How Do We Know What People Know?

Source: Mike Kentz Substack
Author: Mike Kentz
Original source: https://mikekentz.substack.com/p/how-do-we-know-what-people-know

Private backup: the full article text is archived in the private repository at archives/articles/mikekentz-substack-com-how-do-we-know-what-people-know.source.md. It is not published on the public Quartz site.

Summary

Mike Kentz argues that AI has severed the long-standing link between producing an artifact and demonstrating understanding. Using examples from the digital SAT, homework, higher education essays, admissions, hiring, and corporate compliance training, he claims many institutions now face the same assessment crisis: they can no longer infer knowledge from polished submitted work. The article rejects detection as the primary solution and instead argues for live, interactive, observable demonstrations of thinking. Its central design shift is from evaluating artifacts to observing cognition: “show me what you produced” becomes “show me how you think.”

Big ideas

Claims

Key evidence and examples

  • The article discusses digital SAT cheating, Bluebook lockdown bypasses, and remote-control exam services.
  • Research examples include AI-written exam submissions going undetected and outperforming some real student work.
  • Princeton and Amherst require anchored writing samples because standalone admissions essays are less trustworthy.
  • Examples of alternative evidence include in-class blue book exams, AI-powered voice interviews about research, early live hiring conversations, and interactive demonstrations.

Education relevance

Very high relevance for assessment redesign, academic integrity, admissions, AI-era writing pedagogy, live demonstrations of learning, authentic assessment, and institutional evaluation.

My notes