Hysteresis: A common language for the interpretation of AI

Ferraris, M., & Lawrence, P.
In A. Romele, D. Rodighiero, & P. Lawrence (Eds.)
The Routledge handbook for digital hermeneutics
Routledge

Artificial Intelligence (AI) presents theoretical and practical challenges that require explanations addressing both the technical functioning of AI systems and the interpretive needs of diverse stakeholders. Although the philosophy of AI and Explainable Artificial Intelligence (XAI) provide valuable approaches to understanding AI, they often remain divided across computational, ethical, and user-oriented perspectives.

This paper proposes hysteresis as a philosophical and hermeneutical framework for establishing a common language of AI explanation and interpretation. Drawing on four powers of hysteresis—inscription, iteration, alteration, and interruption—the paper develops a conceptual schema for understanding AI systems. Inscription refers to the data, traces, and prior conditions underlying AI processes; iteration describes the repeated computational operations through which patterns are learned and processed; alteration concerns the emergence of new configurations and outputs; and interruption marks the point at which computational processes terminate in a specific output or decision.

Applied to AI and XAI, these categories connect technical methods of model interpretation with intelligible explanations for developers, experts, policymakers, and users. Hysteresis therefore offers a flexible framework for translating AI processes across disciplinary and technical boundaries, providing a foundation for more inclusive, accessible, and context-sensitive approaches to explainability.

Ferraris, M., & Lawrence, P. (in press). Hysteresis: A common language for the interpretation of AI. In A. Romele, D. Rodighiero, & P. Lawrence (Eds.), The Routledge handbook for digital hermeneutics. Routledge.