When most people hear “escape room,” they think of a fun night solving puzzles with friends.
In our work at the Aristotle University of Thessaloniki (AUTH), we asked a different question:
What if an escape room could become a structured laboratory for studying team behaviour, powered by deep reinforcement learning and psychometrics?
That question led to our paper, “Escape Room Experience for Team Building Through Gamification Using Deep Reinforcement Learning,” presented at the 15th European Conference on Games Based Learning, ECGBL 2021.
The conference took place online on September 23–24, 2021, making September 25 a suitable date for a post-event research reflection.publica.fraunhofer+1
Traditional team-building activities often rely on workshops, trust exercises, or one-off activities. These experiences can be valuable, but they are not always easy to measure objectively.
Escape rooms already contain many of the conditions that matter for teamwork:
Our contribution was to take this format online, instrument the environment, and connect it with artificial intelligence and behavioural analysis.
Instead of treating the escape room only as entertainment, we used it as a structured environment for observing how teams communicate, coordinate, and solve problems.
The online escape-room experience combined puzzles, Raven-inspired intelligence tests, and team communication.
Players had to work together to interpret clues, solve problems, share information, and coordinate their actions. The environment created a common task while allowing individual behaviours to emerge.
The system could record detailed behavioural traces, including:
The aim was not simply to determine whether a team escaped. It was to understand how the team escaped.
Teamwork is difficult to measure through a single questionnaire or one short activity.
A team may describe itself as collaborative, but its behaviour may change when information is incomplete, time is limited, or the first strategy fails.
The escape-room environment creates a repeatable context in which collaboration becomes observable.
For example, one team may distribute information quickly and divide tasks efficiently. Another may depend heavily on one participant. A third may explore many options but struggle to agree on a final decision.
These patterns can reveal different teamwork styles.
The same team may also behave differently under different conditions. A calm environment may produce one pattern, while time pressure or increased complexity may produce another.
The research also introduced Deep Reinforcement Learning (DRL) agents into the escape-room environment.
These agents learn to solve puzzles and collaborate using reward functions that encourage specific behaviours. Through repeated interaction with the environment, they learn which actions help them complete tasks and achieve shared objectives.
The agents can be designed to explore different strategies and cooperation patterns. This creates a controlled way to compare:
The objective is not to claim that AI teams behave exactly like human teams. Rather, they provide computational models that help us study possible behavioural patterns in a repeatable environment.
Comparing human teams and AI teams creates a new lens on collaboration.
Human teams bring emotion, experience, intuition, social interpretation, and improvisation. AI agents provide consistency, scalability, and the ability to repeat controlled scenarios many times.
Together, these perspectives can help researchers investigate:
These comparisons can inform both AI research and the design of serious games for teamwork assessment.
A useful team-building experience should do more than create a memorable moment.
It should help participants reflect on how they communicate, make decisions, share responsibility, and respond when a plan fails.
An instrumented digital escape room can provide a shared experience and a structured basis for discussion. After the game, teams can examine questions such as:
This makes the escape room a potential tool for team development rather than simply a recreational activity.
This research helped establish the foundations for SeriousMind’s work with AI-powered serious games, behavioural modelling, and team assessment.
The core idea remains relevant across products such as MindEscape and RealEscape:
MindEscape focuses primarily on individual personality and behavioural assessment, while RealEscape applies similar principles to multiplayer teamwork, communication, and leadership.
The research journey from online escape rooms to commercial serious games demonstrates how academic work can evolve into practical tools for HR, education, team building, and organisational development.
Escape rooms are engaging because they create a clear goal and immediate challenges.
Research makes them valuable by asking what happens inside the experience.
When gameplay is instrumented carefully, every interaction can become evidence about how people approach problems and work with others. That evidence must still be interpreted responsibly, but it can make teamwork more visible than a traditional discussion alone.
The escape room becomes more than a game night.
It becomes a behavioural laboratory.
© 2026 SeriousMind. All Rights Reserved. Privacy Policy