The Dynamics of Trust in AI Systems: Human–Machine Collaboration Within the MLS Exploitation Process

Abstract

This study examines how trust in human–machine collaboration evolves within the exploitation process of machine learning systems (MLS), spanning from non-ML environments to fully autonomous operation. Building on prior research on the exploitation process of machine learning systems, the study positions trust as a foundational mechanism enabling sustainable human-machine collaboration. Drawing on the trust framework of Siau and Wang (2018), five trust-related dimensions—usability and reliability, collaboration and communication, sociability and bonding, security and privacy protection, and interpretability—were operationalized as survey items measured on a five-point Likert scale. Differences across exploitation phases were examined using ANOVA, and principal component analysis (PCA) was conducted to identify latent user archetypes underlying trust orientations. The results suggest distinct trust configurations across phases of the MLS exploitation process: Visualization, Human-centered ML Assistance, ML-centered Human Assistance, and Autonomy. Early phases emphasize collaboration and communication, highlighting the role of relational trust in initial adoption. As system autonomy increases, interpretability and privacy-related concerns gain relative importance, while in the autonomy phase, security and privacy protection tend to become more salient trust requirements. Interpretability gradually declines as MLS becomes operationally embedded, consistent with a shift from transparency-based trust to stability-based trust. Furthermore, the analysis identifies three latent user archetypes shaping trust orientations in human–machine collaboration: the General Users, the AI Practitioners, and the AI Translators. These findings clarify process-dependent trust dynamics and user-level differentiation within MLS exploitation, offering practical guidance for designing trustworthy AI systems aligned with system maturity and organizational roles

Keywords: Human-machine Collaboration, Software Engineering For Machine Learning Systems, Project Management, Technology And Innovation Management.

DOI: 10.54941/ahfe1007728