[Paper Review] Unraveling the Nuances of AI Accountability: A Synthesis of Dimensions Across Disciplines

Loc H. Nguyen, Sebastian Lins+2 more|arXiv (Cornell University)|Jan 1, 2024

Ethics and Social Impacts of AI1 citations

TL;DR

This paper synthesizes interdisciplinary research on AI accountability, identifying six core themes—trigger, entity, situation, forum, criteria, and sanctions—organized into 13 dimensions that clarify accountability scenarios in AI systems. It provides a structured, multidisciplinary framework to reduce conceptual ambiguity and support future research and practice in AI governance and accountability.

ABSTRACT

The widespread diffusion of Artificial Intelligence (AI)-based systems offers many opportunities to contribute to the well-being of individuals and the advancement of economies and societies. This diffusion is, however, closely accompanied by public scandals causing harm to individuals, markets, or society, and leading to the increasing importance of accountability. AI accountability itself faces conceptual ambiguity, with research scattered across multiple disciplines. To address these issues, we review current research across multiple disciplines and identify key dimensions of accountability in the context of AI. We reveal six themes with 13 corresponding dimensions and additional accountability facilitators that future research can utilize to specify accountability scenarios in the context of AI-based systems.

Motivation & Objective

• To address the conceptual ambiguity surrounding AI accountability due to fragmented research across disciplines.
• To identify and synthesize key dimensions of AI accountability from diverse fields such as computer science, law, and information systems.
• To provide a unified, multidisciplinary framework that supports clearer conceptualization and application of accountability in AI systems.
• To support researchers and practitioners in designing accountability mechanisms by identifying facilitators and structural components of accountability processes.

Proposed method

• Conducted a descriptive literature review of 67 peer-reviewed papers across computer science, law, and information systems.
• Applied a thematic analysis based on Day and Klein’s (1987) categorization framework to identify and structure accountability dimensions.
• Synthesized findings into six overarching themes: trigger, entity, situation, forum, criteria, and sanctions.
• Identified and discussed accountability facilitators such as system transparency, governance structures, and social features that enhance accountability.
• Mapped relationships between dimensions to highlight contextual dependencies (e.g., how situation influences forum selection).
• Validated findings through thematic consistency checks and alignment with established accountability theories (e.g., Bovens, 2007).

Experimental results

Research questions

• RQ1 What are the key dimensions of AI accountability as identified across multiple academic disciplines?
• RQ2 How do different accountability themes—such as trigger, entity, and forum—interrelate in real-world AI systems?
• RQ3 What facilitators enhance or impede accountability in AI-based systems, and how can they be leveraged in design and policy?
• RQ4 How does the current conceptual fragmentation in AI accountability hinder research and practice?

Key findings

• Six core themes—trigger, entity, situation, forum, criteria, and sanctions—were identified as central to AI accountability, each subdivided into 13 distinct dimensions.
• The study reveals that accountability is not a monolithic concept but a multi-layered process shaped by context, actors, and mechanisms.
• Accountability facilitators such as system transparency, auditability, and stakeholder engagement were found to significantly enhance accountability potential.
• The literature shows a strong emphasis on organizational entities as responsible actors, though emerging AI development models (e.g., cloud platforms, open-source communities) challenge this assumption.
• There is a notable lack of empirical validation for the proposed dimensions, highlighting a need for future qualitative and quantitative studies.
• The framework enables better alignment between legal, technical, and social perspectives on accountability, offering a shared vocabulary for interdisciplinary research.