Session 13: Chain-of-Thought Prompting — Making AI Think Step by Step

• Understand what chain-of-thought prompting is and why it produces better results
• Add 'think step by step' and similar phrases to your prompts
• Compare direct-answer responses to step-by-step reasoning responses on the same question
• Apply chain-of-thought prompting to complex academic and analytical questions

Imagine asking a detective to tell you who committed a crime. If they just blurt out a name with no explanation, you'd have no way of evaluating whether their conclusion is trustworthy. But if they walk you through the evidence — explaining what each clue means and how each piece connects to the next — you can follow their reasoning, check it against your own understanding, and decide whether you find it convincing. Chain-of-thought prompting applies this same logic to AI.

Chain-of-thought prompting is a technique where you explicitly ask the AI to show its reasoning process step by step before arriving at a conclusion. The magic phrase is simple: "Think through this step by step" or "Let's work through this together, step by step." Adding these phrases to a complex question reliably produces responses that are longer, more logical, and more thoroughly reasoned than direct-answer responses to the same question.

Why does this work? AI models generate responses word by word, predicting what comes next. When an AI is forced to articulate a reasoning step before moving to the next one, each subsequent step benefits from the explicit logic that came before. It's similar to how writing out your thinking on paper often helps you arrive at better ideas than trying to think everything through in your head at once. The act of externalizing the reasoning process improves the reasoning itself.

For graduate students and academic researchers, chain-of-thought prompting is invaluable for several types of tasks. When you're analyzing a complex argument in a text, step-by-step reasoning helps you see whether the logic holds at each stage. When you're choosing between two research methodologies, a step-by-step comparison evaluates the tradeoffs more thoroughly. When you're planning a dissertation chapter, step-by-step breakdown of how the argument should develop is more useful than a high-level outline.

You'll also notice that chain-of-thought responses are often more honest about uncertainty. When AI is asked to show its reasoning, it's more likely to say things like 'this depends on' or 'there are two competing considerations here' — which is more truthful and more educationally valuable than confident-sounding oversimplifications. This kind of nuanced reasoning is closer to the academic thinking you're developing.

Today's exercise involves asking the same question twice: once for a direct answer, and once with the step-by-step instruction added. The contrast between the two responses will make the value of this technique immediately obvious. Once you've seen it work, you'll use it every time you face a complex analytical question.