Title: Making Your Way as a Researcher in Spoken Dialogue Systems

Abstract: As little as 10 years ago, very few people cared much about interacting with machines using language, but since ChatGPT's release in late 2022, now chatbots are common and mainstream. What is next for spoken dialogue systems? A lot. In my talk, I will highlight some key areas that I think have a bright future including some of my own research, and offer some career advice. 

Website: https://www.caseyreddkennington.com/about

Title: Shared Understanding, Shared Responsibility: Towards Grounded GenAI Models for More Reliable Human-Robot Interactions 

Abstract: As language-capable robots have become more widespread, with GenAI models supporting conversational style as a standard interaction mode, critical questions have arisen concerning contextual understanding, multimodality, accuracy, personalization, trustworthiness, and sustainability of these systems. In this talk, I will present insights from recent projects and ongoing research that explore how GenAI models can enhance human-robot interaction. Starting with the enablements of communication, I will discuss the design and development of real-world applications, focussing on technical aspects related to knowledge graphs and grounding, construction of shared context and integration of information from the environment, and addressing broader goals of creating responsible AI for natural, safe and reliable HRI systems.

Bio: https://orcid.org/0000-0003-1229-239X CDM 

Interact robot videos: https://www.cdminteract.com/videos

Title: Keeping it Real: The Role of Humans in Spoken Dialogue System Research 

Abstract: AI now facilitates the automation of cognitive and communicative tasks in a manner similar to the way the industrial revolution facilitated automation of physical tasks. This automation has been applied to the design, development and evaluation of Dialogue Systems. Automation or simulation scales better than human-performed tasks and can be a valuable contribution to the research lifecycle. However there is a risk in automating too much that some value is lost, especially when current LLMs are subject to problems like hallucination and bias. In this talk I will illustrate some examples of issues that come up in automating too much too soon and how keeping humans involved in spoken dialogue system development and research can lead to better systems and science than pure automation.

Website: https://people.ict.usc.edu/~traum/