Abstract

Several recent successes in reinforcement learning have relied upon input from human users and domain experts. Humans can provide information that helps RL algorithms learn effective policies for well-specified tasks. They can also help specify the task to be solved by RL, increasing the alignment of learned policies with the interests of human stakeholders.

This workshop focuses on two challenging questions:

1. How can RL agents be designed to leverage feedback, guidance, and other information from humans that they will learn from, interact with, and assist?

2. How can interactions and experiments be designed to be reproducible and compassionate to the humans involved?

We bring together interdisciplinary experts in interactive machine learning, RL, human-computer interaction, cognitive psychology, robots, and related social sciences to explore and discuss these challenges and what we can bring from our various fields of expertise to address them.

Program

• 1:00pm Welcome (10 min) - Kory Mathewson

• 1:10–2:10pm Session 1 (1 hr) - Peter Stone, Ashley Edwards, Matthew Gombolay

• Break (10 min)

• 2:20– 3:00pm Session 2 (40 min) - Minwoo Lee and Benjamin Poole, Michael Littman

• Break (10 min)

• 3:10– 3:50pm Session 3 (40 min) - Dylan Hadfield-Menell, Brad Knox

• Break (10 min)

• 4:00– 5:00pm Group activity (1 hr) - facilitated by Kory Mathewson

Organizers

Kory Mathewson (DeepMind) and Brad Knox (Google)