Call for Papers

In this workshop, we aim to bring together researchers investigating ToM from different research perspectives to define a unifying agenda for Mutual Theory of Mind (MToM) in human-CUI interactions, where both humans and CUIs exhibit ToM-like capabilities throughout their interactions.

Relevant work may involve CUIs taking on diverse social roles (e.g., teammates, tutors) or propose fresh directions for ToM research in human-CUI interaction. We particularly encourage work that examines how ToM-like abilities can be operationalized, designed, and evaluated, raising questions around which metrics and benchmarks truly capture whether a CUI ‘has’ ToM.

We invite academic and industry researchers to contribute 2–6 page papers (in ACM double-column format) that include position statements, literature reviews, or in-progress empirical studies. We are especially interested in short position papers that make an argument for a viewpoint or perspective about what should be done in ToM in human-CUI interaction to advance the current research and industry landscape.

All submissions will be assessed based on their quality and relevance to ToM in human-CUI interaction, and accepted papers will be made available on the workshop website. At least one author of each accepted submission must attend the workshop in person. For further information, please feel free to contact Qiaosi Wang (Chelsea) at qiaosiw@andrew.cmu.edu

Workshop Themes

We seek submissions that explore the following three broad topics. These topics and questions are meant to be inspirational instead of exhaustive:

1. Building and measuring a CUI’s ToM-like capabilities in human-CUI interactions

   • What are the AI application contexts that would benefit from CUIs having ToM-like capability? (e.g., autonomous driving, personal assistants)
   • What techniques, methods, models, and data can be used to build a CUI’s ToM-like capability? (e.g., machine learning techniques, cognitive models, foundational models)
   • How can CUIs adapt or personalize its conversational responses to a user based on the user’s mental states?
   • What factors from a CUI’s design (e.g., voice pitch and/or gender) could influence user’s perceptions of a CUI’s social roles?
   • How to evaluate CUI’s ToM-like capabilities during human-CUI interactions?
   • What does it mean to design a CUI’s ToM-like capability in an ethical and human-centered manner?

2. Understanding and shaping human ToM in human-CUI interactions

   • What kind of mental states (e.g., beliefs, blame) do people attribute to CUI with varying social capacities?
   • How does people’s mental states attribution to CUI relate to the CUI’s conversational capabilities and features?
   • How does the design of the CUI (e.g., features, functionalities) influence people’s mental state attribution behaviors to the CUI?
   • How do CUIs impact people’s expectations and perceptions of CUI compared to non-conversational AI systems?
   • How do people perceive and react to CUIs that exhibit ToM-like capabilities through conversations?
   • What kind of users (e.g., personality traits) might benefit more from interacting with CUIs equipped with ToM-like capabilities?

3. Envisioning MToM in various human-CUI interaction contexts

   • How can the vision of MToM inform the design of human-CUI interactions (e.g., turn-taking, mutual shaping of dialogue) in different contexts?
   • How does having MToM in human-CUI interactions impact the quality and outcome of human-CUI interactions?
   • What can we learn from human-human conversations to inform the design of MToM in human-CUI interactions?
   • What might MToM look like in conversations that involve multiple humans and multiple CUIs?
   • How can we measure, assess, and evaluate MToM in human-CUI interactions?
   • What are the positive and negative consequences of having MToM in human-CUI interactions?