Back
Research Overview
Intelligent Cognitive Prostheses
Today's rapid technological advances present an unprecedented opportunity to augment the human mind with the help of technology. We refer to such technologies as cognitive prostheses. Our goal is to leverage insights from cognitive science to develop technologies that help humanity overcome its cognitive limitations and enable people to set better goals and pursue them more effectively. To achieve this goal, we are developing digital tools that help people make valuable contributions to society. Our tools assist people in each step from setting prosocial long-term goals [] and planning how to achieve them [] to making good decisions [], staying focused on the goal [], and overcoming psychological obstacles along the way [].
Our basic idea for leveraging AI to help people make better decisions is to align each action's immediate reward with its long-term value to make good decisions easier []. We have developed a mathematical framework for designing incentive structures that can be used to guarantee that the rewards won't accidentally incentivize counter-productive behaviors and a computational method for computing incentives that make it as easy as possible for people to choose the course of action that is best in the long run. We are working on instantiating this approach in an intelligent productivity system that helps people plan their days, weeks, and months by computing optimal hierarchical plans and incentivizing the user's tasks and potential short-term goals accordingly. Initial results suggest that this is a promising approach to helping people prioritize their most important tasks and overcome procrastination []. We are currently scaling up this approach [] and integrating it into a digital companion that helps people achieve their goals. We are also working towards incentivizing planning and goal-setting and computing personalized incentives based on inverse-reinforcement learning []. In another project, we successfully applied this approach to incentivize self-directed learning [][].
We combine these computational approaches with psychological theory to guide people towards socially beneficial goals [] and help them overcome psychological obstacles, such as self-defeating thoughts and emotions [].
Members
Publications