Chat Failures and Troubles: Reasons and Solutions

This paper examines some common problems in Human-Robot Interaction (HRI) causing failures and troubles in Chat. A given use case’s design decisions start with the suitable robot, the suitable chatting model, identifying common problems that cause failures, identifying potential solutions, and planning continuous improvement. In conclusion, it is recommended to use a closed-loop control algorithm that guides the use of trained Artificial Intelligence (AI) pre-trained models and provides vocabulary filtering, re-train batched models on new datasets, learn online from data streams, and/or use reinforcement learning models to self-update the trained models and reduce errors.