Abstract

Abstract Organizations increasingly rely on hybrid human–algorithm decision systems in domains such as hiring, credit evaluation, financial analytics, and risk assessment (Fan & Shu, 2025; Fan & Wang, 2026). These systems are often expected to reduce human judgment errors and improve fairness. However, evidence suggests a more complex dynamic: algorithmic tools may both correct and reinforce human biases, while human oversight can mitigate or amplify algorithmic distortions (Huang et al., 2025). If left unexamined, such interacting biases may accumulate and amplify over time, leading to systematically distorted decisions and downstream harms for stakeholders. Existing research offers limited explanations for these interaction dynamics. Psychological studies show how cognitive heuristics generate systematic judgment errors (Tversky & Kahneman, 1974), while algorithmic fairness research examines how biased data and optimization objectives propagate inequality through automated pipelines (Barocas & Selbst, 2016). Although recent work acknowledges the sociotechnical nature of algorithmic systems (Selbst et al., 2019), current scholarship rarely explains how bias propagates when human cognition and algorithmic computation jointly shape decisions. This paper asks: how do human and algorithmic biases interact within hybrid decision systems? To address this gap, this paper proposes the Hybrid Distortion Architecture (HDA) model, which conceptualizes bias as structured distortion generated by interacting human and algorithmic mechanisms within a shared decision environment. The study adopts a conceptual theory-building approach, integrating insights from cognitive bias, algorithmic fairness, and sociotechnical systems literature. The proposed framework explains how distortions originate in both human and algorithmic processes and how their interaction can amplify and stabilize biased outcomes in hybrid decision environments. By reframing bias as an interactional architecture rather than a single-layer phenomenon, this study contributes a conceptual lens for understanding bias propagation in hybrid sociotechnical decision systems and highlights governance opportunities for addressing bias in such environments.

Download

Share

COinS