The Origins of Common Sense in Humans and Machines
- Kevin Smith, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States
- Eliza Kosoy, UC Berkeley , Berkeley , California, United States
- Alison Gopnik, Department of Psychology, University of California at Berkeley, Berkeley, California, United States
- Deepak Pathak, Machine Learning and Robotics, Carnegie Mellon University, Pittsburgh, Pennsylvania, United States
- Alan Fern, EECS, Oregon State University, Corvallis, Oregon, United States
- Josh Tenenbaum, Brain and Cognitive Sciences, MIT, Cambridge, Massachusetts, United States
- Tomer D. Ullman, Psychology, Harvard University, Cambridge, Massachusetts, United States
AbstractPeople impose structure on raw percepts, filling the world with objects, agents, events, and properties. This reasoning develops early: by their first birthday, infants can determine features of objects such as number and motion (Spelke, 1990; Wynn, 1992), construe agents’ actions as goal-directed (Woodward, 1998; Gergely & Csibra, 2003), and distinguish helpers from hinderers (Hamlin, Wynn, & Bloom, 2007). Agents, objects, events, properties – these are the building blocks of meaning and common sense that allow even young children to rapidly understand and interact with novel scenes.