Affectively Framework

The Affectively Framework is a suite of Open-AI Gym environments with three different games: a racing game, a platformer, and a first-person shooter. While Affectively follows the Reinforcement Learning feedback loop of Open-AI Gym, it incorporates a human-sourced model of arousal into its observation space. This means that the Affectively Framework can be used to train AI agents that include emotions in their decision-making processes.

Relevant Publications

  • Matthew Barthet, Roberto Gallotta, Ahmed Khalifa, Antonios Liapis, Georgios N. Yannakakis: "Affectively Framework: Towards Human-like Affect-Based Agents," in Proceedings of the International Conference on Affective Computing and Intelligent Interaction Workshops and Demos, 2024. PDF BibTex