Towards Closed-World Reasoning in Games - Ultimatum Game Revisited YunQi Xue Abstract: The Ultimatum Game (UG) is one of the widely studied games in experimental economics. Past data shows a consistent deviation from the classical game theory prediction, which suggests a self-interested money maximizing rational agent would accept any nonzero offer as a responder. However, in reality, people often reject low offers that are less than 30% and above zero. Research from neuro-economics claims that such behavior is mostly emotion driven. However an important cross-cultural study shows that the results in these small-scaled societies deviate drastically from the industrialized societies. In addition, there is a high correlation between the deviations and the social systems. Economists call for a better understanding of how people apply social rules when they make economic decision. We believe that we need a framework that can show the following: 1. Preferences are constructed on demand. 2. Social rules are numerous and complex, and tend to trigger non-monotonic reasoning. Formally, we developed the Closed-World Reasoning in Games framework (CWRiG) adopted from Closed-World Reasoning framework (CWR) by van Lambalgen and Stenning. We defined what can be considered a monotone game process. More importantly we show how an agent can make a rational decision change through a non-monotonic reasoning process that brings social rules into the game process. Experimentally, we ran a modified ultimatum game. The results confirms that people's general, even ranked, attitudes towards some social matters are not necessarily equivalent to their preferences of the same kind of matter during the game. We also discover many instances of non-monotonic reasoning patterns used during the game. Keywords: Game theory, Non-monotonic logic, Rationality