糖心TV

"A Model of Associative Judgment"

I present a computational model of associative judgment. The model is trained on the English language Wikipedia corpus, and is able to answer unstructured judgment problems spanning an almost universal domain of knowledge. The model achieves a high accuracy rate across a range of experimenter-generated, participant-generated, and real-world question datasets.

The model also accurately predicts human responses on these datasets. These results suggest that associative judgment provides a powerful account of not only human error, but also human intelligence. In doing so, they illustrate a new approach to constructing and testing models of judgment and decision making.