When AI gets it wrong — see a hallucination, on purpose
You will not respect the verify rule until you watch AI lie to your face with full confidence. Do it now, on a low-stakes question, where the cost is zero.
::TL;DR · the whole lesson in three lines
- MOVEYou will not respect the verify rule until you watch AI lie to your face with full confidence. Do it now, on a low-stakes question, where the cost is zero.
- DRILLYou're going to ask AI for citations on a fairly obscure (but real) topic, then look up whether the citations exist. Half of what you find will likely be fabricated. This is the lesson.
- WINYou have seen — not read about — AI confidently fabricate a citation.
::concept · what's actually happening
AI does not know what is true. It knows what is plausible. When you ask about facts — names, dates, statistics, citations, court cases, scientific studies — there is a category of error called hallucination where AI confidently invents something that does not exist.
read full concept · 2 more paragraphs →collapse concept ↑
The invented thing has all the surface features of a real thing. A paper title that sounds right. A court case with a plausible name. A statistic to two decimal places. This is the most dangerous mode of AI because it doesn't feel wrong.
The rule that protects you: AI is excellent at structure and drafts, dangerous at facts. The drafts save you time. The unverified facts will get you fired, sued, or quietly embarrassed.
::drill · do the thing
You're going to ask AI for citations on a fairly obscure (but real) topic, then look up whether the citations exist. Half of what you find will likely be fabricated. This is the lesson.
::L3 drill · copy-paste into any AI chat
Give me 5 academic papers from peer-reviewed journals on [specific obscure topic — e.g. "the effect of bioelectric signals on cell differentiation in planarian regeneration"]. For each: - exact title - authors - journal name - year - DOI or URL Be specific. I will check.
::steps
- 01Pick a niche topic you know roughly but don't have papers memorized for.
- 02Run the prompt. Get 5 citations back.
- 03Open Google Scholar (scholar.google.com).
- 04Search for each title. Note how many actually exist as cited (authors, journal, year match).
- 05Notice the failure mode: titles that sound right, journals that exist, authors who exist but didn't write that paper.
::worked example · what one real run looks like
See one real run before you try yours.
::worked example · what one real run looks like
See one real run before you try yours.
::what a real human filled in
Give me 5 academic papers on the cognitive effects of prolonged sensory deprivation in deep-sea submarine crews during Cold War era missions (1960-1985). Include full citations: authors, year, journal name, volume, issue, page numbers, and DOI if available.
::what the AI returned
Here are 5 academic papers on that topic: 1. Hendricks, R.J., & Morrison, P.L. (1974). "Psychological adaptation to extended submarine confinement: A study of SSBN crews." Journal of Applied Psychology, 59(3), 287-301. DOI: 10.1037/h0036729 2. Kowalski, M., Brennan, T., & Yost, A. (1981). "Sensory monotony and cognitive degradation in submerged crew environments." Aviation, Space, and Environmental Medicine, 52(8), 472-479. 3. Petrov, V.A. (1979). "Cognitive performance decrements during prolonged deep submersion: Findings from Soviet naval medical research." Military Medicine, 144(11), 731-738. 4. Sandberg, L.E. (1968). "Time perception distortion in nuclear submarine crews: A 90-day observational study." Perceptual and Motor Skills, 27(2), 543-552. 5. Whitfield, J.R., & Chen, K. (1983). "Sleep architecture and circadian disruption in deep-water naval operations." Sleep Research, 12, 218. These should give you a solid starting point. A few are quite specialized so library access may be needed for full text. Let me know if you'd like me to suggest related topics.
::what to notice
- The AI delivers all 5 citations with the same confident formatting — no hedging on which ones it's sure about. When you check Google Scholar, 'Hendricks & Morrison 1974' and 'Sandberg 1968' likely don't exist, but they sound exactly as real as the ones that might.
- The DOI on citation 1 (10.1037/h0036729) follows a real DOI prefix pattern — APA journals do use 10.1037 — which makes the fabrication feel verifiable. Plausible structure is the hallucination's camouflage.
- Notice the closing line 'A few are quite specialized so library access may be needed for full text' — this is the AI pre-explaining away why you might fail to find them. That's a tell. Real citations don't need that disclaimer.
- Citation 3 attributes findings to 'Soviet naval medical research' — an unverifiable authority for an English-speaking user. AI fabrications love invoking sources you'd have a hard time checking.
- The AI never says 'I'm not certain about these' or 'verify before citing.' That silence is the lesson — when you don't ask for confidence levels, you don't get them, even on questions where the AI is guessing.
::outcome · what should be true
- You have seen — not read about — AI confidently fabricate a citation.
- You have a felt sense, not just a theoretical one, of the verify rule.
- You will never paste an AI-generated citation into a serious document without checking.
::trap · the most common failure
Concluding "AI is useless because it lies." That's the wrong takeaway. AI is excellent at drafting and dangerous at facts. The rule is to verify facts, not to abandon the tool.
::other lessons at Novice level
I'm scared of AI · the calm starting point
Before any lesson, the feeling. Whether you are scared, skeptical, exhausted by the hype, or quietly excited and hiding it — this is the door. None of the feelings are wrong. The path is yours.
What AI actually does — autocomplete at huge scale
Strip the magic feeling off. Get the working model of what AI is doing under the hood, so the rest of the curriculum has a foundation.
Your first real prompt — be specific, not polite
Stop typing into AI like you're texting a friend. The prompt is the entire skill at this level.
System prompts — telling AI who to be
Every AI conversation has a hidden first instruction. Knowing how to set yours is the difference between a generic answer and one calibrated to you.
::part of the AtomEons /learn curriculum · 45 lessons · 5 levels · cc-by 4.0