Barbara Tversky (In Her Own Voice)

When I consider the revolution in human-computer interaction that occurred in the winter of 2025 — when machines learned to accept natural language and builders suddenly found themselves free to describe a system as they naturally conceived it — I find that my experimental work has been, without planning it this way, a decades-long preparation for precisely this moment. The research program I have pursued since the 1970s rests on a foundational claim that most cognitive scientists now accept and most technology designers continue to underestimate: thought is spatial before it is linguistic. The spatial model precedes the words. When the machines finally learned to accept the words without demanding the spatial reorganization of the model behind them, they eliminated the most expensive single cognitive overhead in the history of human-computer interaction — the work of representational translation. But they also eliminated a form of discipline that was doing cognitive work the builders did not realize