As Artificially Intelligent systems are becoming more and more present in our surroundings, our ways of interacting with them are also changing. From commercial chatbots to home assistants and robot companions, machines are progressively taking up the role of "communicators", provided with their own agency, and able to interact with their human counterparts in new ways. This workshop aimed at gathering experts in fields relevant to the study of AI systems as communicators, including but not limited to Human-Computer Interaction, Artificial Intelligence, Human-Robot and Human-AI Interaction. It was organized in order to discuss new challenges brought by this recent shift, compare methods and perspectives between different fields, and foster long-term collaborations.
Human Interactions with Artificial Intelligence (AI) systems are becoming part of our everyday life. Generating text and images from prompts, asking for help from a website chatbot, or asking a voice assistant to play our favorite playlists are only a few of the possibilities that interaction with AI systems provide. If designed well, these interactions have the potential to enhance human work, abilities, and well-being. In this workshop, we decided to take the particular viewpoint in which AI systems are not merely a tool for expression or communication, but in which they take the role of "communicators", meaning system with which humans create meaning [1]. This shift creates many new challenges and opportunities to design new ways for humans and AI systems to interact. For instance, such AI communicators may have the agency to initiate communication interactions, and should contribute to such interactions efficiently.
From the AI-development side, the concept of "machines as communicators" have been greatly explored albeit mostly between AI systems, as a way to reduce complexity [2] or coordinate a team of robots [3]. Humans have recently entered the picture [4], especially through the rise of Human-Robot Interaction (HRI) and social robotics, and more recently through the development and fast expansion of Large Language Models. With the perspective of having social robots and conversational agents entering our households or workplaces, developing effective models of interpersonal and human-machine communication becomes a priority [5]. For this reason, the need for a deeper connection between communication theory, HRI and social robotics is more and more acknowledged by the community. Embodied communicators (through social robots or virtual avatars) also present both challenges and opportunities in the broader field of communication, as their embodiment provides room to explore a more complex multi-modality of the interaction [6]. Multi-modality can be used to better convey a concept, to meet the needs of a diverse pool of users who might have difficulties interacting via a uni-modal channel, and to try and recover from a failure happening in one of the different channels. However, it is a challenge in itself and an interdisciplinary effort is needed to direct the design of a multi-modal communication medium [7].
The study of AI communicators should also include a Human-Computer Interaction (HCI) perspective. Indeed, the study of communication as a facet of human-technology interaction have a long history in HCI and semiotic theories informed the engineering of human-computer interaction [8]. The HCI field is making use of advanced user research methods and concepts, design approaches, and conceptual frameworks for analyzing UX and social contexts, which are necessary for the study of human-AI communication.
The primary goal of this workshop was to bridge disciplinary boundaries between various fields, included but not limited to AI, HRI, and HCI, in order to gather a multi-perspective view on the topic of Communication in Human-AI Interaction. In particular, we are interested in exploring the core characteristics of AI communicators and human-AI communication, exchanging research methods, and fostering long-term collaboration between practitioners of different fields. As the study of communication in human-AI interaction is by essence a multidisciplinary approach, we aimed for this workshop to be a multidisciplinary platform where researchers can learn to work together and pave the way to impacting research. We also wished to use this opportunity to draw a tentative disciplinary map of the topic of Communication in Human-AI Interaction, describing different perspectives, research directions, methods, and how these perspectives can be related to one another within the research area as a whole.
The workshop's topics of interests included but were not limited to:
• Concepts and theories of communication in human-AI interaction • Human-AI communication design
The workshop received a total of 6 submissions. Each paper was peer-reviewed in a single-blind process by two members of the organizing committee without any conflict of interest with the authors. The reviewers were instructed to consider how relevant to the workshop the submission was as well as its potential to initiate interesting and fruitful discussions. The committee decided to accept 3 papers. Authors of these papers were asked to bring posters.
The workshop was highly interdisciplinary and designed to encourage interaction and discussion. The afternoon was organized as a World Café, with three tables and discussion points:
1. What is Human-AI Communication? 2. What are the problems encountered when studying Human-AI Communication?
3. What research methodologies could / should be applied to the study of Human-AI Communication?
The first poster presented an annotated multimodal corpus of interactions between an autonomous-looking robot and two humans. The presenter also described the process to collect the data, which consisted in an immersive teleoperation system using Virtual Reality (VR). The human operator is equipped with a VR helmet that recreates the visual perception of the robot (cameras). The operator's head, chin, lips, and eye movement are transmitted to the robot in real-time through motion capture and eye tracking. The result is a more natural interaction between the two human and the teleoperated robot. The present argued that such a method allows bringing the social know-how, language understanding, and sensory-motor abilities of a human to a robot, that can then learn by imitation.
The second poster argued that the use of argumentation is under-addressed in the field of Explainable AI (XAI). The author applied Tolmin's theory of argumentation to a machine learning model (logistic regression), for which he showed how to extract data (specific fact) and warrants (general) for claims (fact, specific). He presented a prototype called MindTone to showcase this approach for an argumentative AI communicator.
The third poster presented a conceptual framework for intent recognition in human-robot collaboration, and highlighted different aspects of this problem that need to be considered in the context of Industry 4.0 / 5.0 -relevant settings. This framework addresses three different aspects: the temporal sequence of actions (an intermediate intention may not be immediately observable by a single action), the granularity of intentions, and deviation from the predetermined tasks. The poster also presented the process required by the framework, namely observation and context analysis, intention recognition, deviation detection, adaptation and reaction, and communication.
From the discussions held during the workshop, it appears that the study and design of AI communicators is indeed a blooming, multi-disciplinary research field. Many aspects need to be considered and human-AI communication encompasses human-side, technical, system-wide, and ethical and societal issues. The World Café also highlighted that many research methodologies applied in different fields may need to be considered and integrated, as a lot of them are usually absent from AI research practices. Examples of such are co-creation approaches or observational studies. The discussions also highlighted a strong interest in researchers focusing on human-AI communication for more interdisciplinary collaboration and widening of the practices. On the networking side, the workshop has been highly successful as all participants are involved in follow-up collaborations.
This workshop was partially supported by: • the Swedish Research Council, under grant numbers 2021-05409 and 2022-04676 • the European Union's Horizon 2020 research and innovation program under grant agreement No 952026 • UKRI TAS Node on Trust (EP/V026682/1)