Emotional intelligence, not technical literacy, best predicts worker productivity: well-being and engagement channel the gains, while AI intensity amplifies but cannot replace human-centered capabilities.

Main Finding

Emotional intelligence (EI) is the single strongest predictor of labor productivity across measures of output, employment quality, and economic resilience, outperforming personality traits, AI/digital literacy, and work-environment factors. Psychological well-being and work engagement are key mediators through which EI (and to a lesser extent personality) translate into productivity gains. Contextual moderators — notably work environment and digital/AI intensity — amplify but do not substitute for human-centered capabilities. Ensemble machine-learning models that capture nonlinearities and interactions provide better predictive and explanatory performance than traditional econometric approaches.

Key Points

• Predictor hierarchy: Emotional intelligence > Personality traits > AI/digital literacy ≈ Work-environment factors (in predictive strength for productivity).
• Mediators: Psychological well-being and work engagement significantly mediate EI → productivity; gains arise via sustained mental health and active involvement rather than isolated skill acquisition.
• Moderation: Digital/AI intensity and work-environment quality strengthen the positive effects of EI but cannot replace them; high technology intensity without human-centered capabilities yields weaker returns.
• Nonlinearity and interactions: Relationships among emotional, psychological and economic variables are complex and non-additive—interaction effects (e.g., EI × AI intensity) and nonlinear dose–response patterns are important.
• Methodological superiority: Ensemble machine-learning approaches (which integrate nonlinearities and feature interactions) outperform conventional linear regression/SEM in predictive accuracy and in uncovering heterogeneous effects.
• Theoretical contribution: Extends human capital theory by incorporating emotional and psychological capital; aligns with behavioral economics’ emphasis on bounded rationality and adaptive performance.
• Policy takeaways: Prioritize well-being-centered education and workforce development, integrate EI-building into digital/AI upskilling, and align workforce strategies with sustainable development goals.

Data & Methods

• Conceptual model: Predictors (emotional intelligence, Big Five personality traits, AI/digital literacy, objective work-environment indicators) → Mediators (psychological well-being, work engagement) → Outcomes (labor productivity, employment quality, economic resilience), with moderators (work environment quality, digital/AI intensity).
• Analytical approach:
  • Machine-learning–first strategy using ensemble models to capture nonlinearities and interactions (ensemble methods such as tree-based boosting/forest ensembles and stacking were employed conceptually).
  • Models evaluated against traditional econometric baselines (e.g., linear regression, conventional mediation/SEM) and found to outperform them on fit and predictive metrics.
  • Mediation and moderation effects were estimated within the ML framework (e.g., via model-agnostic mediation decomposition, conditional partial dependence, or targeted ML approaches), enabling discovery of heterogeneous and nonlinear indirect effects.
• Robustness checks: Comparative model performance, sensitivity analyses to alternative variable definitions, and subgroup heterogeneity (by sector, job type, and AI intensity).
• Data sources (as described conceptually): Individual-level measures of EI and personality, validated scales for well-being and engagement, objective/administrative productivity indicators, and measures of workplace digitalization/AI intensity. (Specific sample size and sampling frame not provided in the summary.)

Implications for AI Economics

• Human–AI complementarity: AI/digital intensity increases returns to the human-centered capabilities embodied in EI and well-being. AI adoption policies should therefore invest in emotional and psychological capacities to realize productivity gains.
• Rethinking skills policy: Workforce-development programs should integrate EI training and well-being support alongside technical AI literacy. Technical reskilling alone will be insufficient for resilient employment and sustainable productivity growth.
• Measurement and modeling: AI economics research should incorporate behavioral and emotional variables and use flexible ML methods to detect nonlinear complementarities and heterogeneous effects. However, prioritize interpretable ML techniques (e.g., SHAP values, partial dependence) to extract policy-relevant insights.
• Labor-market design and platform governance: Algorithmic management and AI-driven workplace technologies must be designed to support psychological well-being and engagement; otherwise, technological intensity can accentuate stress and reduce net productivity gains.
• Equity and policy targeting: Because the returns to AI/digital adoption depend on emotional and psychological capital, interventions should be targeted to reduce disparities in these capabilities to avoid widening inequality.
• Research agenda:
  • Causal identification: Use longitudinal data, natural experiments, and RCTs to establish causal pathways from EI and well-being to productivity in AI-rich contexts.
  • Cross-context validation: Replicate findings across sectors, countries, and organizational forms to assess external validity.
  • Responsible analytics: Combine high-performing ML with transparency and fairness diagnostics to avoid amplification of biases in labor-market policy recommendations.

Overall, the study argues that AI economics must move beyond purely technical and skill-based explanations of productivity and employment to incorporate emotional intelligence and psychological well-being as central determinants of how societies realize the gains from digital and AI technologies.