I remember the first time I tried to build a model to predict how people behave in high-stakes decisions… when to make a major financial move, or step into a life-changing career choice. The data was extensive, the algorithms were reliable, and the compute was more than capable. And yet, there was always a gap…