Outline
Prompt Design — Thinking in Steps
Level
Intermediate
7.1 Learning Goals
By the end of this section, your mandate is to:
Understand the principle of step-based reasoning, which I call chain-of-thought.
Apply “Explain step by step” prompting to complex problems in Maths, Science, and English.
Differentiate between demanding just an answer versus demanding the entire process.
Design prompts that teach you the method, not just the result.
7.2 Why Steps Matter
I believe the final answer to a problem is often the least important part of the solution. Exams, and indeed real-world challenges, reward the method used to reach the answer. If you cannot explain how you arrived at a conclusion, you have not mastered the concept; you have merely memorized a fact.
An amateur asks an AI for the answer. A professional instructs the AI to reveal its steps. This distinction is critical. Demanding the process forces the AI to expose its logic, which you can then use to validate your own thinking and deepen your understanding.
7.3 Chain-of-Thought Prompting
A chain-of-thought (CoT) prompt is a specific instruction that commands the AI to break down a problem into its logical stages and explain them sequentially. It shifts the AI's role from a simple answer machine to a process tutor.
Weak Prompt: “Solve 12 × 15.” (This gives you a number, but teaches you nothing.)
Strong CoT Prompt: “Explain step by step how to solve 12 × 15 using the long multiplication method.” (This gives you a transferable skill.)
7.4 How to Phrase CoT Prompts
To initiate a chain-of-thought response, you must use clear, direct command phrases. I mandate the inclusion of phrases like “Explain step by step,” “Show the reasoning,” or “Walk me through the logic.”
These phrases transform a simple query into a demand for process. For example, asking the AI to “Explain the stages of photosynthesis step by step” is far superior to just asking “What is photosynthesis?” because it forces a structured, logical output.
7.5 Worked Example: Algebra
A common task is solving a quadratic equation. A CoT prompt makes the AI’s logical path visible, allowing you to check it against your own.
Prompt: “Solve for x in the equation x² + 5x + 6 = 0. Explain your reasoning step by step.”
An effective AI output will not just give the answers. It will demonstrate the process, showing you how it first recognizes the quadratic form, then factorizes the expression into (x+2)(x+3), sets each factor to zero, and finally derives the solutions. This output shows process → method → answer, which is the only sequence that builds true understanding.
7.6 Science Example
In scientific explanations, order and sequence are critical. A CoT prompt ensures the AI respects that structure.
Prompt: “Explain step by step how food travels through the human digestive system, naming each primary organ in order and its main function.”
The resulting output should be a clear, sequential journey from the mouth to the oesophagus, stomach, and intestines. This is far more valuable for revision than a generic paragraph about digestion.
7.7 English Example
Even in writing, process is key. The PEEL paragraph structure is a specific method, and a CoT prompt can be used to deconstruct it.
Prompt: “Explain step by step how to construct a PEEL paragraph on the topic: ‘The Importance of Healthy Habits.’”
The AI should then break down the process: how to formulate a clear Point, find supporting Evidence, Explain the connection, and Link it back to the main topic.
7.8 Weak vs Strong Comparison
The difference in output quality is not subtle; it is absolute.
Weak Prompt
Strong CoT Prompt
“Explain digestion.”
“Explain digestion step by step, naming each organ in order.”
“What is mitosis?”
“List the stages of mitosis step by step, including the key event in each stage.”
7.9 Practice Drill
Your task is to rewrite these vague prompts into precise CoT prompts that demand a process.
“Maths: Solve a quadratic.”
“Science: Explain photosynthesis.”
“English: Write an essay.”
7.10 Group Activity
In groups of three, each member will pick a different subject (Maths, Science, English) and write a precise CoT prompt for a topic within it. You will then execute the prompts and compare the outputs. Your analysis should focus on which steps were clear and where the AI may have skipped important details.
7.11 Self-Check
Why is step-by-step prompting a superior method for learning?
Write one effective CoT prompt for a Maths topic and one for a Science topic.
What command phrase must always appear in a CoT prompt?
7.12 Key Takeaway
My view is that final answers are shallow. The process required to reach them is deep. You must train yourself to always demand the process from your tools and from yourself. The law of learning is clear: Process first, answer second.