed just because students are heavy AI users.
RelatedAI literacy · Source evaluation · Bias · Digital agency
A numerical representation of text, image or audio that captures its meaning, letting software search by meaning rather than exact keywords.
In Australian classrooms. Embeddings are the quiet engine behind 'ask your notes' study tools and document Q&A. Understanding them demystifies how an AI can find the relevant passage in a long file.
RelatedRAG (Retrieval-Augmented Generation) · Large language model (LLM) · Token · Context window
When larger AI models display new abilities that were not obvious, or not present, in smaller versions, capabilities that appear to switch on as scale increases.
In Australian classrooms. Emergence is a genuinely open scientific debate, perfect for stretching able students. Edison uses it to discuss why predictions about AI are hard, and why humility about the technology is warranted.
RelatedScaling laws · The Bitter Lesson · Frontier model · Foundation model
Being careful about what you believe, especially when a source sounds confident. The discipline of asking 'how do I know this is true?' before accepting it.
In Australian classrooms. It is the single most protective habit in the age of fluent machines. Edison trains it as a reflex: a confident AI answer is a claim to be checked, not a fact to be copied.
RelatedVerification · Source evaluation · Hallucination · Metacognition
Structured testing to judge whether an AI system performs well across a set of representative tasks, on accuracy, tone, completeness and safety.
In Australian classrooms. Evaluation is the skill most under-taught in AI rollouts and most emphasised at Edison: knowing when to trust, when to verify and when to escalate. It is the Evaluate stage of the Edison Method made concrete.
RelatedOutput evaluation · Benchmarking · Failure modes · Verification
The ability to understand and communicate why an AI system produced a particular output, in terms a person can follow.
In Australian classrooms. Explainability matters most where stakes are high, assessment, health, finance. For students it is also a learning test: if you cannot explain why the answer is right, you have not finished learning it.
RelatedInterpretability · Black box problem · Verification · Responsible innovation
The predictable ways an AI system goes wrong, inventing facts, fabricating citations, defaulting to bias, or sounding most confident when it is least reliable.
In Australian classrooms. Knowing the failure modes is what turns a student from a passive user into a critical one. Edison teaches them by name so they can be anticipated rather than discovered the hard way in an assessment.
RelatedHallucination · Bias · Output evaluation · Epistemic vigilance
Including a few worked examples inside a prompt so the model picks up the format, voice or reasoning pattern you want, without any retraining.
In Australian classrooms. Few-shot is the practical way a student gets consistent, on-brief output: show three strong examples, ask for the next. It teaches by example, which is also how good writing is learned.
RelatedPrompt engineering · Prompt · Fine-tuning · Chain-of-thought
Continuing a pre-trained model's training on a smaller, task-specific dataset so it specialises in a particular voice, format or domain.
In Australian classrooms. Rarely the first move a student needs; clear prompting plus grounding in real sources usually covers it. Edison introduces fine-tuning as an advanced idea, so students know what it is and when it is worth the cost.
RelatedTraining data · Few-shot prompting · Foundation model · RAG (Retrieval-Augmented Generation)
A large, general-purpose AI model trained on broad data that can be adapted to many different tasks. The base layer most AI products are built on.
In Australian classrooms. The chat assistants students use are interfaces over foundation models. Understanding the term clarifies that one underlying model can power dozens of apps, and that the choice of model shapes what a tool can do.
RelatedFrontier model · Large language model (LLM) · Fine-tuning · Closed model
Thinking commercially, creatively and independently: spotting a problem worth solving, scoping it, and taking ownership of getting it built.
In Australian classrooms. Edison's venture-led work cultivates the founder mindset directly, students take an idea from concept to a working prototype they could actually put into the world.
RelatedBuilder mindset · Venture-led learning · Prototype thinking · Applied intelligence
A model at the leading edge of current AI capability. Since 2025 nearly all frontier models use a mixture-of-experts design to scale capacity without a matching rise in cost.
In Australian classrooms. Frontier models move fast, which is why Edison teaches the durable skills, judgement, direction, evaluation, that transfer across whichever model is in front. The interface changes; the thinking does not.
RelatedFoundation model · Mixture of experts · Scaling laws · The Bitter Lesson
Being prepared for a working world reshaped by AI, not by predicting exact job titles, but by building transferable judgement, adaptability and the ability to learn continuously.
In Australian classrooms. Future readiness is the promise parents and schools care about most. Edison frames it honestly: the edge is not access to AI, which everyone will have, but knowing what to do with it.
RelatedAI-native capability · Capability pathway · Strategic judgement · Portfolio learning
AI that produces new content, text, images, audio, code, by predicting likely patterns from what it has learned. The technology behind chat assistants and image tools.
In Australian classrooms. Generative AI is what the Australian Framework for Generative AI in Schools governs. Endorsed by Education Ministers, with its 2024 review re-endorsed in June 2025, it sets the responsible-use principles Edison teaches within.
RelatedLarge language model (LLM) · Hallucination · Foundation model · Responsible innovation
The rules, limits and safety checks placed around an AI system, input filters, output filters, scope limits and human checkpoints, that keep it inside safe boundaries.
In Australian classrooms. Tools such as NSW's NSWEduChat and Queensland's Corella are guardrailed environments built for schools. The same logic applies to a student's own use: decide the limits before you start, not after.
RelatedHuman-in-the-loop · Responsible innovation · Alignment problem · System prompt
When an AI produces confident-sounding output that is factually wrong or invented, fake citations, fabricated details, non-existent facts. A direct consequence of how models predict text.
In Australian classrooms. Hallucination is the main reason unsupervised AI cannot go straight into a student's submitted work. Edison's mitigation is unglamorous and reliable: ground answers in real sources and verify every claim that matters.
RelatedVerification · Failure modes · RAG (Retrieval-Augmented Generation) · Source evaluation
How people and AI work together to produce a better result than either alone, with the human directing, judging and owning the outcome.
In Australian classrooms. Collaboration, not delegation, is the Edison stance. The student stays the author; the AI is a fast, fallible collaborator whose work is always checked and credited.
RelatedAugmented intelligence · Human-in-the-loop · AI fluency · Cognitive apprenticeship
AI designed around people, their needs, ethics and real-world context, rather than around the technology's convenience. Human agency and accountability come first.
In Australian classrooms. A human-centred mindset is the first dimension of UNESCO's AI Competency Framework for Students, and the value Edison puts above tool fluency. The point of the tool is the person it serves.
RelatedAugmented intelligence · Responsible innovation · Alignment problem · Guardrails
A design pattern where AI drafts or proposes and a person reviews and approves before anything is finalised or acted on. The standard safety pattern for consequential work.
In Australian classrooms. Every responsible AI workflow a student builds at Edison has explicit human checkpoints, draft before send, propose before decide. It is also how academic integrity is protected in practice.
RelatedGuardrails · Output evaluation · Agent workflow · Verification
Understanding the inner workings or logic of an AI model, more technical than explainability, which is about communicating the why in human terms.
In Australian classrooms. Interpretability is an active research frontier and a strong destination for students who love the deep end. Edison frames it as the science of opening the black box.
RelatedExplainability · Black box problem · Large language model (LLM) · Alignment problem
A neural network trained on a vast amount of text to predict the next token from the previous ones. Modern LLMs can reason, summarise, draft, code and follow multi-step instructions.
In Australian classrooms. The LLM is the engine under every chat tool a student uses. Knowing it predicts plausible text, rather than looking up true facts, is the single most clarifying idea in the whole glossary.
RelatedToken · Generative AI · Hallucination · Foundation model
Using AI to extend imagination and production, to explore more options, draft faster and prototype ideas, while the creative direction stays human.
In Australian classrooms. Edison teaches students to use AI to widen the search for ideas, then to apply their own taste and judgement to choose and refine. The tool expands the canvas; the student still paints.
RelatedCreative confidence · Prototype thinking · Augmented intelligence · Builder mindset
An AI system retaining useful context across a conversation or workflow, so it can refer back to earlier information rather than starting fresh each time.
In Australian classrooms. Memory makes tools feel more capable but raises a real question students should ask: what is being remembered, and where is it stored? Data-handling judgement is part of responsible use.
RelatedContext window · Token · Guardrails · Responsible innovation
Thinking about your own thinking, planning before you start, monitoring as you go, and checking the result against what you intended.
In Australian classrooms. Metacognition is among the highest-impact, lowest-cost teaching strategies in the evidence base. Responsible AI use is metacognition in modern clothes, and it runs right through the Edison Method.
RelatedCognitive offloading · Epistemic vigilance · Verification · Cognitive apprenticeship
A model design where many specialised sub-networks exist but only a few are activated for any given input, giving huge total capacity while keeping the cost of each response low.
In Australian classrooms. Mixture of experts is now the standard architecture for frontier models. Students do not need to build one, but knowing the idea explains how models keep getting more capable without becoming impossibly expensive to run.
RelatedFrontier model · Foundation model · Scaling laws · Large language model (LLM)
When an AI system becomes less accurate over time because the world, or the data it sees, has changed since it was trained.
In Australian classrooms. Drift is a reminder that AI is not a finished, fixed authority. For students it reinforces the habit of checking currency, a confident answer can simply be out of date.
RelatedTraining data · Verification · Failure modes · Bias
A coordinated set of AI agents, each with a specialised role, that communicate to complete a complex task, for example a planner that decomposes the work and a reviewer that checks it.
In Australian classrooms. Multi-agent designs are an advanced topic Edison's senior students explore, with the same discipline as everywhere else: more agents only when the work genuinely needs more than one specialised step.
RelatedAI agent · Agentic AI · Orchestration · Agent workflow
AI models whose code or trained weights are published, so anyone can inspect, run or adapt them, in contrast to a closed, company-controlled model.
In Australian classrooms. Open models let students see inside the machine and even run one themselves, which is a powerful learning experience. The open-versus-closed question is also a good lens on who controls a technology.
RelatedClosed model · Foundation model · Frontier model · Large language model (LLM)
Coordinating multiple AI tools, agents, data sources and steps into a single working process, deciding what runs when, and where a human checks in.
In Australian classrooms. Orchestration is systems thinking applied to AI, and a natural project for Edison's builders. Designing the flow forces students to be explicit about goals, checks and points of human judgement.
RelatedAgent workflow · Multi-agent system · Systems thinking · Tool use
The everyday discipline of judging a single AI output's quality, accuracy, tone, completeness and safety, before using it.
In Australian classrooms. Where 'evals' is systematic testing, output evaluation is the in-the-moment habit a student needs on every task. Edison's workshops train it directly, because slicker output with shakier understanding is the warning sign, not the win.
RelatedEvaluation (evals) · Verification · Hallucination · Human-in-the-loop
A thought experiment about an AI given a simple goal, make paperclips, that pursues it in extreme, harmful ways because nothing in the goal captures human values. A vivid illustration of the alignment problem.
In Australian classrooms. It is a memorable way into a serious idea, why goals must be specified with care. Edison uses it to open conversations about ethics and unintended consequences without tipping into alarmism.
RelatedAlignment problem · Human-centred AI · Guardrails · Responsible innovation
Learning that produces visible proof of capability, projects, artefacts, prototypes and case studies, rather than only grades on a transcript.
In Australian classrooms. As the graduate bar rises, a body of real work increasingly speaks louder than marks alone. Edison students leave with portfolio-ready artefacts they can explain and defend.
RelatedProject-based learning · Applied intelligence · Builder mindset · Future readiness
Learning by building something real, where understanding is demonstrated through a made artefact rather than recalled in a test.
In Australian classrooms. Project-based learning is how Edison teaches: students make and verify genuine work, the view at the heart of Harvard's Project Zero tradition, where understanding is a performance you can demonstrate, not a fact you can recite.
RelatedPortfolio learning · Applied intelligence · Cognitive apprenticeship · Prototype thinking
The input given to an AI model, the instructions, context, examples and the actual task. The prompt is the primary way a person controls what a model does.
In Australian classrooms. For a student, the gap between mediocre and excellent AI output is almost always the prompt. Edison teaches prompting as structured thinking made explicit: clear context, a clear ask, clear constraints.
RelatedPrompt engineering · Few-shot prompting · System prompt · Chain-of-thought
The craft of writing, testing and improving prompts to get reliable, high-quality output, combining clear instruction, examples, role-setting and a specified output format.
In Australian classrooms. Edison treats prompt engineering as a transferable thinking skill, not a bag of tricks. The framing and constraint it demands are the same skills that make a student a clearer writer and researcher.
RelatedPrompt · Few-shot prompting · Structured thinking · AI fluency
Turning an idea into a rough, testable version quickly, then improving it through feedback, rather than waiting for a perfect plan before starting.
In Australian classrooms. AI compresses the time from idea to first version, which makes prototype thinking more valuable, not less. Edison students learn to ship a draft, test it and iterate, the loop real builders live in.
RelatedBuilder mindset · Creative confidence · Project-based learning · Machine-assisted creativity
A pattern where the model first retrieves relevant documents from a chosen source, then writes an answer grounded in them, rather than relying only on what it learned in training.
In Australian classrooms. RAG is how an AI is made to answer from real, checkable material, a textbook, a set of notes, primary sources, which sharply reduces hallucination. It is the technical backbone of trustworthy study tools.
RelatedEmbedding · Hallucination · Context window · Verification
Building and using AI with ethics, safety and social awareness designed in from the start, rather than bolted on afterwards as a warning.
In Australian classrooms. Responsible innovation maps to the Australian Framework for Generative AI in Schools and to UNESCO's ethics dimension. At Edison it is built into every project, because responsible use is taught, not assumed.
RelatedHuman-centred AI · Guardrails · Alignment problem · Digital agency
Giving learners structured support to attempt something just beyond their current ability, then gradually removing that support as they become capable of doing it alone.
In Australian classrooms. Scaffolding is core learning science and core to how Edison sequences AI capability, support at first, independence by the end. The goal is always a student who can do it with the tool closed.
RelatedCognitive apprenticeship · Capability pathway · Metacognition · Project-based learning
The observation that a model's performance tends to improve predictably as you increase data, computation and parameters, the empirical engine behind the last decade of AI progress.
In Australian classrooms. Scaling laws explain why models kept getting better, and the Bitter Lesson explains why that mattered more than hand-crafted cleverness. Together they are a rich topic for senior students.
RelatedThe Bitter Lesson · Emergence · Frontier model · Foundation model
Assessing whether information is trustworthy, current and relevant, who produced it, why, and how it can be corroborated.
In Australian classrooms. Source evaluation is a long-standing research skill made newly urgent by fluent AI. Edison teaches students to trace an AI's claims to primary sources before they rely on them.
RelatedVerification · Epistemic vigilance · Bias · Hallucination
A critical metaphor, from a 2021 paper by Bender, Gebru and colleagues, suggesting that a language model can generate convincing text by repeating learned patterns without genuine understanding.
In Australian classrooms. It is a sharp counterweight to the hype, and a useful caution for students: fluency is not comprehension. Edison pairs it with the Chinese Room to ground why human verification stays essential.
RelatedThe Chinese Room · Hallucination · Large language model (LLM) · Epistemic vigilance
The ability to decide what matters, what is true, and what to do next, the capacity to direct effort and tools toward a worthwhile end.
In Australian classrooms. Strategic judgement is the most valuable thing a student can carry out the school gate, because it is precisely what AI does not supply. The Edison Method is built to develop it deliberately.
RelatedStructured thinking · Future readiness · AI fluency · Systems thinking
Clear, organised reasoning, breaking a question into parts, ordering them, and working through them deliberately rather than all at once.
In Australian classrooms. Structured thinking is what makes both a strong essay and a strong prompt. Edison treats it as the underlying skill that prompting, research and project work all draw on.
RelatedComputational thinking · Systems thinking · Prompt engineering · Strategic judgement
The behind-the-scenes instructions set by an application, not the end user, that define a model's persona, scope, format rules and safety boundaries for every conversation.
In Australian classrooms. School-built tools like NSWEduChat run inside locked system prompts that enforce safe behaviour. Knowing the layer exists helps students understand why a tool will, or will not, do certain things.
RelatedPrompt · Guardrails · AI agent · Closed model
Understanding how tools, people, workflows, incentives and decisions connect and influence one another, seeing the whole, not just the parts.
In Australian classrooms. Systems thinking is powerful for AI education because real AI value lives in workflows, not single prompts. It is a recurring theme across Edison's senior, build-focused programs.
RelatedComputational thinking · Orchestration · Structured thinking · Strategic judgement
The unit of text an AI model reads and writes, roughly three-quarters of an English word. Token counts drive how much a model can read, how fast it responds, and what it costs.
In Australian classrooms. Tokens are the model's atoms. The idea explains why very long documents can exceed what a model can hold at once, and why concise prompting is often clearer as well as cheaper.
RelatedContext window · Large language model (LLM) · Embedding · Generative AI
When an AI can reach beyond text to use external tools, search, a calculator, code, files, a calendar, to get a task done, rather than relying only on what it already knows.
In Australian classrooms. Tool use is what turns a chatbot into something that can act, and the point where supervision matters most. Edison students design which tools an AI may use, and where a human signs off.
RelatedAI agent · Agentic AI · Orchestration · Agent workflow
The text, code and other content a model learns from. Its composition, breadth and recency largely determine what a model knows and where its biases sit.
In Australian classrooms. Off-the-shelf models over-index on US content, which is why Australian context and spelling have to be prompted for. Edison grounds models in Australian and student-specific material rather than assuming the defaults fit.
RelatedBias · Fine-tuning · Model drift · RAG (Retrieval-Augmented Generation)
Alan Turing's 1950 proposal that a machine could be judged 'intelligent' if a person could not reliably tell it apart from a human in conversation. A test of appearance, not understanding.
In Australian classrooms. A foundational idea for any AI education, and a great prompt for debate now that systems routinely pass casual versions of it. Edison uses it to separate sounding human from thinking, the distinction students must hold onto.
RelatedThe Chinese Room · Stochastic parrot · Emergence · Explainability
Learning by building real products, projects or concepts that could genuinely exist in the world, with the ambition, scoping and ownership that implies.
In Australian classrooms. Venture-led learning is where Edison's Innovators stretch furthest, taking an idea through to a defensible prototype. It fuses the builder and founder mindsets with real AI-enabled execution.
RelatedFounder mindset · Builder mindset · Prototype thinking · Portfolio learning
Checking whether an AI's output is accurate, credible and usable, treating every claim that matters as a draft to be confirmed against a real source.
In Australian classrooms. Verification is the reflex Edison works hardest to install, because it is the habit that keeps a student on the right side of the cognitive-offloading curve. Assume the confident answer might be confidently wrong.
RelatedEpistemic vigilance · Source evaluation · Hallucination · Output evaluation
