Overview
This series asks whether a learner’s confidence with AI can stand in for AI literacy. It argues that confidence is a readiness signal, not proof of understanding or judgement. The framework defines literacy through knowledge, human skills, responsible attitudes, and agency across engaging, creating, managing, and shaping AI. One academic year of ChatGPT logs from 36 undergraduates at one U.S. university shows why verification, repair, metacognitive checking, trust calibration, and reasoned restraint add evidence that self-report lacks. A practical measure should place confidence beside task-based evidence, while recognising that the framework is normative and the study is not population-wide.
Learning outcomes
- Confidence can signal readiness to try AI, but cannot by itself demonstrate AI literacy.
- AI literacy combines knowledge, human skills, responsible attitudes, and agency across four connected domains.
- Explaining limits, choosing AI’s role, verifying outputs, repairing failures, and reflecting on consequences provide stronger evidence.
- The framework is normative, and the ChatGPT study of 36 undergraduates at one U.S. university is not population-wide evidence.
Keywords
Preview every chapter · Unlock full text, audio, and video from chapter 4
PlansFAQ
Can confidence measure AI literacy?
No. Confidence may signal readiness to try AI, but it cannot show whether a learner understands the system, verifies outputs, repairs failures, or takes responsibility.
What does AI literacy include?
AI literacy includes knowledge, human skills, responsible attitudes, and the ability to engage with, create with, manage, and shape AI in context.
What evidence shows AI literacy better than confidence?
Evidence comes from what learners explain and do: assess system limits, choose AI’s role, verify claims, repair or abandon weak workflows, monitor thinking, and consider consequences.
What are the limits of current evidence about confidence and AI literacy?
The sources do not establish a universal statistical link. The framework is non-binding, while the empirical study examined 36 undergraduates at one U.S. university using ChatGPT and did not measure confidence.