A neural network walks into a lab: towards using deep nets as models for human behavior

• DOI: 10.48550/arxiv.2005.02181
• Publication Date: 2020-01-01
• Authors: Wei Ji; Benjamin Peters
• Keywords: Toolbox; Joke

What might sound like the beginning of a joke has become an attractive prospect for many cognitive scientists: use deep neural network models (DNNs) as human behavior in perceptual and tasks. Although DNNs have taken over machine learning, attempts to them are still early stages. Can they versatile model class scientist's toolbox? We first argue why potential be interesting behavior. then discuss how that can more fully realized. On one hand, we cycle training, testing, revising needs revisited through lens goals. Specifically, methods assessing goodness fit between DNN date been impoverished. other science start using complex tasks (including richer stimulus spaces), but doing so beneficial DNN-independent reasons well. Finally, highlight avenues where traditional process may show productive synergy.