Ai systems should support human building before generating answers
The generative paradox: the Anti-Prompt paradigm for human capability
Artificial Intelligence
Summary
When using AI to help people learn or solve problems, there is a risk that the AI’s quick answers stop people from fully understanding or being able to fix things later on. The paper calls this issue the generative paradox and suggests that AI should first help people prepare to build or think through solutions themselves, instead of just giving immediate answers. This approach, called the Anti-Prompt paradigm, aims to keep people skilled and responsible even after AI support ends. The authors show how this method can work in areas like math, science experiments, coding, and lesson planning.
What this means in practice
- •For educational software developers: Design AI tutoring tools that guide users to construct understanding before giving direct answers, preserving long-term learning and problem-solving skills.
- •For software development teams: Create AI-assisted debugging systems that encourage developers to diagnose problems themselves with AI setting up conditions rather than giving immediate fixes.
A position paper. It proposes an approach and reports no results.
Authors
Margarita Trifonov Vadim E. Levit
Abstract
Generative AI in education is usually judged by the output of human-AI teams. That focus overlooks a second outcome: what people can still understand, challenge, repair, and adapt when assistance changes or disappears. We call this tension the generative paradox and distinguish four manifestations: performance, assistance, ownership, and capability stock. We propose developmental function allocation, which complements operational function allocation and meaningful human control by treating retained human capability as a design objective alongside current joint performance. The Anti-Prompt paradigm reverses the default interaction order: when immediate generation would preempt a consequential human act, AI should first create the conditions for human construction. We distinguish the validity of a product, responsibility for its construction, and evidence of capability after support changes. Representative Anti-Prompts, a canonical interaction sequence, and stress tests spanning mathematical reasoning, experimental science, software diagnosis, and lesson design make these distinctions concrete. We conclude with falsifiable predictions about retained capability and a system-level hypothesis about the reproduction and distribution of expertise.