Do Children Trust AI, and Should They? Designing and Validating a Child-Centred K-AI Trust Scale for Intelligent Systems

Paper Title

Do Children Trust AI, and Should They? Designing and Validating a Child-Centred K-AI Trust Scale for Intelligent Systems

Publication Info

• Topic area: Human-Computer Interaction (HCI), focusing on child-AI trust measurement.
• Keywords: Trust in AI, child-computer interaction, intelligent systems, K-AI Trust scale, dispositional trust, situational trust, child-centred design, psychometric validation, ethical AI, transparency.

Background and Problem

• Problem / challenge: Existing trust metrics for intelligent systems are adult-centric and fail to account for children’s developmental, cognitive, and interpretive differences. There is a lack of validated tools to measure children’s trust in AI systems, particularly in educational and everyday contexts.
• Significance: As AI systems increasingly interact with children in learning and communication environments, understanding and calibrating trust is critical to ensure these systems support critical thinking and ethical engagement rather than blind reliance or premature rejection.
• Motivation and related work: Prior research has focused on adult users, framing trust as dispositional rather than situational. Child-centred trust constructs, integrating ethical principles such as transparency and fairness, remain underexplored. This paper builds on child-rights frameworks and automation trust models to address this gap.

Solution

• Proposed approach: Development and validation of the K-AI Trust Questionnaire, a child-centred tool that measures dispositional and situational trust in AI systems, grounded in ethical principles and developmental psychology.
• Novelty:
  1. Creation of a developmentally appropriate trust scale tailored for children aged 9–11.
  2. Integration of dispositional (baseline) and situational (post-interaction) trust components.
  3. Alignment with child-rights principles (e.g., transparency, fairness, privacy) and ethical AI frameworks.
  4. Iterative psychometric validation across three studies, ensuring reliability and ecological validity.
• Procedure and key techniques:
  • Study 1: Adapted the Propensity to Trust Technology (PTT) scale for children, assessing dispositional trust.
  • Study 2: Developed and tested a preliminary K-AI Trust Questionnaire after child-AI interaction, exploring situational trust.
  • Study 3: Finalised and validated the K-AI Trust Questionnaire, refining items based on psychometric analysis and thematic insights.
  • Mixed-methods approach combining exploratory factor analysis, confirmatory factor analysis, and qualitative thematic analysis.

Results

• Concrete findings:
  • Revised PTT scale demonstrated moderate trust levels (M = 2.83) but low internal consistency (α = 0.59).
  • Preliminary K-AI Trust Questionnaire revealed a tentative two-factor structure (Usefulness and Reliability, Transparency and Critical Awareness), with moderate reliability (α = 0.70).
  • Final K-AI Trust Questionnaire supported a unidimensional structure with improved reliability (α = 0.75) and measurement invariance across gender and age groups.
• Advantage over baselines:
  • K-AI Trust captures situational trust more effectively than the adapted PTT, aligning with children’s lived experiences and ethical concerns.
  • Improved psychometric properties and developmental appropriateness compared to adult-centric trust measures.
• Experiments / evaluation:
  • Sample sizes: 289 children in Study 1, 85 children in Studies 2 and 3.
  • Age range: 9–11 years.
  • Methods: Psychometric validation (EFA, CFA), thematic analysis, and measurement invariance testing.
  • Metrics: Reliability (Cronbach’s alpha), factor loadings, convergent and discriminant validity, and thematic coding of qualitative responses.
• Limitations and future work:
  • Limited sample diversity (homogeneous socio-cultural context).
  • Potential familiarity effects due to repeated participation in multiple study phases.
  • Need for longitudinal studies to trace trust trajectories and refine ethical constructs like transparency and agency.
  • Expansion to other domains (e.g., clinical or safety-critical settings) and cross-cultural validation.

Summary

This paper introduces the K-AI Trust Questionnaire, a child-centred tool designed to measure trust in intelligent systems. Through iterative psychometric validation across three studies, the authors developed a reliable, unidimensional scale that integrates dispositional and situational trust components. Results highlight children’s holistic trust evaluations, blending functional, relational, and ethical cues, and underscore the importance of developmentally appropriate measures. The findings contribute to child-centred HCI by providing a robust tool for evaluating trust and fostering ethical, empowering AI interactions for children. Future work will focus on refining constructs, expanding sample diversity, and exploring longitudinal and cross-cultural applications.