Getting the 'Why' from AI: A Guide to Human-Centric AI Interactions

AI is powerful—but only if you ask the right questions. Generic prompts yield mechanical, step-by-step responses. To get meaningful, human-focused explanations, ask for the "why" and include context.

The Issue with Generic Prompts

Example JavaScript function:

function calculateTotalPrice(items) {
  let total = 0;
  for (let i = 0; i < items.length; i++) {
    total += items[i].price * items[i].quantity;
  }
  return total;
}

A prompt like "Explain this code" often just lists what the code does, missing its purpose.

How to Get Better Explanations

User Perspective:

Explain as if talking to a customer calculating their shopping cart total.

Purpose Focus:

What problem does this code solve? Why is it written this way?

Context:

Assume this code is part of an online store’s order processing. Describe its role.

Example Prompts

User-Centric:

Explain this code as if explaining to a user on an e-commerce site.

Purpose-Focused:

Explain the purpose of this code. What problem does it address?

Contextual:

Assume this code is part of an online store. Describe its role in order processing.

Iterative Refinement

If the answer remains too technical, ask follow-up questions, for example: What does the items array represent here?

Summary

Focus your prompts on "why" rather than just "how." This approach yields explanations that are practical, contextual, and user-centered.

Why this matters

• Clear “why” leads to better “how.” AI can surface trade-offs you might miss.
• Teams align faster when rationale is written down.
• Decisions become teachable artifacts, not tribal knowledge.

How to use this today

• Ask AI to list options with pros/cons and a recommendation.
• Require a short rationale in PRs for non-trivial changes.
• Save rationales in ADRs (Architecture Decision Records).

Common pitfalls

• False certainty: ask for confidence levels and unknowns.
• Cherry-picking: compare at least two viable paths.
• Stale decisions: revisit when assumptions change.

What to try next

• Generate ADR templates prefilled from issues.
• Ask AI to link decisions to metrics and alert thresholds.
• Run quarterly reviews of old decisions for drift.