You've taken hundreds of test questions in your life. Every single one of them was written by someone, checked for fairness by someone, and eventually scored by some system — and almost nobody ever explains how any of that actually happens. Today, I sit down with someone who's been inside that machine for over 30 years.
Mr. Simmy Ziv-El built one of the very first classroom tests that could grade itself, back in 1991 — decades before anyone imagined a computer could read an essay. He went on to work inside some of the biggest names in testing, including ETS and ACT, helped oversee the scoring of a real worldwide test of creative thinking, and — years before ChatGPT made "AI" a household word — started building the AI systems now used to actually write test questions. He was also a student himself, in Johannesburg, South Africa, back in 1976, which gives him a firsthand, personal window into a completely different country's education system.
In this episode: what actually makes a test question "fair," how a computer learns to grade something as human as an essay, what it's like watching AI go from science fiction to something quietly writing your homework questions, and what one person's path through nearly every era of testing — paper-and-pencil to artificial intelligence — can teach the rest of us about how much work goes on behind a single test score.
We also get a sneak peak into the Soweto Protests In South Africa, arguably one of the most iconic and important student led protests of history, from someone who stood on the streets himself.
Podden och tillhörande omslagsbild på den här sidan tillhör
Samarth Chitta. Innehållet i podden är skapat av Samarth Chitta och inte av,
eller tillsammans med, Poddtoppen.
