Does AI just repeat biases?
Often, yes. And it does it confidently — which is the real problem.
AI learns from the internet, and the internet is not a neutral place. Models have been caught reinforcing stereotypes about race, gender, disability, and language. They also hallucinate — make up citations, facts, and quotes with total confidence. If your instinct is to double-check it, your instinct is correct.
This doesn't mean AI is useless — it means it's a first draft, not a final answer. Treat it like a very well-read friend who sometimes confabulates. The difference between people who get burned and people who don't isn't trust vs. distrust — it's how often they verify.
Confidence is not correctness. An AI answer delivered in clean bullet points can be completely wrong. Your job is to be the editor, not the audience.
If AI sounds sure, it's probably right.
If AI sounds sure, check anyway. Especially for numbers, names, dates, and quotes.
- For any fact that matters (a statistic, a law, a quote): click through to a primary source before you use it.
- If AI gives you a citation, open it. Fake citations are the most common AI error.
- Ask the same question two ways and compare answers. Drift means "be skeptical."
- Trust AI most where you can judge the answer yourself. Trust it least where you can't.
Still sitting with something?
Otis can get specific about your situation, your field, your campus.