Exploitation prompts and GPT keywords concentrate in selected rather than discarded scenarios¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Exploitation prompts were predominantly used in scenarios participants selected (95%) rather than discarded (5%). → Pablo Flores Romero 2025

Evidence¶

Pablo Flores Romero 2025¶

Pablo Flores Romero, Kin Nok Nicholas Fung, Guang Rong & Benjamin Ultan Cowley. (2025). Structured human-LLM interaction design reveals exploration and exploitation dynamics in higher education content generation. npj Science of Learning. https://doi.org/10.1038/s41539-025-00332-3

q2 · i? · design · r2

Descriptive analysis of 25 structured GPT interactions. Table 4 reports 38 exploitation prompts (95%) in the 25 selected scenarios versus 2 (5%) in the 11 discarded ones; 14 selected scenarios (56%) contained GPT keywords versus 1 (9%) discarded.

"Lastly, Table 4 shows a predominant presence of both exploitation prompts and GPT keywords within the selected scenarios, contrary to the discarded ones."

Discussion¶

Related Claims¶

• Exploration prompts exclusively use socially-sourced guide keywords while exploitation prompts mainly use GPT-generated keywords — related
• Cooperativity positively correlates with GPT keyword use in younger, less-experienced participants (Group 1)
• GPT keyword use is negatively correlated with essay topic diversity, suggesting exploitation narrows reflective focus
• Exploration prompt frequency is positively correlated with keyword topic diversity in structured human-LLM content generation
• Participants favor socially-sourced others' keywords over their own in exploration prompts, deviating from personalized learning expectations