Older workers who feel judged by algorithms report feeling dehumanized and less satisfied at work; the harmful link is weaker for employees with more proactive personalities, according to a three-wave survey of 223 Chinese financial-sector staff.

Main Finding

Perceived algorithmic evaluation (PAE) — employees’ sense that their work is being matched, recorded, and rated by algorithms — increases organizational dehumanization among older employees in traditional firms, which in turn lowers job satisfaction. A proactive personality buffers this process: employees higher in proactivity show a weaker PAE → dehumanization link, and the indirect negative effect of PAE on job satisfaction is reduced for more proactive individuals.

Key Points

• Conceptual contributions
  • Extends the PAE construct (previously studied mainly among gig/app workers) to traditional, hierarchical organizations (financial sector).
  • Introduces organizational dehumanization as a dignity-based mediator linking PAE to job satisfaction (a symbolic/relational pathway distinct from workload or burnout).
  • Identifies proactive personality as a personal resource that moderates the first stage (PAE → dehumanization), producing moderated mediation on the PAE → dehumanization → job satisfaction path.
• Core empirical results
  • PAE positively predicts organizational dehumanization.
  • Organizational dehumanization negatively predicts job satisfaction.
  • Proactive personality weakens the positive association between PAE and dehumanization.
  • The indirect (harmful) effect of PAE on job satisfaction via dehumanization is smaller at higher levels of proactive personality.
• Distinctions emphasized
  • Organizational dehumanization is conceptually and empirically separable from job satisfaction (distinct factor, reported correlation r ≈ -0.67).
  • PAE framed as a socio-cognitive job demand (visibility, comparability, externalized accountability) rather than simply surveillance or HR analytics.

Data & Methods

• Sample: 223 employees aged 45+ in 12 financial-services firms across five Chinese regions (Beijing, Shanghai, Zhejiang, Jiangsu, Guangdong).
• Design: Three-wave time-lagged survey (T1, T2, T3).
• Measures: Perceived algorithmic evaluation (PAE); organizational dehumanization; proactive personality; job satisfaction. Confirmatory factor analysis supported construct separability.
• Analysis: Moderated mediation tested using PROCESS Model 7 (Hayes) to examine first-stage moderation (PAE → dehumanization) and the indirect effect of PAE on job satisfaction through dehumanization at different levels of proactive personality.
• Main limitations noted by authors: single-sector (finance) and country (China) sample of older workers; self-report measures and observational design (time-lagged helps temporality but does not prove causality); generalizability to younger workers or other industries remains to be tested.

Implications for AI Economics

• Worker heterogeneity matters for the net returns to algorithmic HR systems. The same algorithmic evaluation can produce productivity/efficiency gains but also symbolic costs (dehumanization) that lower job satisfaction—especially for less proactive, older workers—potentially reducing retention, discretionary effort, and firm human capital value.
• Complementarities and internalities:
  • Firms deploying algorithmic evaluation should pair it with complementary investments (transparency, human oversight, feedback channels, training, job-crafting opportunities) to mitigate relational/dignity harms and preserve long-term human capital.
  • The value of algorithmic HRM depends on endogenous worker responses; unaddressed dehumanization can generate hidden costs (turnover, loss of tacit knowledge) that offset short-term efficiency gains.
• Policy and regulatory considerations:
  • Design guidelines (human-in-the-loop, explainability, right to appeal, participation/voice mechanisms) can reduce symbolic costs and unequal burdens across worker groups (e.g., older employees).
  • Labor-market and welfare analyses of AI adoption should incorporate dignity- and identity-related welfare effects, not only wages and hours.
• Future economic modeling directions:
  • Formal models of algorithmic HR adoption should include worker-level heterogeneity in personal resources (proactivity) and dynamic effects on retention, productivity, and firm-level returns.
  • Empirical work linking PAE and dehumanization to objective outcomes (turnover, performance, sick leave, productivity metrics) would clarify the magnitude of externalities and the cost-benefit calculus of automation in HR.
• Practical takeaway for firms: To maximize the economic benefits of algorithmic evaluation, organizations must design evaluation regimes that preserve employees’ sense of personhood—especially for less proactive, late-career workers—through transparency, opportunities for voice, and human oversight.