Can You Trust This Answer?

According to the passage, how does an AI language model generate its responses?

• It searches the internet in real time to find the most current information

• It looks up facts from a built-in database of verified knowledge

• It predicts the most statistically likely response based on patterns in its training data

• It reasons through problems the same way a human expert would

According to the passage, when an AI produces information that sounds plausible but is simply fabricated, researchers call this (a)   .

A student asks an AI to explain a scientific discovery that was announced last month. Based on the passage, which reason best explains why the AI's answer might be inaccurate?

• Hallucination

• Bias

• Training Data Cutoff

• Overconfident Tone

A student uses the Part 3 checklist to evaluate an AI response about climate policy. The AI gives a detailed answer but only cites the position of one government. Which checklist item should the student flag?

• Accuracy

• Recency

• Completeness

• Bias

Look at the real example from Part 4.

Student prompt: "What are the health effects of eating too much sugar, and what does recent research say about artificial sweeteners as an alternative?"

The student found that one of the three studies cited by the AI does not appear to exist. Using the passage's vocabulary, name the specific problem this represents and explain in your own words why it makes AI responses difficult to trust.

The passage states that AI produces text that looks like a confident answer, but that appearance of confidence is built into the (a)   , not into actual knowledge.

The passage describes four reasons AI can get things wrong.

1). Choose two of those reasons and for each one, describe a real school situation where that specific problem could cause harm to a student, and

2) name which checklist item from Part 3 would help catch it.