Priming human shadows as math lovers or haters shifts LLM emotional tone about mathematics¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Emotional tone in LLM interview responses is influenced by priming, with positive scores under the Math Lover condition and reversal under the Math Hater condition for some models. → Esposito 2026

Evidence¶

Esposito 2026¶

Esposito, N., Tricarico, A., Porzio, L., Aghazadeh Ardebili, A., & Stella, M. (2026). Math Education Digital Shadows for Investigating Learning with GenAI: Mathematics Performance, Anxiety, and Confidence in LLMs. arXiv preprint. https://arxiv.org/abs/2604.27618 doi:10.48550/arxiv.2604.27618

q2 · i? · design · r2

Sentiment analysis of Task 1 interview responses using VADER compound scores, averaged over Q1, Q2, Q3, and Q7 and pooled by model within priming groups. The article reports that "Scores are always positive under the “Math Lover” condition" while Grok 4.1 Fast and Magistral Small receive negative scores under the Math Hater condition.

"Fig. 2b shows that emotional tone is influenced by priming. Scores are always positive under the “Math Lover” condition. Mistral Small 4 and Mistral Small 3.2 achieve the highest scores, while Granite 4 Tiny achieves the lowest."

Discussion¶

Related Claims¶

• LLMs simulating survey respondents under AI-assistant instructions report higher mathematics self-efficacy and lower anxiety than under human-shadow conditioning — related
• Human-shadow prompting generates broad psychometric variance while AI-assistant responses are uniform — related
• LLM digital shadows reproduce the human pattern of positive science perceptions anchored in negative mathematics associations — related
• Censoring appears to affect negative-valence ratings of mathematics-related cues between model variants — related
• A general-purpose LLM assessing team emails' emotional tone was biased toward interpreting messages as anxiety-related only — related