Intelligent Systems Lab

Author(s): G. Liapis, K. Zacharia, K. Rrasa, I. Vlahavas

Keywords: Personality Assessment, Serious Escape Room Game, OCEAN Five, Reinforcement Learning Agents

Tags:

Abstract: Personality traits are essential parts of human behavior anal- ysis and may be applied in various scientific domains and contexts like job screening. Nowadays, several organizations utilize self-assessment methodologies to evaluate people or groups in order to establish pro- ductive teams. Even though extensive study has been done on ques- tionnaires and other self-assessment techniques to profile a candidate or an employee, they are frequently mundane and biased depending on personal beliefs. In this study, we present a serious 3D Escape Room game with the goal of analyzing behaviors that underpin each personal- ity trait, based on the OCEAN Five Personality Traits model. This model encompasses an individual’s behavior on five dimensions: openness, con- scientiousness, extraversion, agreeableness, and neuroticism. We created a number of rooms that have their own characteristics to monitor the player’s gameplay style in order to develop customized models that com- pute the amount of each personality trait. These models use gameplay data generated by deep reinforcement learning agents that emulate hu- man behavior, as a ground truth for each personality trait. However, the final profile is based on a mix of measurements from the various rooms because the behaviors are common and overlapping, but may be analyzed in different settings. Undergraduate and postgraduate students from Greece and Italy took part in our pilot study and the results of the game are correlated to the baseline that weighted questionnaires have set. The results show that there is indeed a correlation between the profiles from the questionnaires and the game.