@inproceedings{leusmann2024comparing,
	title = {Comparing Rule-based and LLM-based Methods to Enable Active Robot Assistant Conversations},
	author = {Jan Leusmann and Chao Wang and Sven Mayer},
	year = {2024},
	booktitle = {Proceedings of the CUI@CHI 2024: Building Trust in CUIs - From Design to Deployment},
	publisher = {CUI},
	pages = {1--5},
	url = {https://sven-mayer.com/wp-content/uploads/2024/06/leusmann2024comparing.pdf},
	date = {2024-05-11},
	abstract = {In the future, robots will be a ubiquitous part of our daily interactions. In contrast to home assistants, robots will assist us by responding to questions and requests and interacting with the environment and humans directly. For this and with the help of recent advancements in AI, we propose shifting from passive to active robots, which can communicate with humans to understand how and when they should perform supportive tasks, which will be especially supportive in collaborative settings. In this paper, we envision two different approaches to how this can be implemented. (1) Rule-based approaches where the dialogue is implemented as a state machine and the conversation procedure is static and predictable. (2) LLM-based approaches, which can dynamically adapt to any situation and ambiguous human input. We compare these two ideas of how robots can first detect the state in which a human is and how they can engage in conversations, and then discuss the advantages and disadvantages of these approaches regarding trust, usability, performance, user agency, and perception of the robot.},
	keywords = {robot assistants, active conversations, rule-based methods, large language models, human-robot interaction, conversational AI, user trust, usability, collaborative robotics, adaptive interactions}
}