The Origins of Commonsense in Humans and Machines

July 29, 2020 | 42nd Annual Meeting of the Cognitive Science Society

For all their recent successes, artificial systems do not yet perform basic commonsense reasoning at the level of even young children. By understanding the origins of common sense in humans, we hope to understand how to recapitulate it in machines. In turn, by looking at the successes and failures of machines, we can make scientific progress towards understanding the initial state and learning mechanisms of human intelligence.

This workshop serves as a forum for considering theories and approaches for understanding and building common sense, presenting experimental research that probes the foundations of common sense in people, and reporting on progress on building artificial agents with infant-like commonsense reasoning capabilities.

While the workshop is over, videos of each of the speakers are linked below in the Talks section.

View the full workshop description here.

Talks

Speaker	Title
Alan Fern	A model-based reinforcement learning perspective on common sense learning
Stanislas Dehaene	Advances in understanding human geometrical intuition
Joshua Tenenbaum	Tools for reverse engineering common sense
Ori Ossmy	A behavioral approach to the development of common sense
Matthew Botvinick	Learning intuitive physics (almost) from scratch
Moira Dillon	Cognitive Artificial Intelligence: Building better machines… and babies!
Deepak Pathak	Intelligence without a brain
Jessica Sommerville	Developing a moral sense: probing the extents and limits of infants’ socio-moral cognition and behavior
Eliza Kosoy	Curiosity and exploration in children’s maze playing using DeepMind Lab
Tucker Hermans	Can common sense guide autonomous robot exploration?
Kevin Smith	Building models of infants’ physical understanding
Shari Liu	Origins of social intelligence in human infants
Allison Gopnik	Children are MESS-y: Model-building, Exploratory, Social learning systems