Addressing Trust in AI Systems through Education: A Didactic Perspective

Research identifies a critical challenge in Machine Learning (ML) education: the presentation of AI systems as 'black boxes' leading to superficial understanding and hindering the development of calibrated trust among users. A new didactic framework, ICE-T, is proposed to address these issues by integrating intermodal transfer, computational thinking, and explanatory thinking, aiming to foster deeper comprehension and appropriate reliance on AI.

Why it matters

The ability to foster calibrated trust and deeper understanding of AI systems is critical for their responsible deployment and integration across various sectors. Without it, individuals and organizations may either over-rely or under-rely on AI, leading to suboptimal outcomes or significant risks. This framework offers a structured approach to enhance AI literacy, which is foundational for effective human-AI collaboration and strategic decision-making.

Key insights

• Current ML education often treats AI as an opaque 'black box,' limiting learners' understanding.
• This opacity prevents users from developing calibrated trust necessary for appropriate reliance on AI systems.
• The ICE-T didactic framework is introduced, based on intermodal transfer, computational thinking, and explanatory thinking.
• The framework aims to improve understanding and cultivate appropriate trust in AI.
• It connects to existing empirical literature on algorithm aversion, AI literacy, and mental model formation.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.02453