Executing Instructions in Situated Collaborative Interactions
This work addresses the challenge of human-AI collaboration in dynamic, situated interactions, which is incremental as it builds on existing instruction-following methods by focusing on error recovery and multi-goal reasoning.
The paper tackles the problem of enabling a system to execute instructions in collaborative scenarios where users act alongside it, requiring error recovery and handling of multi-goal instructions, and demonstrates the approach through a game environment with human user evaluations.
We study a collaborative scenario where a user not only instructs a system to complete tasks, but also acts alongside it. This allows the user to adapt to the system abilities by changing their language or deciding to simply accomplish some tasks themselves, and requires the system to effectively recover from errors as the user strategically assigns it new goals. We build a game environment to study this scenario, and learn to map user instructions to system actions. We introduce a learning approach focused on recovery from cascading errors between instructions, and modeling methods to explicitly reason about instructions with multiple goals. We evaluate with a new evaluation protocol using recorded interactions and online games with human users, and observe how users adapt to the system abilities.