Evolving personas for player decision modeling

Abstract

Abstract-This paper explores how evolved game playing agents can be used to represent a priori defined archetypical ways of playing a test-bed game, as procedural personas. The end goal of such procedural personas is substituting players when authoring game content manually, procedurally, or both (in a mixed-initiative setting). Building on previous work, we compare the performance of newly evolved agents to agents trained via Q-learning as well as a number of baseline agents. Comparisons are performed on the grounds of game playing ability, generalizability, and conformity among agents. Finally, all agents' decision making styles are matched to the decision making styles of human players in order to investigate whether the different methods can yield agents who mimic or differ from human decision making in similar ways. The experiments performed in this paper conclude that agents developed from a priori defined objectives can express human decision making styles and that they are more generalizable and versatile than Q-learning and hand-crafted agents.

Procedural Personas as Critics for Dungeon Generation

“…": if the persona's chosen action matched the human's, the agreement ratio increased. Summarizing the results of [6], most players had the highest agreement ratio with the Treasure Collector persona, while a smaller number of players matched the Monster Killer persona.…”

“…Using the MiniDungeons game as a testbed, 38 participants played all 10 levels of Fig. 1, as covered in detail in [6]: a few participants managed to collect all the treasures in every level (see Fig. 2a), while others rushed to the exit (see Fig.…”

“…For the purposes of this paper, generated dungeons will be evaluated by personas evolved on all dungeons of Fig. 1 as per [6]. The controller for each persona is a combination of 7 linear perceptrons, with inputs being the hero's HP and distance to different elements (e.g.…”

“…The values of C m , C t and C r are taken directly from the fitness function which guided the evolution of each persona's controller 1 ; the persona was evolved on 10 authored levels (see Fig. 1) and was evaluated on how it represents an archetypical decision making style (a Monster Killer that kills most monsters, a Treasure Collector that collects most treasure) [6]. Inversely, the scores of eq.…”

“…In this paper, the personas have been evolved on a set of authored dungeons, according to different fitnesses that match archetypical decisions-making priorities. The testbed game, named MiniDungeons, is a simple turnbased roguelike game; the game has been tested by human users and a close match between procedural persona playstyle and human playstyle was found [6].…”