Limitation Learning: Catching Adverse Dialog with GAIL
This work addresses the problem of identifying adverse behavior in dialog models for AI and NLP researchers, but it appears incremental as it applies existing imitation learning techniques to a new domain.
The authors applied imitation learning to conversation, recovering a policy for talking to users and a discriminator that classifies between expert and synthetic conversation, with results indicating limitations in dialog models.
Imitation learning is a proven method for creating a policy in the absence of rewards, by leveraging expert demonstrations. In this work, we apply imitation learning to conversation. In doing so, we recover a policy capable of talking to a user given a prompt (input state), and a discriminator capable of classifying between expert and synthetic conversation. While our policy is effective, we recover results from our discriminator that indicate the limitations of dialog models. We argue that this technique can be used to identify adverse behavior of arbitrary data models common for dialog oriented tasks.