Agents Behavior and Preferences Characterization in Civilization IV

Abstract

Player Modeling is becoming an important feature in Digital Games. It basically consists in understanding and modeling the player characteristics and behaviors during the game and has been mainly used to improve the games artificial intelligence, making games more adaptable to different players. In this paper, we try to characterize the preference of the players using a novel approach in games: we use mathematical regressions to characterize players behavior, looking for functions that best fit these behaviors. Using AI controlled players in Civilization IV as a testbed, this characterization is performed by extracting game data (score and resources, for example) at the end of each turn and generating functions that characterize the data evolution during the game. We were able to obtain models that distinguish the agents preferences showing the effectiveness of this approach.

A binary classification approach for automatic preference modeling of virtual agents in Civilization IV

“…Since a match may have, at most, 460 turns, we can summarize a whole match with 460 observations. In [7], we were able to show that the evolution of some of these indicators is able to represent different preferences when analyzed along several matches.…”

“…Finally, in a previous work [7], we characterized some Civilization IV agents behaviors with linear regressions. We showed that some agents' game indicators, when analyzed individually, can be linearly separable between agents with different preferences, and drew conclusions about preferences in relation to the time in the game (passage of turns), other preferences, and the gameplay.…”

“…Once the game platform that is going to be used is defined, a study of the selected data may be useful to assure that the assumptions of the data relevance are correct. An example of this type of study is [7], where we analyzed the discriminative capacity of some Civilization IV indicators for the preference modeling task.…”

“…A complete list, also including features' range, is presented in [8]. Notice that the decision to use the same features used in [6] and [8] was based on the work in [7], where we showed that some indicators are able to distinguish different behaviors. We made regressions on these indicators showing that the curves obtained are statistically different for different preferences when analyzed individually.…”

“…The authors also suggest that preference models are seldom used due to the fact that modeling this concept is much harder than modeling actions, since preferences are not clearly measurable. One approach we can use to model agent's preferences is to observe its actions as an indirect evidence of its preferences, as done in [6], [7] and [8], for example.…”