A Taxonomy for Curiosity in Humans and AI

From Day 2 (July 14th, 2021) of the mediaX Conference, Building Reciprocity: Curiosity and Learning in Humans and Machines, Nick Haber, Assistant Professor, Stanford Graduate School of Education discusses his research around...

Human learning is interactive learning. We learn by interacting with our environments and the people within them, and we have evolved to be very good at interacting in order to learn. By embracing how humans learn through interaction, and in particular how they learn through social interaction, we can design AI that learns with and from us — and teaches us — more naturally.

In engineering AI that learns through interaction, one important component to design is curiosity, or intrinsic motivation: how we decide what to attend to and interact with in order to learn. We might be motivated by novelty, or difficulty, or some measure of expected increase in understanding, or something else entirely.

Here we present a taxonomy for curiosity as well as a study that compares human behavior to AI implemented with various types of curiosity. Through work like this, AI can help us better understand how we learn as well as how different people learn differently.

Видео A Taxonomy for Curiosity in Humans and AI канала mediaXstanford