Bias-aware business process management for AI adoption: multi-country evidence on calibrated reliance and responsible process performance
Ling Xu, Xianghan ZhangPurpose
This article examines how perceived algorithmic-bias salience, automation bias and algorithm aversion are associated with artificial intelligence (AI) use in organisational work processes. It develops a bias-aware business process management (BPM) model of calibrated reliance, process integration and responsible performance.
Design/methodology/approach
A six-country professional survey produced 600 usable responses. Survey constructs are triangulated with respondent-transcribed KPI summaries, 90 de-identified incident records and 3,600 process-month observations. Partial least squares structural equation modelling, measurement invariance of composite models, country models, bootstrapping and process-clustered panel tests are used.
Findings
Perceived algorithmic-bias salience is negatively associated with calibrated reliance, while algorithm aversion is negatively associated with process integration. The automation-bias path is unsupported; exploratory behaviour tests do not validate H2. Bias-aware BPM governance is associated with reliance discipline and integration, but its direct panel association is non-significant. Key performance indicator and panel evidence are triangulation, not causal proof.
Originality/value
The article defines bias-aware BPM capability as the capacity to place, audit and revise responsible-AI controls at process decision points. It specifies the roles, records, review triggers and feedback routines through which AI outputs become challengeable and correctable in workflow execution.