Promoting Critical Thinking With Domain-Specific Generative AI Provocations

Why Read This from an HCI Perspective

This article demonstrates from an HCI perspective that GenAI can be more than a simple answer generator—it can function as a ‘thinking tool’ that disrupts users’ judgments and prompts them to revisit and reconsider. In particular, it examines how questions grounded in domain knowledge, interaction friction, and user-contribution gates can stimulate critical thinking, making it highly relevant to both UX designers and researchers.

CIT's Commentary

From a CIT perspective, the core message of this piece is that ‘good AI’ is not about smoother automation, but about designing context-appropriate discomfort. ArtBot and Privy test the same principles across different domains, and it’s convincing that domain-specific questions aligned to concepts, norms, and evaluation frameworks are far more interpretable and actionable than generic prompts that simply ask ‘Why?’. However, what matters here is not whether friction exists, but who receives it, how much, and at what pace. For experts, it can feel like over-explanation, while for beginners it can be misread as authority. Therefore, CIT views this work as the starting point for adaptive interaction design rather than ‘static prompting’. In other words, supporting critical thinking requires modeling not only domain knowledge, but also users’ skill levels and expectations.

Questions to Consider While Reading

• Q.When designing domain-specific ‘productive friction’, how can we distinguish between novices and experts and adaptively adjust the experience?
• Q.Where should we draw the boundary between GenAI’s role in asking questions and the role users play in thinking for themselves?
• Q.How can we validate that the effect of increasing critical thinking translates into actual work quality or long-term learning?