Skip to main content
Skip to content
Menu

Responsible AI Intermediate 3 parts

Audit your AI prompts for hidden incentives misaligned with your values

You'll have a checklist of your 5-10 most-used AI tasks with annotations showing what metric or outcome each prompt is actually optimizing for versus what you intended.

What you'll make

Prompt Values Audit

  1. Current Prompt Inventory: task + hidden optimization target
  2. Misalignment Analysis: where optimization conflicts with values
  3. Redesigned Prompts: uncertainty-surfacing alternatives

The 3 parts

Each part ends with a check you can see.

  1. Map Your Optimization Patterns

    Inventory your most-used AI tasks and identify what each prompt actually optimizes for

    Done when

    You have 5-10 documented AI tasks with the hidden metric each prompt rewards, not what you think you're asking for

  2. Find Your Alignment Gaps

    Test your prompts against your stated values to identify where optimization targets conflict

    Done when

    You have documented misalignments between what your prompts reward and what you actually value, with specific examples of each conflict

  3. Surface What You Don't Know

    Redesign your most problematic prompts to expose uncertainty rather than project false confidence

    Done when

    You have 3+ redesigned prompts that surface tradeoffs and uncertainty instead of optimizing for single metrics, plus evidence they change AI output quality

    Capstone: you put together your Prompt Values Audit.

About this lesson

What you'll be able to do

  • Identify hidden optimization targets in your actual AI prompts
  • Distinguish value-aligned from value-misaligned prompt patterns in your workflow
  • Redesign 3+ prompts to surface uncertainty rather than false confidence

Who it's for

  • Knowledge workers who use AI regularly but haven't examined what their prompts actually optimize for
  • People who sense their AI habits might be steering them away from their values
  • Professionals making decisions with AI support who want to audit their cognitive surrender patterns

What it won't cover

  • How to write better prompts from scratch
  • AI safety theory or technical alignment research
  • Organization-wide AI governance or policy design

Skills you'll practise

  • Prompt audit methodology
  • Value-alignment detection
  • Uncertainty surface design
  • Cognitive surrender pattern recognition