Celeste Kidd

Celeste Kidd is an Assistant Professor of Psychology at UC Berkeley who investigates learning and belief formation. The Kidd Lab is one of few in the world that combines technologically sophisticated behavioral experiments with computational models in order to understand knowledge acquisition. They employ eye-tracking and touchscreen testing with human infants among other methods to show how learners sample information from their environment and build knowledge over time. She has published in PNAS, Neuron, Psychological Science, Developmental Science, and elsewhere. Her work on ethics in AI and tech has been featured in keynotes at NeurIPS and the World AI Summit. Her lab has received funding from NSF, DARPA, Google, the Jacobs Foundation, the Walton Family Foundation, the Human Frontiers Science Program, the Templeton Foundation, and the Hellman Foundation.

She received the American Psychological Science Rising Star designation, a 2021 Janet Taylor Spence Award for transformative early career contributions to psychology, the Glushko Dissertation Prize, and the Cognitive Science Society Computational Modeling Prize. She is an Association for Psychological Science Fellow. Kidd was named one of TIME Magazines 2017 Persons of the Year as one of the “Silence Breakers” for her advocacy for greater equality, diversity, and inclusion in higher education.

Research focus
Celeste Kidd studies how children navigate the sea of information that surrounds them to learn. She uses a blend of computational models and behavioral experiments to study the core cognitive systems that people use to guide their learning—including attention, curiosity, and metacognition (thinking about thinking). Her work shows even infants select information based on its expected utility and become certain with feedback. She also studies why people sometimes hold unjustified beliefs.

What have I achieved during my fellowship?
The most important achievement of the fellowship period has been the realization that there is no distinction between knowledge acquisition and belief formation. Everything anyone knows represents their best guess based on a very limited quantity of data they have sampled from the world. “Knowledge” is not as stable as we thought. Human concepts are far more flexible and adaptive and change across time and context. We have a klugey system that prevents us from hyper focusing our learning on just one thing by inducing boredom when heuristics suggest we understand something approximately well enough. It works pretty well most of the time in the naturalistic world, but it leads to systematic errors in certain contexts. For example, we become very confident that a new idea is correct if we get a few pieces of confirmatory feedback, and once we reach that threshold level of certainty, we stop sampling and checking to see if we should remain confident. We have published a lot on this topic (in PNAS, Cognition, Open Mind, Psychonomic Bulletin & Review, PLOS ONE, Cognitive Science, the American Journal of Human Biology, and elsewhere), secured funding to continue this work (from Templeton, DARPA, NSF, and the Berkeley Center for New Media), and initiated interdisciplinary collaborations with other Jacobs Fellows that have added depth to how we approach these questions.

My plans for the future
Immediate future plans include investigating:
(1) the origins and nature of common-sense reasoning abilities in human infants (DARPA),
(2) the evolution of curiosity via lifespan and cross-cultural studies (Templeton),
(3) why people sometimes hold unjustified beliefs in the face of evidence that they should not (Berkeley Center for New Media), and
(4) the origin and influence of children’s certainty in their learning (Jacobs Foundation).