Affect Modeling via Privileged Information

This study on affective computing assumes that privileged information (available in lab settings) can aid models of affect to perform well in real-world settings where privileged information is absent. Two games (Survival Shooter, Space-Maze) are tested. The games' priviledged information (teacher model) is game telemetry data, heart rate, electrodermal activity. The games' non-privileged information (student model) is the pixels of the gameplay footage. In a followup study, webcam videos were used as non-privileged information while audiovisual and physiology metrics were used as privileged information.

Relevant Publications

  • Konstantinos Makantasis, David Melhart, Antonios Liapis and Georgios N. Yannakakis: "Privileged Information for Modeling Affect In The Wild," in Proceedings of the IEEE International Conference on Affective Computing and Intelligent Interaction, 2021. PDF BibTex

  • Konstantinos Makantasis, Kosmas Pinitas, Antonios Liapis and Georgios N. Yannakakis: "From the Lab to the Wild: Affect Modeling via Privileged Information," in IEEE Transactions on Affective Computing 15(2), 2024. PDF BibTex