Beyond Watzlawick: axioms of pragmatic interaction for second-order cybernetics

Purpose To formalize Watzlawick, Beavin and Jackson’s pragmatics of human communication as a level-separated, testable axiom system suitable for second-order cybernetics, so that paradox, reflexivity and context-dependence are treated as phenomena to be modeled rather than pathologies to be excluded. Design/methodology/approach We restate five classical propositions as axioms for unavoidability, meta-inevitability, code choice, observer-dependent punctuation and relational dynamics (A1–A5), and extend them with five interactional axioms (A6–A10) for turn-taking, repair, common ground, relevance defaults and public–private channel separation. We derive short theorems, specify counter-models to test axiom independence, and propose interpretive posterior entropy as an observable for code-variance and leakage effects. Findings The axiom set preserves Watzlawick–Bateson insights while making them empirically usable. It explains how divergent punctuation can produce reciprocal blame, how repeated meta-signals stabilize complementary or symmetric dyads and how code/leakage choices constrain interpretive uncertainty. The framework applies to human–human and socio-technical (human–AI, multi-agent) interaction. Research limitations/implications The paper is primarily theoretical; empirical validation requires interactional corpora annotated for relational/meta content, timing/repair moves and code/leakage features, plus experimental manipulations of observer segmentation. Future work can implement belief-update operators (e.g. AGM-style) and compare entropy-based predictions across media. Practical implications The axioms can inform the design of dialogue protocols, training for mediation/facilitation and human–AI interfaces that accommodate mis-punctuation and repair. Implementing A6–A8 in dialogue managers should reduce human–machine misunderstandings. Social implications By formalizing how escalation and relational locking emerge from observer-dependent segmentation, the framework helps diagnose and pre-empt interactional pathologies in organizational, intercultural and participatory/SDD settings. Originality/value The article shows that pragmatic interaction can be rendered in an observer-dependent, second-order cybernetic calculus that yields testable predictions — something often considered impossible for paradox-rich communication. It bridges Bateson–Watzlawick pragmatics, conversation analysis and contemporary systems/cybernetics.