Sherlockian Way of Thinking
The method of elimination removes options that conflict with facts or logic until only one candidate survives. Simple, powerful, and especially useful under uncertainty.
Holmes’s dictum from A Study in Scarlet is classic: “When you have eliminated the impossible, whatever remains, however improbable, must be the truth.” This is not rhetoric; it is a working algorithm.
In prompts, elimination means presenting alternatives, checking each against constraints or evidence, and discarding misfits in sequence.
1. Lay out the candidate set.
2. Test each against conditions or evidence.
3. Remove conflicts until a best remaining option stands.
Use this for bounded decisions, mutually exclusive options, and process-tracing.
1. Career triage: Given an introverted person preferring repetitive solo work, evaluate (A) retail manager, (B) data analyst, (C) customer service; remove misaligned roles and justify the survivor.
2. Suspect pruning: With three suspects and partial alibis, eliminate those who cannot be guilty before naming the lead.
3. Laptop choice: Four models, fixed constraints; cross off non-compliers, leave the final pick.
4. Holmesian techniques: From four skills, remove those not evidenced in Holmes’s practice to find the outlier.
Elimination prompts embed a process in the question, guiding the model to reason by subtraction rather than by blind selection.
We have re-cast six of Holmes’s methods as prompt strategies for generative AI. A prompt can be far more than an instruction; it can be a device for structuring thought. Like Holmes, our questions must stand on logic and speak the language of design—not merely of data.
Yet understanding the forms is only the beginning. What matters is practice: composing questions, observing the model’s responses, and refining one’s own habits of mind through the exchange.
In the next chapter, we will put these tools to work. Prompts are born in the mind, but their power emerges when written by the hand and tested by the eye. Now, let us train ourselves to think—and to question—like Sherlock Holmes.