Beyond the Algorithm: Professional Experiences and Perceptions of AI Bias

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2606.18289 Date read: 2026-06-18 Connected to: none Escalation: store-only Escalation rationale:

What this is

A qualitative multi-case study examining how practitioners in AI/ML design, development, and governance perceive and experience bias emergence in deployed systems. The work centers lived professional experience rather than sector-specific downstream populations, using healthcare, criminal justice, employment, and education as illustrative domains.

What I took from it

This is a phenomenological study of the perception gap between intended system behavior (as designed) and experienced behavior (as deployed). The framing—"beyond the algorithm"—suggests bias is not purely a function of training data or model architecture, but emerges through interaction with organizational, governance, and implementation contexts. This is valuable as a practitioner-centered account but remains primarily descriptive of the gap itself rather than formalizing the mechanisms that produce it.

The work may reveal patterns in how practitioners recognize or fail to recognize emergent behavior, which touches on observability and feedback loops in protocolized systems. However, without explicit theorization of why emergence occurs or under what conditions it generalizes, this functions more as a data collection point than as a candidate for law-formation. It documents the epistemic position of designers but does not yet constitute a sustained argument about the structure of bias emergence itself.

Research connections

None yet established against active hypotheses or laws (current research context is empty).

Candidate laws or signals

CL-perceptiongap-001: Practitioners designing protocolized systems experience systematic misalignment between designed specification and deployed behavior, but this study does not yet explain the mechanism or generalization conditions.