Research areas

Cognition

Cognition

Since the earliest days of Artificial Intelligence, the study of human cognition has played an important role. The very first working code presented at the Dartmouth Summer workshop in 1956 where the term “artificial intelligence” was first used, was a model of how humans learn, implemented by Herb Simon and Alan Newell. Since then the study of human intelligence (and, more broadly, human cognition) and the study of “machine cognition” or artificial intelligence, have been mutually beneficial.

How people learn continues to be one of the important topics addressed in this thematic area, using more contemporary AI tools such as deep learning algorithms and large language models to model human cognitive abilities. In this context a number of  projects address how infants learn language. Here, unsupervised and self-supervised algorithms model the process of infant language learning in diverse contexts, and using diverse data sources including text, raw audio and audio+video. In turn, these studies of infants can lead to more flexible and data efficient machine learning algorithms. A number of other projects address the learning of pragmatics (the use of language, rather than its structure) by children in middle childhood. Here multimodal deep learning algorithms are given both the language and the nonverbal behaviors (such as eye gaze shifts and facial expressions) of pairs of individuals as input into models of the building of interpersonal rapport, or social bonds. In turn, these models are used to implement embodied conversational agents that can successfully build rapport with their users. The most recent set of studies of rapport-building behavior add additional evidence from the neuroscience technique called hyperscanning, where pairs of individuals are scanned simultaneously while working alone and while collaborating, and evidence is gathered about their inter-brain synchrony in these different contexts. Studies such as these can contribute to building truly collaborative AI systems, and also to better understanding the role of the brain in natural human interaction.