"Dataset for Managerial Override of Artificial Intelligence in Operations: An Accountability and Decision Escalation Model "

"Artificial intelligence is increasingly embedded in operational planning and execution, yet many organizations fail to scale value because managers frequently override AI recommendations at the point of decision. Prior work on trust in automation shows that reliance depends on vulnerability and uncertainty, not accuracy alone [1]. Behavioral research also indicates that observing algorithmic error can trigger algorithm aversion, even when algorithms outperform humans [2]. At the same time, evidence of algorithm appreciation suggests that managers may prefer algorithmic advice under certain task conditions [3]. This study develops and tests a decision escalation model that explains AI override as a governance and behavioral control outcome shaped by task criticality, outcome uncertainty, and accountability pressure. Building on escalation of commitment [4] and accountability theory [5], we hypothesize that override increases with criticality and uncertainty, and that accountability strengthens these effects even when perceived AI accuracy is high. We propose a two-study design combining operational system logs with a multi-respondent survey and estimate multilevel logistic models with robustness and endogeneity checks. The study contributes to engineering management by reframing AI governance as decision rights and escalation design rather than a technical tool issue."