LLMs are more intrusive and more consistent than humans, eager to view neutral discussions as opportunities for positive reinforcement¶

Claim · All claims
Evidence · 1 study · associational r1 · 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? LLMs tend to over-intervene and are more consistent among themselves than humans, interpreting neutral narrative discussions as chances for positive reinforcement. → Dimitris Tsirmpas 2026

Evidence¶

Dimitris Tsirmpas 2026¶

Dimitris Tsirmpas, Katerina Korre, John Pavlopoulos. (2026). To Facilitate or not to Facilitate: Human and LLM Facilitator Tendencies in Online Discussions. https://arxiv.org/abs/2607.28643 doi:10.48550/arxiv.2607.28643

q2 · i? · associational · r1

The same survey repeated with six open-source LLMs on PEFK discussion chunks. In 36 cases no human agreed with the LLM majority vote despite LLM agreement of at least 0.66; qualitative analysis suggests LLMs "being eager to view neutral discussions as opportunities for positive reinforcement."

"LLMs are more intrusive and more consistent compared to humans, being eager to view neutral discussions as opportunities for positive reinforcement."

Discussion¶

Related Claims¶

• Human-LLM sentiment divergences were patterned: the LLM tended to assign affective labels to statements a human coder read as neutral — possibly the same claim (merge candidate)
• LLM coders replicate many human discursive moves but lack questioning, adaptive responsiveness, and emotional expressiveness — related
• Systems Thinking Practices is the hardest CT construct to code reliably for both human raters and the LLM, requiring more explicitly operationalized rubrics — related
• Making trustworthiness metrics and visualizations explicit increased inter-rater reliability among learning engineers evaluating LLM responses — related