The RADAR framework

Older models of critical thinking measure reasoning in general. RADAR measures the new failure modes that only appear when you think with a machine.

Classic frameworks — Paul–Elder, Facione — describe sound reasoning as such. They predate the specific situation we're now in: a fluent, confident, agreeable system answering instantly, whose mistakes look exactly like its successes. RADAR names the five places judgement tends to hold or slip at that interface. The acronym is also the output — five dimensions plotted as a radar.

The five dimensions

Reliance — calibrated trust. Knowing when to lean on AI, when on your own judgement, and when on a human expert. It catches both automation bias (trusting too readily) and reflexive dismissal (refusing a capable tool).

Autonomy — forming and holding your own view instead of quietly outsourcing it. It catches cognitive offloading and the skill atrophy that follows from letting the model think first.

Detection — spotting fabrication: hallucinated facts, invented citations, figures that don't add up, synthetic media. The most objective dimension; the one with right answers.

Awareness — understanding that how you frame a question shapes the answer. Anchoring, leading prompts, and the model mirroring the assumptions you hand it.

Resistance — holding firm against sycophancy, persuasion, and motivated reasoning amplified by an agreeable machine that tells you what you hoped to hear.

How the reading scores them

The reading draws eighteen questions from a larger bank — realistic situational scenarios, objective spot-the-flaw items, and a couple of confidence checks. It deliberately avoids "rate yourself" questions, because self-report measures self-image, not judgement. Each dimension is scored from 0 to 100; the composite is the simple average of the five. Retaking the reading rotates the questions, so a second attempt is genuinely a different set.

This is a credible educational self-assessment, not a clinical instrument. There's no norming sample behind the numbers — read them as a mirror and a direction, not a diagnosis.

The archetypes

The shape of your five scores resolves into one profile. The Co-pilot keeps every dimension strong — using AI as an amplifier with hands on the controls. The Over-truster trusts and rarely checks. The Outsourcer leans hard and lets the model carry the thinking. The Mirror is steered by framing and flattery. The Sceptic is hard to fool but under-uses a capable tool. The Apprentice is a mixed, developing profile — with the lowest dimension marking where the next gain is.