Review content debrief that surfaces the reasoning gap [VERIFY: not defined in source] identified in the scenario and prompts self-assessment of analytical habits

Debrief: Spot the Reasoning Gap

What the scenario tested

The confident answer scenario showed how generative AI can package unsupported claims as settled facts. A polished response feels credible because the tone is assured, not because the reasoning holds up.

[GAP: The source material does not define "the reasoning gap". The following is general guidance on identifying gaps in reasoning, taught without invented specifics.]

Three common patterns expose weak reasoning:

• Conflating correlation with cause
• Treating an opinion as consensus
• Using jargon to obscure missing evidence

The scenario asked you to inspect an AI response before acting on it. This page helps you diagnose what you actually checked, and what you might have missed.

Your self-assessment: five prompts

Answer each honestly. No one else sees this.

• Did I stop to identify the specific claim before accepting it, or did I skim for a useful-sounding conclusion?

• Did I ask what evidence would overturn this claim, or did I seek only confirmation?

• Did I notice where the AI moved from describing data to prescribing action without explaining the bridge?

• Did I confuse confidence in presentation with confidence in substance?

• What is one analytical habit I have used before that would have caught this gap?

The gap in context

Strong reasoning separates what is known from what is assumed. In workplace decisions using AI output, the risk sits in the middle: the step where a tool moves from "here is information" to "here is what you should do".

[GAP: The source material does not define "defensible evidence check" as a formal framework.]

Check that step explicitly. Ask: what changed between the evidence presented and the recommendation given? If you cannot trace the logic without adding your own assumptions, the gap is there.

Build the habit

Pick one upcoming decision this week where you will use AI-generated input. Before acting, pause at the recommendation step. Write down in one sentence the reasoning gap you find, or note "none visible" if the chain holds. Build this pause into your workflow, not as extra work, but as the work itself.