Zoom link: https://mit.zoom.us/j/2711902511

***
Speaker: Shengyi Wu

Affiliation: Schulz Lab

Title:  Children track variability in adult attention and plan interventions accordingly

Abstract: Humans are social beings and we care a great deal about the attention of others. Prior research on young children’s sensitivity to others’ attention has overwhelmingly taken a first-person perspective, focusing on how children experience and respond to adults’ attention. Studies show that children are highly responsive to this attention, benefiting from its presence and suffering in its absence. However, as social organisms, children are also exposed to a world of interpersonal interactions where they have plenty of opportunity to learn from the distribution of attention in third-party interactions – even when they themselves are not directly part of the social dynamic. Here I will present three studies showing that older (6–7 years old), but not younger (4–5 years old) children, expect others to value others’ attention, track the factors that covary with adults’ attention, and can plan interventions to hold adult attention. I will also introduce some new work on the impact of adult attentiveness on children’s learning.

***
Speaker: Ben Lipkin

Affiliation: Fedorenko & Levy

Title: Modularity, Systems, and Cognitive Architecture 

Abstract: LLMs have emerged as a dominant paradigm in the design of systems that interact through text. While the capabilities of LLMs in isolation are astounding, some of the most powerful applications have come from their combination with classical symbolic systems. In this talk, I will highlight how leveraging principles from classical cognitive science and AI, particularly modular cognitive architecture, can yield more robust and reliable systems to interact with the world of text. Across a few case studies from my work and others, I will present approaches that intersect language models with interactive theorem provers to yield provably accurate logical reasoning and probabilistic programming languages to yield uncertainty-aware semantic parsers. I will also present some early next steps working on resource-rational dispatch of computation when multiple target processes are available and an approach for online information seeking via question generation.