The philosopher of science Thomas Kuhn asks: what must nature be like in order for science to be possible at all? As a cognitive scientist, my research asks a parallel question: what must we humans assume the world be like for us to comprehend it?
Both questions stem from this challenge: the impossibility of determining what is in the world from data alone, regardless of how much data we have. Our minds avoid potential cognitive paralysis so well that we find the problem counterintuitive. Where is the problem? How can my compelling perception of the 3-D world be a mere inference?
Our minds meet the challenge by adopting cognitive constraints — basic assumptions that narrow the infinite space of possibilities. Perceptual illusions provide a glimpse of the challenge. When we show our undergraduates the Necker cube — a 2-D image whose front and back squares are identical — we take the flipping between two cube percepts (as if viewed from above or below) from a single retinal image as evidence that perception involves inference. However, given the geometry of light projection, infinitely many 3-D shapes can project the same retinal image. The wonder, then, is why we perceive only two cubes. A still deeper wonder is why we perceive cubes at all. Why doesn’t the (equal-sized) “back” square loom larger than the front with each flip, to obey 3-D perspective? Our percept defies not only an individual’s lifetime of visual experience, but also the visual experience of the species — and of any intelligent system.
The rubber-hand illusion illustrates an analogous defiance of experience-based expectations. Even octopuses experience the illusion. When a hidden tentacle and a visible gel tentacle are stroked synchronously for a few seconds, the octopus changes color and flees when the fake tentacle is pinched. Why is a brief intervention capable of changing the perception of something so familiar and intimate as the boundary of one’s body?
For both illusions, one explanation is that our minds, to avoid cognitive paralysis, bet on the simplest coherent explanation.
Working hand-in-hand with the coherence constraint, the causal invariance assumption holds that cause-and-effect relations are (ideally) stable across contexts. Our induction process begins with the simplest causal model consistent with the data, and revises it only when predictions fail. Because only causal models support predictions about actions, even preschoolers rationally revise representations in response to violations of causal, rather than merely associative, expectations.
My work has mostly focused on the causal-invariance constraint. Recently, by leveraging both basic constraints, my lab has developed brief (15- to 35-minute) science-education interventions that effectively promote climate and biodiversity-conservation actions. The biodiversity materials aim to enable the construction of a coherent narrative of one’s place in the deeply interconnected web of life, fostering a capacious self that encompasses other living things. Study participants, including self-identified conservatives, encountered prediction errors about puzzling yet indisputable phenomena that invited the construction of their own coherent explanations. Compared to control participants, treatment participants reported taking more climate and biodiversity actions when probed a year or even two years later.
In the infinite space of possible representations, just as an octopus’s sense of its bodily boundary can be expanded by a brief intervention, so can humans’ sense of their moral self.
I was born to parents from the same village in rural China who married in their mid-teens in an arranged marriage. The sea change between my conception of the world and theirs tells the story of how malleable minds are. Bearing witness to that transformation likely animated my study of cognitive constraints and their harnessing to make the world a better place. After receiving a B.A. from Barnard College and a Ph.D. from the University of Michigan, both in cognitive psychology, I completed a postdoctoral fellowship in computer science at Carnegie Mellon University. Outside of work, I enjoy concerts and museums, hiking, swimming, and sharing good food with family and friends.
To learn more about my lab’s work, visit: https://www.psych.ucla.edu/faculty-page/pcheng/
Category: Spotlight