Bridging Formal and Perceived Fairness: Development of an Interdisciplinary Framework in Algorithmic Decision-Making

Research on algorithmic systems primarily focuses on formal fairness metrics and technical bias mitigation within Computer Science. However, fairness is a subjective, context-sensitive human judgment influenced by cognitive heuristics, mental models, normative expectations, and sociotechnical factors. This creates a critical divergence where systems satisfying formal technical fairness requirements may still be perceived as unjust by stakeholders.

Why it matters

This distinction between formal and perceived fairness is strategically important as it highlights a potential failure point for algorithmic systems, irrespective of their technical precision. Organizations must address both technical compliance and stakeholder trust to ensure the legitimacy and effective adoption of AI and automated decision-making processes.

Key insights

• Algorithmic fairness research is largely driven by Computer Science, emphasizing formal metrics and technical mitigation strategies.
• Fairness is not solely a technical attribute but a subjective, context-sensitive human judgment.
• Human perceptions of fairness are shaped by cognitive heuristics, mental models, normative expectations, and sociotechnical factors.
• A significant gap exists between an algorithm's formal fairness compliance and stakeholder perceptions of its justice.
• Systems can meet technical fairness criteria yet still be deemed unjust by those affected.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.03853