Game Data and Annotations from Agent and Virtual Human in a Split-or-Steal Interaction

Dataset of manuscript: How Personas Can Influence Agents to Play Split or StealIt contains raw data of conversations between players (gemma3:12b, gemma4:e4b, ministral3:3b, and Phi4:14b) against a VH (gpt-4.1 mini) in a scenario of split-or-steal across 15-round game sessions. Data differs in terms of temperature, persona injection, and model chosen. Language: Portuguese (Dialogs), English (insights)