Emotional prompting boosts creativity, hurts technical accuracy

This research empirically validates that "persona prompting" isn't a universal performance hack—it actively trades precision for divergence.

• –Injecting high curiosity (0.95) combined with mild frustration (0.20) yielded the best results for open-ended reasoning, outperforming pure curiosity.
• –The technique degrades performance on strictly technical tasks, proving that emotional prompts should be scoped to the specific domain.
• –Manipulating "confidence" levels didn't alter raw quality scores but dramatically changed epistemic style, causing 4x more linguistic hedging under low confidence.
• –The open-source Cortex-Nexus framework provides a rigorous methodology to test prompt engineering claims instead of relying on subjective vibes.