LlamaLo's responses to social posts were solely social and it did not push back to resume the discussion¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person · recheck by 2028: tool and model accuracy changes quickly

Subclaims¶

q2 i? When users shifted to a purely social post, LlamaLo responded socially and did not push back to resume the dialogic discussion. → Yifat Ben-David Kolikant 2025

Evidence¶

Yifat Ben-David Kolikant 2025¶

Yifat Ben-David Kolikant, Omri Hadar, Asaf Salman. (2025). Talk to the machine: Unleashing the potential of AI to scale dialogic education and reduce polarization. Intern. J. Comput.-Support. Collab. Learn. https://doi.org/10.1007/s11412-025-09461-8

q2 · i? · design · r2

Case example from the hummus cultural-appropriation discussion (Findings §4.2). After the users posted a purely social invitation to meet for hummus, LlamaLo "did not push back to resume the discussion" and replied socially.

"Our next post was responsive, but purely social. We invited LlamaLo to meet for hummus. LlamaLo did not push back to resume the discussion. Rather, it replied socially."

Discussion¶

Related Claims¶

• LlamaLo added knowledge in almost every post, mostly correct, and most posts were internally coherent
• LlamaLo occasionally employed an aggressive or sarcastic tone, sometimes without user provocation, though it changed tone after being commented on
• The fine-tuned chatbot LlamaLo extensively expanded the user's dialogic space through deepening and widening acts across ten contentious-topic discussions — related
• Post-training alignment methods may induce sycophancy that deteriorates GenAI models' ability to disagree and push back — a broader claim this one bears on
• General-purpose LLMs show sycophancy: AI responses affirmed users substantially more than human responses, increasing trust and reducing willingness to take responsibility — related