As part of an individual specialization, I am working on an individual project in collaboration with the graduation project of one of the “Animation Director”-students at the Danish Filmschool.
The project at the Filmschool spans two semesters (Fall 2021 and Spring 2022), and is an AI-driven horror game with a heavy emphasis on narrative and empathy.
The game is named “Church in the Wild”, themed around a nightmarish virtual reality controlled by an Artificial Intelligence, and developed in Unreal Engine 4 for the PC.
The project at the Filmschool is a collaboration between students from the Filmschool, several differnet Universities, and other higher educations, where each student brigns their own expertise to create and enhance parts of the project.
In the individual project I am working on, we (the two programmers) are researching and developing Machine Learning models for modelling the level of fear that a player experiences in a horror game, and use this player model in an AI-director that can adaptively direct the pacing and events of the game.

We have created a prototype of a horror game for gathering data, which is used to train the Machine Learning models that power the player fear modelling.
As for data, we are gathering telemetry about what the player is doing in the prototype game, we are recording the screen while players are playing through the prototype, and we are gathering biometrics data through a Bitalino HeartBIT kit.
The biometrics we are collecting is skin conductance (measured through an EDA-sensor) and heart-rate data (measured through a PPG-sensor).

For the machine learning we are running the data through a suite of simple Machine Learning models (like Support-Vector Machines, Decisition Trees, etc.), both for classification and regression, and we are planning to also try some more advanced ML models like Convolutional Networks and “Learning under Privileged Information”, where a teacher and student Neural Networks are trained, with the teacher having access to types of data that the student does not, and then the student is trained with input from the teacher network.

– Jonathan E. Mogensen

Church in the Wild:
– Alexander Tange
– Michał Pikulski
– Stella Vaka
– Mads Hølmer
– Niels Faaborg
– Vaeis Omar
– Ronja Tange
– Troels Jørgensen
– Magnus Laursen
– Sophie Hahn