ETHOS-SE: A SOCIO-TECHNICAL FRAMEWORK FOR MITIGATING DEEPFAKE DISINFORMATION

Abstract

The advancement of generative Artificial Intelligence has expanded the production and dissemination of synthetic content, making deepfakes a major challenge to information reliability in digital environments. Although the literature reports significant progress in automated detection techniques, content provenance, algorithmic transparency, digital accessibility, and the integration of these dimensions into software development remain insufficiently explored as a unified problem. This study proposes the socio-technical framework Ethical Technical Handling of Synthetic Entities (ETHOS-SE), conceived from a Software Engineering perspective to support the mitigation of deepfake-based disinformation. Methodologically, the research is qualitative, exploratory, and propositional, grounded in a structured narrative literature review and an exploratory bibliometric analysis of scientific production indexed in Scopus. The analysis identified a predominance of approaches centered on automated detection and gaps related to provenance, explainability, and accessibility. Based on these findings, ETHOS-SE was organized into four complementary layers: Provenance, Detection and Analysis, Explainability, and Accessibility with Cognitive Friction. The main contribution is a conceptual architecture that integrates traceability, transparency, interpretability, and accessibility requirements, extending traditional approaches to disinformation mitigation and providing guidelines for the development of more reliable, transparent, and user-centered systems.