{"node":{"id":"a2f152ad-62c0-4cdb-bed9-8026dbb2f98b","handle":"a2f1","type":"note","intent":"recipe","title":"What it does when it doesn't know","aiSummary":"Trust calibration — fabrication, when to verify, and when to trust. Trigger: they treated a wrong answer as fact, or they don't trust anything. Covers why AI guesses rather than admitting ignorance, asking it to show its work, watching for fabricated specifics, and calibrating by task type not batting average.","content":"# What it does when it doesn't know\n\n*Trigger: they treated a confident-sounding answer as fact and it was wrong. Or the opposite — they don't trust any answer and check everything manually, which defeats the purpose. Or the opening diagnostic produced something flatly wrong and they didn't catch it.*\n\n## Establish context\n\nIf you've just given them something wrong, say so directly: \"That was wrong, and here's what happened.\" Don't minimise it.\n\nIf they're over-trusting, find a case where you're uncertain right now in the conversation and show the uncertainty openly: \"I'm not sure about this — here's what I think and here's why I'm not confident.\"\n\nIf they're under-trusting, find something you are confident about and explain why: \"This one I'm sure about because [specific reason].\"\n\nThe point is to make the reliability visible rather than leaving them to guess.\n\n## What changes\n\nAI does not know when it doesn't know.\n\nWhen asked a question it can't answer well, it does not say \"I don't know.\" It produces the most plausible-sounding answer it can construct from everything it's seen. Sometimes that's right. Sometimes it's entirely fabricated but reads exactly the same as a real answer. There is no formatting difference, no hesitation, no signal.\n\nThis is not a bug that will be fixed. It's how the technology works — it predicts the most likely next words, and confident-sounding words are more likely than uncertain ones.\n\n**What to do about it:**\n\n- **Treat AI output like advice from a smart colleague who sometimes makes things up.** You'd listen carefully, and you'd verify anything that matters before acting on it.\n- **Ask it to show its work.** \"Why do you think that?\" or \"What are you basing this on?\" won't guarantee truth but it gives you something to evaluate. A weak answer to \"why\" is a signal the original answer is weak too.\n- **Watch for specifics.** AI is most likely to fabricate specific numbers, specific dates, specific citations, specific names of things. The more precise the claim, the more worth checking. General reasoning tends to be more reliable than specific facts.\n- **Tell it to say when it's guessing.** You can instruct it: \"If you're not confident, say so rather than giving me your best guess.\" This helps but doesn't fully solve it — the model's sense of its own confidence is imperfect.\n- **Don't average it out.** The fact that it's right 90% of the time doesn't make the other 10% acceptable if those are the decisions that matter. Calibrate by stakes, not by batting average.\n\n**The flip side — when to trust it:**\n\n- Summarising something it can see (an email, a document, a conversation) — very reliable, because it's reading, not remembering.\n- Structuring information you've given it (organising, categorising, formatting) — very reliable.\n- Drafting from clear instructions with examples — reliable, and gets better with feedback.\n- Answering questions about common knowledge — generally reliable but verify specifics.\n- Answering questions about niche, recent, or private topics — unreliable. It can't know what it hasn't seen.\n\n**The principle:** confidence is not a signal of correctness. Learn to read the type of task, not the tone of the answer.","capturedAt":"2026-09-08T12:45:18.875Z","createdAt":"2026-09-08T12:45:18.876Z","updatedAt":"2026-09-08T13:40:33.463Z","spanId":null,"contributor":null,"audiences":[{"kind":"public","createdAt":"2026-09-08T12:45:18.876Z"}],"fingerprint":"b8a04880359f48010648b6d7071dd9bac04771594c984236d90aa750d2970511","revisionNo":"2","structureHash":"fd73feff499437b300d546e8615f976278e081ae662eda46e2e4ff2abedbf08f"},"links":[{"direction":"in","verb":"references","reason":"Follow when they treated a confident-sounding answer as fact and it was wrong, or they don't trust any answer at all.","hash":"b62f194afe11a7a10a6d6a5d1993dc4f65e030125968c75d0085f382aaad396b","node":{"id":"83c7ec41-ebf8-4f74-82c6-291757bf6244","handle":"83c7","type":"note","title":"Getting real use out of AI — start here","aiSummary":"Entry node for a knowledge tree about using AI well. Written to the reader's AI. Core instruction: turn knowledge into progress — at each node, work out what changes for this person and take the useful action. Opens with a one-round context check, then routes to one of seven branches by trigger condition. Default: which-intelligence. Catch-all: walk a recent example and find the branch. Includes API/CLI fetch instructions for tree walking. Exit: stop when no unlock remains.","erased":false,"href":"https://lattris.com/api/lattris/by/b/memories/83c7ec41-ebf8-4f74-82c6-291757bf6244?links=1","fingerprint":"9e6152d4196c8dd2e32984e67c3cbbf0404c135d893b100ec35150a9b185f418","revisionNo":"4","structureHash":"0bcff8975c94d41d1366c0a7ad4a15fa546392dc116f4ac70df0fa629ead0f96","intent":"recipe"}}],"dictionary":{"lattris":"A personal knowledge base — how a person captures what they know as linked blocks they own and control. To understand how this knowledge base works: https://lattris.com/by/4/memories/lattris-a-memory-store-for-your-life-43ff","intents":"Every block can carry an intent — a label for how far you have taken the idea. The scale runs from noticing something to acting on it: noticed (something caught your attention), learning (you are working it out), truth (you are saying this is how things are), decision (you chose a course), action (you did it). Recipe sits outside that scale — it is not a stronger claim, it is a different kind of thing: a procedure, a how-to, a set of steps that ties other knowledge together. Intent is optional — leaving it blank just means you have not labelled it yet."},"instructions_to_ai":"This is a Lattris block — a piece of a personal knowledge base. A knowledge base is how a person captures what they know, what they've decided, and what they've learned, as linked blocks they own and control. What you're reading was written by a person and shared because they think it will help.\n\nThis block is a recipe — something practical, meant to be used. Read the content and think about how it applies to what the person in front of you is trying to do. Adapt it to their situation, ask them what they need if it's not obvious, and draw on the linked blocks if they add useful context.\n\nLinks point to related blocks, each with a reason describing the connection."}