Sherlockian Prompt Design (6)
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.
7. Elimination: Rule Out Impossible and Narrow to Answer
Exercise ① Eliminating Unsuitable Travel Destinations
Scenario. You must choose one of five cities for a vacation. Conditions are:
Your Prompt:
(Write here.)
Design Note.
Instruct the model to (1) list explicit criteria, (2) test each city against each criterion, and (3) eliminate step by step with a short justification trail. Require a final tie-breaker rule to avoid indecision.
Exercise ② Selecting a Hiring Candidate
Scenario. You are an HR manager with three finalists. Each has strengths, but only one can meet all company criteria: team-centric collaboration, strong efficiency within regular hours (no overtime culture), and fast work tempo.
Your Prompt:
(Write here.)
Design Note.
Ask for a criteria × candidate matrix, then a sequential elimination (least fit removed first), ending with the best-remaining and residual risk notes. This preserves transparency and prevents halo bias.
Exercise ③ Tracing the Cause of a Product Fault
Scenario. Customers report that a specific function in a shipped electronics product is failing. The production process has four stages, with three potential failure points. As after-sales support, you must identify the likeliest cause.
Your Prompt:
(Write here.)
Design Note.
Have the model test each suspect stage against the observed failure signature (“one function, repeatable”). Require disconfirming evidence (what would make this cause unlikely). Elimination gains strength from negative tests, not just positives.
Exercise ④ Finding the Culprit Inside a Story Structure
Scenario. In a short mystery, there are three suspects.
Your Prompt:
(Write here.)
Design Note.
Order the model to compare motive, opportunity, and timeline against the fixed constraints of the plot; eliminate inconsistencies first, then articulate a Holmes-style rationale for the survivor. This makes the conclusion inevitable by exclusion, not assertion.
7. Elimination — Sample Prompts
Exercise ① Travel destination
Sample Prompt.
“From the cities below, eliminate unsuitable options step by step using these conditions: flight time within 5 hours, average July temperature below 30°C, and presence of nature-oriented sights. At each elimination step, briefly state the reason. Leave one final candidate.”
Design Note.
Stepwise elimination + justification trail = an auditable path to the answer, not a black box.
Exercise ② Hiring decision
Sample Prompt.
“Here are the traits of candidates A, B, and C. The company needs someone who is highly collaborative, productive within non-overtime hours, and fast in execution. Compare each candidate against the criteria, eliminate the least suitable first, and identify the final choice.”
Design Note.
A structured rubric reduces bias; forced elimination order exposes trade-offs and clarifies why one candidate prevails.
Exercise ③ Troubleshooting a product fault
Sample Prompt.
“You’re the after-sales lead. A specific feature in our product is failing repeatedly. Suspect processes: (1) component assembly, (2) QA testing, (3) final packaging. Given that the failure is isolated to a single function, review each process and eliminate the less likely causes one by one. Conclude with the most probable cause and your rationale.”
Design Note.
Tying each step to the failure signature suppresses speculation and keeps reasoning strictly evidence-constrained.
Exercise ④ Identifying the culprit in story structureSample Prompt.
“Among the three suspects below, infer the culprit by ruling out inconsistencies against the story’s known constraints (motive, opportunity, timeline). Remove the least compatible suspect first and justify each elimination. Present the remaining suspect with a concise, Holmes-style rationale.”
Design Note.
Elimination turns mystery-solving into a constraint satisfaction task; the final answer is the last consistent survivor.
