Middle Managers and AI-Assisted Performance Appraisal in Chinese Enterprises: Explainability, AI Literacy, and the Case for Human Oversight

Abstract

As AI-based appraisal tools move from pilot projects into daily HR routines, middle managers face a practical dilemma. They are asked to use algorithmic scores and model-generated recommendations in rating, feedback, promotion, and compensation decisions, yet they remain accountable for explaining those decisions to employees. This article offers a structured review of research on AI-assisted performance appraisal, algorithmic management, explainability, AI literacy, and procedural fairness. Rather than treating managerial confidence as a simple matter of technological acceptance, the article argues that what matters is calibrated reliance: the ability to use AI outputs seriously without surrendering judgment to them. Three conditions recur across the literature. First, explainability reduces procedural opacity and makes appraisal outcomes easier to defend. Second, AI literacy equips managers to interpret outputs, detect limitations, and resist both blind trust and reflexive rejection. Third, human oversight preserves accountability, contextual correction, and respectful treatment in employee-facing decisions. Building on these themes, the article develops an integrative framework linking explainability, AI literacy, and oversight to middle managers' reliance on AI-assisted appraisal. The discussion then considers practical implications for Chinese enterprises, where performance evaluation often carries strong consequences for pay, promotion, and internal mobility. The paper contributes to the AI-HRM literature by shifting attention from adoption alone to the managerial conditions under which AI-supported appraisal becomes usable, legitimate, and organizationally defensible.