Monitoring Trust in Human-Machine Interactions for Public Sector Applications

Trust in AI systems can be measured through biological signals: The research demonstrates that physiological responses captured through EEG (brain activity) and GSR (skin conductance) sensors can reliably detect whether humans trust AI systems, offering an objective way to monitor trust beyond self-reported surveys.

Explainability is essential for building trust: The paper emphasizes that understanding why an AI system makes decisions about trust levels is crucial—using explainable AI methods (LIME) makes HMI systems more transparent, which paradoxically helps promote trust in the technology itself.

Less data can achieve better results: By identifying which specific features of physiological data actually matter for predicting trust, researchers can reduce the amount of training data needed, making trust-monitoring systems more practical and cost-effective for real-world public sector deployment.

Public sector applications need practical solutions: The research specifically addresses governmental and public sector use cases, recognizing that trust monitoring systems must be implementable in actual work environments—not just theoretically sound in laboratories.

Current AI tools lack transparency about what drives predictions: Traditional machine learning classifiers (neural networks, support vector machines) can predict trust levels but don't explain which factors are most important, making it difficult to improve human-machine relationships; explainable AI methods address this gap.