We are pleased to announce the release of a new publication by DIVERSIFAIR researchers Ilina Georgieva, Tessa Bruijne, Lieke Dom, and Steven Vethman, published in Oxford Intersections: AI in Society and edited by Philipp Hacker.
The paper, “Situating Harms in the AI Governance Landscape”, highlights a growing challenge in AI governance: risk-based approaches often fail to capture the real-world harms experienced by individuals and communities. These harms are shaped by social context, system design, and ongoing interaction, not only by technical risk classifications.
To address this gap, the authors introduce three key characteristics that help make sense of how AI harms emerge:
- Intentionality: Was the system designed or deployed with harmful intent?
- Temporality: Does the harm appear as a single incident, or does it build over time?
- Interactivity: Does harm arise from direct user interaction with the system?
Understanding these dimensions can help policymakers design more effective governance frameworks, support researchers in structuring their analyses, and guide industry and civil-society actors in recognising and addressing potential harms.
This publication is the first in a series of studies on AI harms produced within the DIVERSIFAIR project. Together, these studies aim to advance a more inclusive and context-aware approach to AI governance.
