The concept of "interaction" in debates on human-machine interaction

Recent advancements in AI, ML, neurotechnology, and social robotics have spurred extensive debates on HMI, its implications for engineering, science, ethics, and politics. While informatics and engineering initially focused on efficient user interfaces for data transfer (HCI), the emergence of BCIs, virtual reality, and ubiquitous computing shifted the focus to a more nuanced understanding of interaction, often framed as a form of communication or dialogue between humans and machines. This broader perspective, referred to as HMI, necessitates contributions from diverse fields such as sociology, philosophy, psychology, and communication science, all striving to simulate human communication characteristics in machines. However, the lack of a unified understanding of "interaction" hinders productive discussions about ethical and philosophical aspects of HMI. This paper aims to address this issue by providing a framework for a more coherent and fruitful interdisciplinary discourse.

The authors discuss the existing literature on the concept of "interaction," highlighting its ambiguity and vagueness, particularly in scientific contexts. They point to the work of Janlert and Stolterman (2017), who noted the colloquial understanding of "interaction" contrasts with its ambiguous usage in scientific discussions. They also refer to existing work which uses related terms such as interactivity, interactability, and interactiveness. The review highlights the lack of a single, universally accepted definition of interaction, which impedes effective debate on ethical, political, engineering, and scientific questions surrounding HMI.

To address the ambiguity surrounding "interaction," the authors introduce a four-dimensional SMPC model. This model provides a framework for analyzing the concept along four dimensions: 1. **Subjects:** Who or what entities are involved in the interaction? 2. **Modes:** How do the subjects interact? What methods or means are employed? 3. **Purposes:** Why does the interaction occur? What are the goals or intentions? 4. **Contexts:** Where and under what circumstances does the interaction take place? The authors then apply this SMPC model to analyze the concept of interaction within various disciplines including informatics/computer science, game theory, sociology, philosophy, psychology/cognitive science, and media studies/communication science. For each discipline, they examine how the four dimensions of the SMPC model are interpreted and applied, highlighting both similarities and differences in understanding.

The analysis reveals a diverse range of understandings of "interaction" across disciplines. **Informatics/Computer Science:** The evolution of interaction is categorized into three stages: (1) early focus on algorithms and engineering; (2) HCI emphasizing user interfaces for data exchange; and (3) HMI encompassing communication and dialogue. This transition highlights a shift from a primarily user-centered, asymmetrical view of interaction to a more symmetrical one. **Game Theory:** Interaction is analyzed through the lens of rational decision-makers employing strategies to maximize rewards within specific contexts, making it useful for understanding strategic HMI. **Sociology:** Traditional sociological views limited interaction to human-human relations. Contemporary approaches (ANT and STS), however, broaden this to encompass human-non-human interactions, highlighting the role of technology in shaping social structures. Various theories are reviewed and categorized according to preconditions for interaction (intelligence, intentionality, embodiment, adaptability, symbolic exchange, intensity/friendliness). **Philosophy:** The philosophical approaches explore the notion of interaction within post-phenomenology, technoscience, enactivism, and analytical philosophy. Post-phenomenology emphasizes the mutual shaping of humans and technology through interaction. Technoscience introduces "intra-action" replacing the concept of interaction to emphasize the inseparable entanglement of agents. Enactivism focuses on the situated and co-regulated coupling of autonomous agents. Finally, analytical philosophy approaches interaction from ethical and epistemological perspectives. **Psychology/Cognitive Science:** Interaction is understood as a dynamic process involving mutual adaptation and the development of complementary behaviors. The importance of cognitive (mental states) and emotional factors is emphasized. However, transferring the understanding of human-human interaction to HMI reveals significant challenges, particularly in the context of social interaction. Different interaction types (instructing, conversing, exploring, responding) are considered for analyzing HMI. **Media Studies/Communication Science:** This field distinguishes between message-centered and structural approaches, with message-centered approaches focusing on user control over content and structural approaches viewing interactivity as a property of technology. Perceptual approaches consider the subjective experience of interactivity.

The study's findings demonstrate the multifaceted nature of "interaction" and highlight the limitations of applying a single definition across diverse fields. The SMPC model offers a framework for resolving ambiguity by systematically analyzing the four dimensions of interaction in specific contexts. The diverse disciplinary perspectives provide valuable insights into understanding and designing HMI. Understanding these variations is crucial for avoiding misunderstandings and for making informed decisions in normative debates on HMI.

This paper offers a comprehensive analysis of the concept of "interaction" within the context of HMI, using a novel four-dimensional SMPC model. It reveals the multifaceted nature of "interaction" and the need for a nuanced understanding across disciplines. This work emphasizes the importance of acknowledging the varying interpretations of "interaction" to promote more fruitful and effective interdisciplinary discussions on HMI. Future research should focus on further refining the SMPC model and applying it to specific HMI contexts to develop more nuanced theories and designs.

While the SMPC model provides a valuable framework, it does not offer a definitive solution to the problem of defining "interaction." The application of the model to different disciplines reveals complex and nuanced interpretations that cannot be easily reduced to a single, universally applicable definition. Future research is needed to further explore the interactions between the four dimensions of the SMPC model and their implications for different types of HMI.