The Association Between Human–AI Interaction and Leadership: A Structural Equation Modeling Analysis in Colombian Organizations

Zarate-Torres R.A.; Rey-Sarmiento C.F.; Acosta-Prado J.C.; Moncada-Niño A. · Technologies (2026) · DOI: 10.3390/technologies14080460

Resumen

Artificial Intelligence (AI) has evolved from a tool for automation into a strategic component of organizational decision-making. However, the extent to which the dimensions of Human–AI Interaction are associated with leadership remains underexplored, particularly in emerging economies. This study examines how interaction quality, productivity enhancement, user experience, organizational impact, ethical governance, and innovation are associated with Leadership Effectiveness and Organizational Sustainability in Colombian organizations. We applied Structural Equation Modeling (SEM) to data collected from 170 participants using a purpose-built 30-item instrument designed to measure eight dimensions of the human–AI relationship through a five-point Likert scale. Six dimensions assessed Human–AI Interaction (Interaction Quality, Productivity and Efficiency, User Experience and Acceptance, Organizational Impact, Ethical Governance, and Innovation and Transformation), while two dimensions assessed leadership (Leadership Effectiveness and Organizational Sustainability). Estimation used maximum likelihood (ML/FIML) as the primary method, with robust ML (MLR) and an item-level WLSMV estimator as sensitivity checks. Correlation analysis (Pearson, with Spearman as a robustness check) revealed consistent positive and significant associations among all construct indicators. Confirmatory Factor Analysis (CFA) confirmed convergent validity, with innovation ((Formula presented.)) emerging as the highest-loading dimension. The structural model demonstrated a significant association between Human–AI Interaction and leadership (two-parcel model: (Formula presented.), (Formula presented.)). Under a more conservative six-item leadership specification, the association attenuates to (Formula presented.) ((Formula presented.)), which we treat as the substantive estimate. Common-method-variance diagnostics (Harman’s first factor (Formula presented.) ; a common latent factor accounting for ≈41% of variance) and an elevated RMSEA (= 0.13) signal the need for expanded measurement models and longitudinal designs to address causality. The findings suggest that ethical governance and innovation-oriented AI interaction are primary correlates of leadership effectiveness in digital organizations operating in emerging economies.

Instituciones: Pontificia Universidad Católica del Perú