Enhancing AI Assisted Writing with One-Shot Implicit Negative Feedback

[paper] [arXiv]

We propose a method for improving AI writing assistance by leveraging one-shot implicit negative feedback. When a user rejects a suggestion, this provides valuable information about what not to generate. Our approach uses this implicit feedback to dynamically update the model, leading to more personalised and useful suggestions without requiring explicit user annotations.

Download paper here

Recommended citation: Enhancing AI Assisted Writing with One-Shot Implicit Negative Feedback (Towle & Zhou, EMNLP 2024).