Escape Room Experience for Team Building Through Gamification Using Deep Reinforcement Learning

Georgios Liapis, Aristotelis Lazaridis and Ioannis Vlahavas. “Gamified Escape Room Experience for Simulating Team Building using Deep Reinforcement Learning”. 15th European Conference on Game-Based Learning (ECGBL 2021), Brighton, UK, 23-24 September, 2021.

Gamification, serious game, escape room, team building, machine learning, deep reinforcement learning

Gamification, which is considered to be an efficient practice for learning through play, can be significantly expanded by Artificial Intelligence methods, and particularly Machine Learning. Nowadays, different industries employ a variety of applications based on gamification to create coherent and effective teams, e.g., by assigning roles based on the knowledge, understanding, and relationships between members. In this paper, we explore an online Escape Room experience that incorporates a variety of Raven-inspired intelligence tests and team-members communication, combined with Machine Learning methods. More particularly, we implemented state-of-the-art Deep Reinforcement Learning (Deep RL) agents, which are used for emulating human-like behaviour to navigate and interact with the 3D rooms, as well as to solve the tests. The RL agents simulate behavioural elements based on OCEAN personality traits model, such as openness, conscientiousness, and neuroticism, while also generating a big number of gameplay data. Analysis shows that their particular behavioural patterns have a significant effect on their performance, stability and time required to solve tasks. These findings allowed us to produce new performance metrics for a generic escape room model, which can categorize human play styles according to the OCEAN Five personality trait model. This approach effectively analyses the teams’ behaviour concerning both individual and overall performance.