Limitations of Human and Artificial Cognition

One of the most remarkable features of human cognition is the ability to rapidly adapt behavior in a changing world. Mechanisms underlying this function are summarized under the term cognitive control. They are engaged across various domains of cognition, including perception, attention, learning, and memory, and appear to be fundamental to many of the faculties that distinguish human mental function from that of other species (and continue to distinguish it from machines), including problem-solving, planning, and language processing. Yet, humans are strikingly limited in how many control-demanding tasks they can perform simultaneously (e.g., reading a document while listening to a friend) or how intensely they can focus on a single task (e.g., parsing an equation in a noisy environment). Our research program focuses on understanding limitations in the capacity of the human mind to exert mental effort and the consequences of these limitations for natural and artificial cognition. To examine these limitations, we apply a combination of neurocomputational modeling, behavioral experimentation, and closed-loop machine learning. 

Relevant Work

Musslick, S., & Cohen, J. D. (2021). Rationalizing constraints on the capacity for cognitive control. Trends in Cognitive Sciences, 25(9), 757–775. doi: https://doi.org/10.1016/j.tics.2021.06.001

Petri*, G., Musslick*, S., Dey, B., Öczimder, K., Turner, D., Ahmed, N., . . . Cohen, J. D. (2021). Topological limits to parallel processing capability of network architectures. Nature Physics, 17(5), 646–651. doi: 10.1038/s41567-021-01170-x

Musslick, S., Saxe, A., Özcimder, K., Dey, B., Henselman, G., & Cohen, J. D. (2017). Multitasking capability versus learning efficiency in neural network architectures. In Proceedings of the 39th Annual Meeting of the Cognitive Science Society (pp. 829–834). London, UK.

Autonomous Empirical Research

The integration of empirical phenomena into mechanistic models of human cognition and brain function is a fundamental staple of cognitive neuroscience. Yet, researchers are beginning to accumulate increasing amounts of data without having the time or monetary resources to integrate these data into scientific theories and/or to test the resulting theories in follow-up experiments. Our research group strives to enhance and accelerate scientific discovery by automating each step of the empirical research process, from constructing a scientific hypothesis to conducting novel experiments. To this end, we will devise an open-source programming language for autonomous empirical research. We seek to co-create and share this system with empirical scientists across disciplines who seek to advance their research without sacrificing scientific standards, such as reproducibility and transparency. We also seek to lead the forefront of open science by developing methods for the automated documentation and dissemination of steps taken in the empirical research process.

Relevant Work

Musslick, S., Cherkaev, A., Draut, B., Butt, A., Srikumar, V., Flatt, M., & Cohen, J. D. (2021). Sweetpea: A standard language for factorial experimental design. Behavior Research Methods. doi: https://doi.org/10.3758/s13428-021-01598-2

Musslick, S. (2021c). Recovering quantitative models of human information processing with differentiable architecture search. In Proceedings of the 43rd Annual Meeting of the Cognitive Science Society (pp. 348–354). Vienna, AT.

Department of Cognitive, Linguistic, and Psychological Sciences

Carney Institute for Brain Science