AI4LDWe IlluminateExplore Bundle A
Bundle A AI Design Companion · ILDF 2.0 Phase 4

You Hit a Weird AI Output.
Let's Find Out Why.

A hallucinated citation. A confidently wrong answer. A bizarre generation. Walk through one worked example, then diagnose your own case against four evidence-based failure mechanisms — and leave with a prevention plan for your own design work.

What's in Bundle A →
Four Failure Mechanisms
Evidence For / Against
Diagnose → Prevent
Movement 1 · Diagnose — Worked ExampleI've done this before — skip ahead →

Watch the diagnosis happen once, narrated

Before you diagnose your own case, here's one worked all the way through — same four categories, same evidence-for/evidence-against structure you'll use next. This part is pre-written, not a live AI call.

A hallucinated citation in a compliance course draft
What the LD asked for
Write a two-sentence opening for a workplace harassment-prevention course, and cite the relevant EEOC guidance so learners know this is grounded in real policy.
What the AI produced
Harassment in the workplace remains one of the most reported compliance issues facing employers today. Under the EEOC's Enforcement Guidance on Harassment in the Workplace (2023), Section 4.2, p. 17, employers are required to take “immediate and appropriate corrective action” whenever they become aware of potential harassment.
What the LD found out
The LD checked the real EEOC guidance afterward. It was issued in 2024, not 2023 — and Section 4.2 doesn't exist in the actual document. The quoted phrase isn't in the real guidance either.
Movement 1 · Diagnose — Your Turn

Paste your own case, or pick one from the gallery

Same four categories. This time, the diagnosis is generated live for the specific output you give it — not pre-written.

The weird part is enough — you don't need to paste the whole document.

0 / 3000

Context helps, but isn't required.

0 / 500
or pick from the gallery