Normalize prompts via domain-consistent terminology substitution and clinical narrative restructuring that add no new information¶

Strategy · All strategies
Evidence · 1 claim (1 for) · 1 study (1 causal), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

Qualitative analysis of PET rewrites identifies two concrete normalization mechanisms. First, domain-consistent terminology: an informal phrase like 3 HIV medicines is rewritten as three-drug antiretroviral regimen, improving linguistic precision without introducing new medical information. Second, clinical information restructuring: a long unstructured narrative is reorganized into semantically equivalent sections. Overall, these examples suggest that PET improves accessibility primarily by enhancing the presentation of existing information rather than altering its content.

Design Implications¶

Context¶

Requirements¶

• Rewrites must preserve the original meaning and may not modify or introduce clinical facts; only presentation and organization of existing information change.

Constraints¶

• Mechanisms are illustrated with two representative medical question-answering examples rather than tested as separate experimental conditions.

Target Learners¶

• users who describe clinical concerns in informal everyday language
• users submitting long, unstructured narratives to AI systems

Target Learning Goals¶

• expressing the same clinical question in terminology consistent with clinical literature and LLM training data
• making diagnostically relevant evidence more salient to the language model

Affordances¶

• Prompt Privilege Prompt Equity Framework

Claims¶

• After PET normalization, no statistically significant performance differences remain between user cohorts [+W]

Related Strategies¶

Examples¶

-

Key Sources¶

• Lier Jin, Lan Hu, Binqi Shen, Hanyu Cai, and Yuting Xin. (2026). Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access. https://arxiv.org/abs/2608.08942