Take a technique validated on machines, forced bisociation then keep the far-and-good, and run the same protocol on people. Measure whether human ideation moves further from the average while staying useful.
Models let us test cognitive techniques at the scale of thousands, cheaply, in hours. The winners are carried into a human protocol and tested with rated human ideation.
ML labs will not do this; they do not care about humans. Cognitive science cannot, not at this scale. The intersection is the Institute’s ground: the mirror gives you back a better way to think.