Work with AI
effectively,
ethically,
and safely.

Understanding Generative AI

You don't need to be an engineer, but knowing the basics helps you set realistic expectations and spot problems before they matter.

Use this table to decide which mode fits your current task.

• Is this task well-defined enough for AI to attempt without confusion?
• Which mode fits best — automation, augmentation, or agency?
• Does this task require my domain expertise, relationships, or judgment?
• What are the risks if AI gets this wrong?
• Am I the right person to review and take responsibility for the output?

Not every prompt needs all six elements — but the more you include, the less AI has to guess.

Adapt this template when disclosing AI involvement in your work.

Every term from the course

Click any term to expand its definition. Filter by category using the buttons below.

• AI FluencyFramework +
  The ability to work with AI systems in ways that are effective, efficient, ethical, and safe.
• The 4DsFramework +
  The four core competencies of AI Fluency: Delegation, Description, Discernment, and Diligence.
• DelegationFramework +
  Deciding what work should be done by humans, by AI, or collaboratively — including understanding goals, AI capabilities, and making strategic choices about task distribution.
• DescriptionFramework +
  Effectively communicating with AI systems — defining desired outputs, guiding processes, and specifying how AI should behave during your collaboration.
• DiscernmentFramework +
  Thoughtfully evaluating AI outputs, processes, and behaviors — assessing quality, accuracy, relevance, and the quality of the AI's reasoning.
• DiligenceFramework +
  Using AI responsibly and ethically — making thoughtful choices, maintaining transparency about AI's role, and taking full accountability for AI-assisted work.
• AutomationFramework · Mode +
  A modality of Human-AI interaction where AI performs specific tasks based on direct human instructions. The human defines; the AI executes.
• AugmentationFramework · Mode +
  A modality of Human-AI interaction where humans and AI collaborate iteratively as thinking partners. Both contribute to the outcome.
• AgencyFramework · Mode +
  A modality where humans configure AI to work independently on their behalf, establishing AI knowledge and behavior patterns rather than specifying exact actions each time.
• Generative AITechnical +
  AI systems that create new content — text, images, code, audio — rather than simply classifying or analyzing existing data.
• Large Language Model (LLM)Technical +
  Generative AI systems trained on vast amounts of text data to understand and generate human language. Claude, GPT-4, and Gemini are all LLMs.
• ParametersTechnical +
  The mathematical values within an AI model that determine how it processes and generates information. Modern LLMs contain billions or even trillions of parameters.
• Transformer ArchitectureTechnical +
  The breakthrough AI design from 2017 that enables LLMs to process sequences of text in parallel while attending to relationships across long passages. The foundation of all modern LLMs.
• Context WindowTechnical +
  The amount of information an AI can consider at one time, including the full conversation history and any documents you've shared. Larger context windows = more working memory.
• HallucinationTechnical +
  When an AI confidently states something that sounds plausible but is factually incorrect. A major reason why Discernment and Deployment Diligence are essential.
• Knowledge Cutoff DateTechnical +
  The point after which an AI model has no built-in knowledge of the world, based on when its training data was collected. Events after this date are unknown unless provided in the prompt.
• Pre-trainingTechnical +
  The initial training phase where an AI learns patterns from vast amounts of text, developing foundational language understanding and broad world knowledge.
• Fine-tuningTechnical +
  Additional training after pre-training where models learn to follow instructions helpfully, communicate clearly, and avoid harmful outputs. This is what makes a raw LLM into a useful assistant.
• TemperatureTechnical +
  A setting that controls the randomness of AI responses. Higher temperature = more creative and varied; lower temperature = more predictable and focused.
• RAG (Retrieval Augmented Generation)Technical +
  A technique that connects AI to external knowledge sources at query time, allowing the model to access up-to-date or domain-specific information and reduce hallucinations.
• BiasTechnical +
  Systematic patterns in AI outputs that unfairly favor or disadvantage certain groups, perspectives, or viewpoints — often reflecting patterns present in the training data.
• Scaling LawsTechnical +
  The observed pattern that as AI models grow larger and train on more data, performance improves consistently — and entirely new capabilities can emerge at certain scale thresholds.
• PromptPrompting +
  The input given to an AI model — including your instructions, context, examples, and any documents or data you share. The quality of the prompt heavily determines the quality of the output.
• Prompt EngineeringPrompting +
  The practice of designing effective prompts to produce desired outputs — combining clear communication skills with AI-specific techniques like role-setting, chain-of-thought, and few-shot examples.
• Chain-of-Thought PromptingPrompting +
  A technique that encourages AI to work through a problem step-by-step before giving its final answer, leading to more accurate and well-reasoned responses on complex tasks.
• Few-Shot Learning (N-Shot)Prompting +
  Teaching AI by showing examples of the desired input-output pattern within the prompt itself. The "N" is the number of examples provided. Even 1–2 examples (one-shot/two-shot) can dramatically improve results.
• Role / Persona DefinitionPrompting +
  Specifying a character, expertise level, or communication style for AI to adopt. A well-defined role shapes not just the content but the tone, depth, and framing of the response.
• Output ConstraintsPrompting +
  Explicitly specifying the desired format, length, structure, or characteristics of AI's response. Without constraints, AI will make its own guesses — which may not match what you actually need.
• Think-First ApproachPrompting +
  Asking AI to work through its reasoning before providing a final answer. Leads to more thorough, self-checked responses — especially valuable for complex or multi-step problems.