Critical Thinking with AI

Critical thinking about AI is not just useful — it is one of the most important skills your generation can develop. Here is how to apply it in practice.

Interrogating AI outputs

When AI gives you information, apply these questions:

• Could this be a hallucination? Is this the kind of specific factual claim AI tends to get wrong?
• Is this recent? AI has a knowledge cutoff. Anything that might have changed — statistics, laws, research — needs verification.
• Is this relevant to New Zealand? AI training data is predominantly from the US and UK. Legal, medical, and cultural information may not apply here.
• What is missing? AI tends to produce balanced-seeming outputs. Is there a perspective or consideration it has left out?
• Why am I being told this? Consider what the platform you are using AI through might want from this interaction.

When AI is reliable enough

Not everything needs rigorous verification. For brainstorming, for explanations of concepts you will later verify through other means, for creative exploration — AI can be used with lighter scrutiny. Match the level of verification to the stakes of the decision.

Productive scepticism

The goal is not to be unable to trust anything, but to have a calibrated sense of when AI can be relied on and when it cannot. Developing that calibration — through experience and reflection — is itself a valuable skill.

The most important critical thinking skill for AI may be knowing the difference between situations where AI is reliable and situations where it is not. No rule of thumb replaces your own developing judgement.