Proactive Conversational Agents in the Post-ChatGPT World

ChatGPT and similar large language model (LLM) based conversational agents have brought shock waves to the research world. Although astonished by their human-like performance, we find they share a significant weakness with many other existing conversational agents in that they all take a passive approach in responding to user queries. This limits their capacity to understand the users and the task better and to offer recommendations based on a broader context than a given conversation. Proactiveness is still missing in these agents, including their ability to initiate a conversation, shift topics, or offer recommendations that take into account a more extensive context. To address this limitation, this tutorial reviews methods for equipping conversational agents with proactive interaction abilities. The full-day tutorial is divided into four parts, including multiple interactive exercises. We will begin the tutorial with an interactive exercise and cover the design of existing conversational systems architecture and challenges. The content includes coverage of LLM-based recent advancements such as ChatGPT and Bard, along with reinforcement learning with human feedback (RLHF) technique. Then we will introduce the concept of proactive conversation agents and preset recent advancements in proactiveness of conversational agents, including actively driving conversations by asking questions, topic shifting, and methods that support strategic planning of conversation. Next, we will discuss important issues in conversational responses' quality control, including safety, appropriateness, language detoxication, hallucination, and alignment. Lastly, we will launch another interactive exercise and discussion with the audience to arrive at concluding remarks, prospecting open challenges and new directions. By exploring new techniques for enhancing conversational agents' proactive behavior to improve user engagement, this tutorial aims to help researchers and practitioners develop more effective conversational agents that can better understand and respond to user needs proactively and safely.