Skip to main content

AI learning has a curriculum; human intelligence does not

What human intelligence should actually teach, why judgement erodes alongside capable AI, and the behaviors that keep AI use aligned with an organization’s values

Most organizations can describe their AI learning in detail; prompting, tool selection, data handling, and what not to paste into a public model. Ask what their human intelligence learning contains, and the answer is often a list of qualities rather than a program.

Human intelligence does not develop automatically. Research shows that confidence in AI can lower scrutiny, experience does not reliably protect against bad outputs, and underlying skills can decay with repeated exposure. The answer is a practical curriculum of behaviors people can learn, practice, observe, and reinforce: verification, ethical judgement under pressure, decision-surfacing, drawing insights from data, deliberate delegation, ownership, and translation.

Get the white paper

What this white paper reveals

  1. Human intelligence is a set of behaviors, not a list of virtues. Verification, ethical judgment, decision-surfacing, deliberate delegation, ownership, translation, and drawing insights from data can all be taught, practiced, and observed.
  2. AI adoption can weaken judgment if people become too confident in the system, defer to its output, or stop exercising the underlying skill. In one Wharton study, people followed a chatbot’s answer roughly 80% of the time when it was wrong across nearly 10,000 reasoning trials.
  3. Experience is not a reliable defense. A 2023 radiology study found that experienced radiologists performed at 82% accuracy when an AI suggestion was correct and 45% when it was wrong, showing how people can be pulled toward the machine’s error.
  4. Skill decay is measurable. Among experienced endoscopists, unassisted detection fell from 28.4% to 22.4% after months of working with AI detection tools. In knowledge work, developers using AI assistance scored 50% on a comprehension test versus 67% for those working unaided.
  5. More AI learning alone will not solve the problem. AI literacy teaches people how to operate the tool; human intelligence learning determines what they do with the output, especially when the work is ambiguous, consequential, or under pressure.

AI fluency is not the full solution

The organizations best prepared for the AI era will not simply train people to use new tools. They will build the habits that keep judgment in the work: checking outputs, challenging assumptions, surfacing decisions, choosing deliberately what to delegate, and owning the outcome. They will also design workflows and leadership practices that make those behaviors possible and measurable.
Download Human Intelligence in the AI Era