Preparadigmatic Field Dynamics

AI/ML as a field exhibits characteristics of preparadigmatic science: competing frameworks, inconsistent terminology, rapid methodological churn, and lack of settled foundations.

Kuhnian Framing

Thomas Kuhn distinguished between:

• Preparadigmatic: No dominant framework, competing schools
• Normal Science: Shared paradigm, puzzle-solving within framework
• Revolutionary: Paradigm crisis and shift

Current AI/ML has elements of all three, but the foundational instability suggests preparadigmatic characteristics.

Evidence

Terminology Chaos: “Prompt engineering,” “prompt design,” “prompt programming” (same concept, different names Method Proliferation: New techniques weekly, unclear which matter Evaluation Disagreement: No consensus on what “good” means Hype Cycles: Rapid oscillation between breakthrough and disillusionment

Implications for Practice

• Don’t expect stability; build for change
• Be skeptical of claimed “best practices”
• Document decisions and rationale for future revision
• Maintain awareness of competing approaches

Strategic Opportunity

In preparadigmatic periods, practitioners who develop coherent frameworks (even provisional ones) create disproportionate value. The field needs organizing perspectives.

Related: 05-atom—evaluation-metric-limitations, 00-source—schulhoff-2024-prompt-report