Appendix B — Glossary

This glossary provides definitions for key terms used throughout “Intentional Prompting.”

B.1 AI-Related Terms

AI Coding Assistant: Software tools that use artificial intelligence to help programmers write, understand, and debug code. Examples include GitHub Copilot, Amazon CodeWhisperer, and conversational AI models like Claude and ChatGPT when used for programming tasks.

Chain-of-Thought Prompting: A technique that encourages AI to break down complex reasoning into intermediate steps, producing more comprehensive and accurate outputs.

Few-Shot Prompting: Providing an AI with examples of the desired input-output behaviour before presenting the actual task, helping guide the model’s response format and approach.

Foundation Model: Large AI models trained on vast datasets that serve as the basis for more specialised applications, including coding assistants.

Large Language Model (LLM): AI systems trained on massive text datasets that can generate human-like text and code based on prompts.

Prompt Engineering: The practice of crafting effective instructions for AI systems to produce desired outputs, involving specific techniques and patterns.

Zero-Shot Prompting: Asking an AI to perform a task without providing examples, relying on the model’s pre-existing knowledge. A useful technique on its own. However, chaining zero-shot prompts back-to-back without pushback or verification between exchanges is delegation in volume, not conversation. Intentional prompting treats each prompt as one move in a structured exchange, not a self-contained transaction.

B.2 Human-AI Collaboration Concepts

Human Agency: The principle that the human remains the architect and decision-maker throughout an AI-assisted process - the capacity to choose, weigh, and own the outcome. In intentional prompting, agency is preserved by framing the problem yourself, evaluating outputs critically, and making final integration decisions based on your own understanding rather than AI convenience.

Human Oversight: The practice of reviewing, validating, and correcting AI-generated output before accepting it. Where agency is the principle (the human is in charge), oversight is the everyday practice that keeps the principle real (the human actually checks).

Human-in-the-Loop (HIL): A system-design and AI-governance term for arrangements where a person remains an active participant in decisions an AI system could otherwise automate. HIL is the structural counterpart to human agency: the same stance described from the perspective of the system being built rather than the person using it.

Directive vs. Delegative Approach: Two contrasting ways of working with AI coding assistants. The delegative approach hands problems off and accepts the AI’s solution with minimal scrutiny. The directive approach guides the AI through a structured process, retains control over architecture and design decisions, and critically evaluates outputs at each step. Intentional prompting is built on the directive approach.

B.3 Intentional Prompting Concepts

Intentional Prompting: A deliberate approach to programming that leverages AI tools while maintaining human control over the development process. It focuses on understanding, guiding AI through structured methodology, using AI as a learning tool, and maintaining the human developer as the architect and decision-maker.

Six-Step Methodology: The structured approach at the core of intentional prompting: 1. Restate the problem 2. Identify input and output 3. Work the problem by hand 4. Write pseudocode 5. Convert to code 6. Test with data

Challenge Prompts: Deliberately introducing programming challenges to test understanding and explore potential issues.

Clarification Prompts: Questions that refine requirements and explore edge cases before implementation.

Foundation Prompts: Initial prompts that establish baseline requirements for a programming task.

Refinement Prompts: Prompts that push for code improvements based on best practices and efficiency considerations.

Scaffolding Prompts: Prompts that support learning by breaking down complex concepts into understandable components.

B.4 Programming Concepts

Code Smell: Patterns in code that may indicate deeper problems or potential for improvement.

Edge Case: A problem or situation that occurs only at an extreme operating parameter, often requiring special handling in code.

Pseudocode: A simplified, high-level description of a computer program or algorithm, using structural conventions of programming languages but intended for human reading rather than machine reading.

Refactoring: The process of restructuring existing code without changing its external behaviour, typically to improve non-functional attributes like readability, reduce complexity, or improve maintainability.

Technical Debt: The implied cost of additional rework caused by choosing an easy (but limited) solution now instead of using a better approach that would take longer.

B.5 Teaching and Learning Terms

AI-Proof Assignment: Assignments designed to be difficult for AI coding assistants to solve directly, often focusing on conceptual understanding or novel problems.

Documentation-First Methodology: An approach where students create detailed specifications and documentation before writing any code, using AI to evaluate completeness and clarity.

Process-Based Assessment: Evaluating students based on their problem-solving approach and understanding rather than just the final code output.