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

Rapid advances in AI, ML, neuro- and self-tracking technologies, and social robotics have intensified debates about conditions for successful human–machine interaction (HMI) and its implications for engineering, science, ethics, and politics. Historically, work focused on human–computer interaction (HCI) as data and information transfer via interfaces. Since the 1980s, developments such as BCIs, AR/VR, ML, and ubiquitous computing fostered a broader notion of HMI that invokes communication or dialogue and draws on sociology, philosophy, psychology, cognitive science, media and communication studies. Despite its centrality, the concept of "interaction" remains vague and ambiguous across and within disciplines, hindering clear ethical, political, and design deliberation. The paper sets out to analyze and clarify the different meanings and dimensions of "interaction" relevant to HMI by: (1) introducing a four-dimensional SMPC model—subjects, modes, purposes, contexts; (2) distinguishing related terms (interactivity, interactability, interactiveness); and (3) surveying disciplinary understandings (informatics/computer science, game theory, sociology, philosophy, psychology/cognitive science, media and communication studies), followed by (4) an analysis mapped to the SMPC model to support clearer intra- and interdisciplinary discourse.

The paper reviews how "interaction" is conceptualized across key disciplines relevant to HMI and clarifies related terms: - Related terms: Interactivity (either disposition for interaction or ongoing process), interactability (specific interaction-enabling properties of artifacts in context), and interactiveness (a system’s propensity to engage users). - Informatics/computer science: Three historical stages. Stage 1 (1840s–1950s) emphasized engineering and algorithms. Stage 2 (1950s–1980s) focused on HCI—interfaces for explicit data/information exchange to solve user-defined problems (interactability/interactiveness). Stage 3 (late 1980s–present) broadened to HMI—responsive environments, dialogue, agency, and parameters like pace, independence, receptivity, predictability, enforcement; humans and machines increasingly treated as more symmetrical subjects. - Game theory: Models strategic interactions among rational decision-makers, with strategies chosen to maximize payoffs in given decision contexts; useful for analyzing strategic HMI (e.g., control, mutual understanding). - Sociology: Traditionally limited to human–human interaction; STS and ANT broadened to include non-human entities (machines) as actors. Modes include mutual adaptation, symbolic exchange, role-taking, mutual awareness; purposes include exchange of goods, creation of order and meaning, and construction of structures, all context-dependent. - Philosophy: Post-phenomenology (Verbeek) emphasizes mediation relations and context; technoscience (Barad) replaces interaction with intra-action (agents emerge through relations); enactivism (Gallagher) frames interaction as mutually engaged, co-regulated coupling between autonomous agents; analytical philosophy of information (Floridi) treats interactivity as agent–environment mutual action and as an epistemic criterion via internal/external interactions. - Psychology/cognitive science: Interaction viewed as dynamic coupling, continuous mutual adaptation, complementary behavior; modes include gesture, affect, cognition, dialogue; highlights limits of transferring human–human social interaction models (e.g., ToM, social penetration) directly to HMI; proposes interaction types (instructing, conversing, exploring, responding). - Media studies/communication science: Message-centered (user control over content, mediated conversations), structural (interactivity as media/technology property enabling influence), and perceptual approaches (user’s experienced interactivity/telepresence), typically in mediated contexts; German media theory focuses on technology-immanent operations.

Conceptual and analytical synthesis. The authors introduce the SMPC model (Subjects, Modes, Purposes, Contexts) to structure the concept of interaction, clarify related terms (interactivity, interactability, interactiveness), and conduct a cross-disciplinary literature analysis. They historically and thematically review major disciplinary treatments (informatics/computer science, game theory, sociology, philosophy, psychology/cognitive science, media and communication studies), then map core elements of each to the SMPC dimensions. No empirical data collection or quantitative analysis was performed.

- A four-dimensional SMPC model clarifies interaction by specifying: who interacts (subjects), how (modes), why (purposes), and where/under what conditions (contexts). Variation across these dimensions explains the term’s ambiguity. - Informatics/computer science: Identifies three stages from algorithmic/engineering focus to HCI (explicit information exchange) and to HMI (dialogue, agency, responsive environments), shifting from asymmetrical user–computer relations to more symmetrical human–machine relations and expanding purposes beyond epistemic problem-solving to simulation and environmental change. - Game theory: Frames interaction as strategic choice among rational agents to maximize payoffs in defined decision contexts, offering analytic tools for HMI control and understanding. - Sociology: Expands interaction partners to include non-human entities (ANT/STS), emphasizes mutual adaptation, symbolic interpretation, and context-specific outcomes (from meaning-making to institutional structuring), cautioning against uncritical transfer from interpersonal contexts to HMI. - Philosophy: Post-phenomenology and technoscience foreground relationality and context (interaction/intra-action); enactivism defines interaction as co-regulated coupling of autonomous agents; analytical philosophy (Floridi) treats interaction both ethically (agent–environment effects) and epistemologically (interaction as existence criterion). - Psychology/cognitive science: Defines interaction as dynamic coupling and mutual adaptation; highlights cognitive, emotional, and social mechanisms, while identifying limits of directly porting human–human social models (e.g., ToM, social penetration) to HMI; delineates distinct interaction types (instructing, conversing, exploring, responding). - Media/communication studies: Distinguishes message-centered, structural, and perceptual interactivity approaches; emphasizes mediated contexts and user control, technological affordances, and perceived interactivity/telepresence. - Related terms organized: Interactability (artifact-specific enabling properties), interactiveness (propensity to engage users), and interactivity (either disposition or ongoing process) are distinct yet connected, helping disentangle conditions vs processes of interaction. - No single, universal definition of interaction is feasible across disciplines; explicit articulation of the operative understanding is essential for ethical, political, engineering, and scientific debates on HMI.

By mapping disciplinary usages of interaction onto the SMPC model and distinguishing related terms (interactivity, interactability, interactiveness), the paper resolves sources of ambiguity and enables clearer communication about HMI. The findings show how divergent assumptions about subjects (human-only vs including machines), modes (information exchange, dialogue, strategic choice, symbolic interpretation, co-regulated coupling), purposes (problem-solving, meaning-making, relationship formation, environmental transformation), and contexts (technical, social, mediated) shape debates and outcomes. This clarity supports more precise normative analysis (e.g., agency and moral status in HMI), guides design and evaluation (e.g., aligning interactability/interactiveness with intended contexts), and helps avoid misapplied theories (e.g., uncritical transfer of human–human social models to HMI).

The paper contributes a synthesizing framework (SMPC) to analyze and compare concepts of interaction across disciplines central to HMI, and clarifies the linked notions of interactivity, interactability, and interactiveness. Surveying informatics/computer science, game theory, sociology, philosophy, psychology/cognitive science, and media/communication studies, it shows that multiple legitimate understandings exist, precluding a single definition but necessitating explicitness in usage. This conceptual groundwork supports better interdisciplinary dialogue and more robust ethical, political, and design decisions in HMI. Future work can operationalize SMPC dimensions for specific HMI scenarios, develop measures for interactability/interactiveness across contexts, and explore emerging AI-based interaction forms where agency, autonomy, and social cognition are contested.

Conceptual, non-empirical analysis based on selected literatures; no datasets or quantitative validation. The review, while broad, is not exhaustive, and some approaches discussed do not explicitly address HMI but inform it indirectly. The paper does not propose a single definition, focusing instead on comparative clarification.