Sherlockian Prompt Design(3)
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.
4. Hypothetico-Deductive Reasoning:
Form Hypothesis, Then Test It
Hypothetico-deductive reasoning begins with a hypothesis and logically predicts what must follow if the hypothesis is true; it then tests those predicted consequences against reality to evaluate the hypothesis. This is the workhorse of scientific inquiry, and in prompt design it is powerful wherever we construct “if…then…” scenarios and trace their logical outcomes—policy analysis, customer-behavior prediction, strategy simulation, and creative plotting.
Exercise ① Validating a User-Behavior Scenario
Scenario. You run an education app. Quiz accuracy has suddenly dropped. Your hypothesis: “Longer ad videos are degrading user focus.” Design a prompt that helps test whether this is true.
Your Prompt:
(Write here.)
Exercise ② Predicting Responses in a Dialogue System
Scenario. You are designing an empathetic chatbot. Hypothesis: “When users use negative expressions, they prefer analytical responses over empathic ones.” Build a prompt that predicts likely outcomes.
Your Prompt:
(Write here.)
Exercise ③ Forecasting Policy Effects
Scenario. Your team is designing an urban transport policy. Hypothesis: “Expanding bike-priority lanes can reduce traffic congestion.” Ask the AI to examine this via simulation.
Your Prompt:
(Write here.)
Exercise ④ Re-plotting a Story by Counterfactual
Scenario. You are writing a detective novel. Premise: “If the protagonist knew the culprit from the start, the ending would be entirely different.” Prompt the AI to redesign the plot line accordingly.
Your Prompt:
(Write here.)
Closing note: A hypothesis gives thought a direction
Hypothetico-deductive prompts trade haste for a designed path of reasoning. Whatever you ask an AI, embedding a “What if X were true?” will make the response more disciplined and directional. As Holmes would say, “If he did X, then what follows?”—structure the inquiry for the AI, and the answer begins to read like the product of real thinking.
4. Hypothetico-Deductive — Sample Section
Exercise ① User-behavior validation
Sample Prompt.
“Hypothesis: Quiz accuracy fell because longer ad videos reduced users’ immersion. To test this, which datasets should we compare? Identify the single most decisive metric and explain possible interpretations.”
Design note. Do not ask “Why did it happen?” Present a hypothesis and have the AI design the experiment. Hypothesis-led questioning naturally yields thought experiments.
Exercise ② Dialogue response preference
Sample Prompt.
“Premise: When a user says ‘I’m exhausted and it all feels pointless,’ they will prefer an analytical response (‘Shall we map what’s hardest?’) to an empathic one (‘That must be tough’). Compare the two responses and predict which yields more positive feedback, providing at least two reasons.”
Design note. Shift from mere response generation to scenario-based evaluation.
Exercise ③ Policy effect forecast
Sample Prompt.
“If bike-priority lanes are expanded by 30% over current levels, predict changes in vehicle throughput, public-transit ridership, and perceived inconvenience. Ground the predictions in plausible reasoning, and summarize positive vs. negative scenarios.”
Design note. Reframe “What will happen?” as “If this assumption holds, what changes?” to deepen predictive reasoning.
Exercise ④ Counterfactual plotting
Sample Prompt.
“Rebuild the ending in three stages under the premise that ‘the protagonist knew the culprit from the beginning.’ At each stage, reflect the protagonist’s actions, emotions, and the reader’s expectations. Aim for a conclusion more tense and more logical than the original.”
Design note. Use a hypothesis as a creative starting point and restructure the entire arc—logic and imagination working in tandem.
