Learning a metacognition for object perception
This work addresses the problem of identifying unreliable percepts for AI systems, offering an incremental step towards more robust perception by incorporating self-awareness.
This paper introduces MetaGen, an unsupervised model that learns metacognition for object perception by modeling how a perceptual system generates noisy percepts. MetaGen jointly infers world objects and its own perceptual system, leading to improved accuracy on simulated data compared to models without metacognition.
Beyond representing the external world, humans also represent their own cognitive processes. In the context of perception, this metacognition helps us identify unreliable percepts, such as when we recognize that we are seeing an illusion. Here we propose MetaGen, a model for the unsupervised learning of metacognition. In MetaGen, metacognition is expressed as a generative model of how a perceptual system produces noisy percepts. Using basic principles of how the world works (such as object permanence, part of infants' core knowledge), MetaGen jointly infers the objects in the world causing the percepts and a representation of its own perceptual system. MetaGen can then use this metacognition to infer which objects are actually present in the world. On simulated data, we find that MetaGen quickly learns a metacognition and improves overall accuracy, outperforming models that lack a metacognition.