Sherlockian Prompt Design(2)
Sherlockian Way of Thinking
☞ How to use this chapter
For each method, read the “Scenario” carefully. Under “Your Prompt,” draft the best prompt you can. At the end of the chapter, compare with the sample prompts and design notes, and then revise your prompt once more.
3. Induction: Stack Cases to Surface Pattern
Induction observes individual cases, detects recurring elements or common structure, and from them derives general rules or tendencies. In prompt design, induction is better for sensing “flow” and “trend” or for proposing new taxonomies than for forcing sharp, single-point answers.
We will practice listing multiple cases and prompting the AI to detect and summarize patterns—or to generate a new case that follows from them.
Exercise ① Analyzing Tone & Style
Scenario. You will show the AI five newsletters you have written so far and ask it to summarize your tone and stylistic signatures, so that future issues remain consistent.
Your Prompt:
(Write here.)
Exercise ② Extracting Purchase Drivers From Reviews
Scenario. You market dishwashers. You will provide ten customer reviews and ask the AI to extract the primary purchase drivers.
Your Prompt:
(Write here.)
Exercise ③ Designing Characters via Sentiment Hotspots
Scenario. You are writing a long web novel. From chapters written so far, you will collect the scenes that provoked the strongest reader emotions and look for “what scene structures heighten emotional immersion.”
Your Prompt:
(Write here.)
Exercise ④ Inferring Dispositions from Interview Answers
Scenario. You have answers from three CEOs. You want to infer their leadership styles and the values that guide their decisions.
Your Prompt:
(Write here.)
Closing note: Cases are mirrors for thought
Inductive prompts do not ask for “the one right answer.” They invite the AI to read a flow in the repetition, interpret structure, and derive new rules or direction. Holmes could infer the trade and habits of a man from tiny creases and dust on a worn hat because he had seen many such hats; so too, with AI, our inductive questions are designs that make it detect, summarize, and project patterns—linking them to creative prediction.
3. Induction — Sample Section
Exercise ① Tone & style
Sample Prompt.
“These five pieces are my newsletters. Summarize three common features in voice, diction, and structure. Then propose two cautions to keep the style consistent in future issues.”
Exercise ② Review-based purchase drivers
Sample Prompt.
“These are ten reviews of our dishwasher. Focusing on recurring words or phrases, summarize the top three reasons customers chose to buy. Include one example sentence for each reason.”
Design note. The AI reads not mere word repetition but recurring meanings, activating its pattern-recognition strengths.
Exercise ③ Emotion-driven scene design
Sample Prompt.
“These five scenes elicited the strongest reader reactions. Identify two or three common features in their structure or technique. Then propose one new scene idea that can achieve a similar effect.”
Design note. “Emotional” scenes are not defined by many emotion words; they often hinge on recurrent structures—contrast, silence before eruption, the unexpected—which the AI can induce.
Exercise ④ Disposition from interviews
Sample Prompt.
“These are key answers from three CEOs. Based on recurring words, expressions, and thinking patterns, summarize each CEO’s leadership style in three keywords. For each keyword, predict what managerial choices they are likely to make.”
Design note. What is nebulous for humans becomes tractable for AI by inductively tying recurring traits to behavior predictions.
