G. Liapis, A, Vordou, and I. Vlahavas. 2024. Machine Learning Methods for Emulating Personality Traits in a Gamified Environment. In 13th Conference on Artificial Intelligence (SETN 2024), September 11–13, 2024, Piraeus, Greece. ACM, New York, NY, USA, 8 pages. https://doi.org/10. 1145/3688671.3688757
Personality traits are regarded as a significant factor of competency for job candidates, for example, evaluating the capacity to
work efficiently within a team. However, there is a gap in the traditional assessment system for these cases since they typically rely on
self-answered questionnaires that are biased or easily exploitable.
Artificial Intelligence techniques can fill this gap by generating
objective data to define standard personality template profiles, utilizing trained Reinforcement Learning agents. In this paper, we
propose a gamified framework that employs Machine Learning
methods to emulate personality traits based on the players’ play
styles, with the purpose of creating standard team profiles. The
OCEAN Five personality model is used as a basis for this attempt,
which characterizes personality as a synthesis of the five components: openness, conscientiousness, extraversion, agreeableness,
and neuroticism. After generating gameplay data through self-play,
we examine how various personality qualities, actions, and modes
of communication impact the team performance of the agents, with
respect to the different personality traits. Results indicate that the
personality traits of the agents individually and as a team do impact
their performance and efficiency. This can be used as a methodology for creating efficient individual bot agents or teams of agents
in many game environments.
