When the Keyboard Lies: Two Decades of Predictive Text Research and the Cognitive Gap It Exposes
Academic inquiry into mobile text entry has accumulated substantial evidence that autocorrect and predictive text systems were engineered around a fundamentally mistaken model of how human beings translate thought into language. Computational linguistics and HCI studies converge on a troubling conclusion: current input mechanisms prioritize throughput metrics at the direct expense of the exploratory, iterative nature of human cognition. Emerging research into probabilistic language models and in