← CHI 2026 map

CHI '26 · Best paper · full-paper review · confidence high

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

Grazia Ragone , Paolo Buono , Judith Good , Rosa Lanzilotti

DOI PDF Program page Page JSON

This is a strong CHI measurement contribution because it does two things well: it demonstrates that adult trust scales transfer poorly to children, and it replaces that assumption with a carefully refined child-centred instrument. The paper is especially valuable for grounding trust measurement in children’s developmental realities and post-interaction experience rather than abstract adult formulations.

Video Figure

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper stands out because it treats measurement as a substantive HCI contribution rather than a procedural afterthought. The authors begin from a common but weak assumption in AI evaluation: that trust scales developed for adults can simply be simplified and reused with children. Their evidence shows that this assumption does not hold well. The adapted PTT baseline has only modest psychometric performance, and the discussion makes clear that adult-centric wording and abstractions do not map cleanly onto how children understand technology. That negative result is important on its own, because it warns the field against importing adult trust constructs without developmental scrutiny. The stronger contribution is the positive one: the iterative construction of K-AI Trust as a child-centred, post-interaction instrument. The paper positions trust as something children judge in relation to actual experience with an AI system, not merely as a stable trait. That framing is well matched to HCI practice, where designers often need to know whether a specific interaction felt safe, fair, understandable, and worth relying on. The validation story is also credible: the paper reports multi-study refinement, psychometric analyses, factor-analytic evidence, and invariance testing, culminating in a shorter final scale with better reliability and a coherent unidimensional structure. The discussion usefully interprets that unidimensionality not as oversimplification but as evidence that children often integrate practical, relational, and ethical cues into a holistic judgment. The limitations are real and should temper generalization. The sample is narrow in age and setting, all studies were run in the same primary school, and the paper itself acknowledges that the adapted PTT likely needs substantial redesign. Some ethical-agency items also still need refinement. Even with those constraints, the contribution is substantial: it gives researchers and designers a better instrument for studying child-AI trust, and it makes a broader methodological point that developmental fit is central to trustworthy measurement in HCI.

What Changed

Canon before

Most adult trust measures assume complex reasoning and language aligned with adult cognition and treat trust as dispositional or trait-like, overlooking children’s developmental and situational experience with AI systems. Trust in child-AI interaction is often seen as adult trust in automation or a static trait without capturing children’s real interactional trust processes and ethical considerations.

Departure from common sense

The paper breaks the expectation that trust in AI can be directly measured with adult-oriented dispositional scales by showing these scales are developmentally misaligned and unreliable for children. Instead, it shows children’s trust is better captured through a child-centred, post-interaction measure that reflects how they integrate usefulness, fairness, transparency, and lived experience.

Actual novelty

The paper’s main novelty is the iterative design and psychometric validation of the K-AI Trust Questionnaire, a child-centred trust instrument for intelligent systems that combines a dispositional baseline with situational post-interaction assessment and is grounded in child-rights principles. It contributes a developmentally informed measurement tool rather than reusing adult trust scales.

Evidence

The paper supports its claims with iterative questionnaire development across three studies. The abstract reports psychometric analyses on 289 children and confirmatory factor analysis on a subsample of 85. Study 1 shows the child-adapted PTT has questionable internal consistency and a weak item, motivating refinement. Study 3 reports reliability analysis, exploratory and confirmatory factor analysis, construct validity, and measurement invariance testing for the final 9-item K-AI Trust scale. The discussion frames the contribution as a child-centred situational trust measure and explicitly notes remaining limits, especially the need to redesign the adapted PTT and refine ethical-agency items.

Limits

Method limits

The studies were conducted in the same primary school with year 4 and 5 pupils aged 9 to 11, which narrows demographic and contextual diversity. The adapted PTT showed low internal consistency and likely needs substantial redesign, and some ethical-agency items still require refinement for child comprehension.

Deployment limits

The validated instrument is best supported for immediate post-interaction assessment of children’s trust in intelligent systems, especially conversational or similar child-facing AI contexts. Use beyond this age range, school context, or cultural setting requires further validation.

Boundary conditions

Findings are bounded to children aged 9 to 11 in one primary school context and to trust judgments made after direct interaction with an AI system. The final K-AI scale is most directly justified as a situational, post-interaction measure rather than a general dispositional trust instrument.

Position in field

This paper fills an important gap in child-AI and HCI measurement by operationalising trustworthy-AI and child-rights principles in a validated child-centred trust questionnaire. Its contribution is less a new theory of trust than a strong measurement intervention showing why adult-centric instruments are insufficient and how child-appropriate trust assessment can be built.

Abstract

Most trust metrics for intelligent systems are developed for adults, relying on complex reasoning and language that do not align with children’s developmental stages. As intelligent systems increasingly engage with young users, evaluating trust in child-AI interaction has become an urgent concern in HCI. In this paper, we present the iterative refinement and validation of the K-AI Trust Questionnaire, a child-centred instrument that integrates dispositional and situational trust components grounded in child-rights principles. Dispositional trust is captured through a child-adapted Propensity to Trust Technology (PTT), while situational trust is assessed through post-interaction items reflecting children's experience with AI. Starting with a sample of 289 children, we conducted psychometric analyses and exploratory testing, culminating in a confirmatory factor analysis on a subsample of 85 children. Results supported a unidimensional structure consistent with the PTT, and highlighted the limitations of adult-oriented scales, underscoring the need for developmentally appropriate tools for trustworthy child-AI design.