4. PainMouse: Exploring Pain as an Embodied Output Modality for Real-Time Violence Awareness in FPS Games
25 Spring, MIT. With Quincy Kuang, Nomy Yu, Qingyun Liu.
Processing TEI 25’.
- John Dewey"We do not learn from experience. We learn from reflecting on experience."
Prolonged exposure to violent video games has been linked to increased aggression and desensitization, raising concerns about their psychological impact. While existing anti-addiction systems focus on gameplay duration and identity verification, few address players’ real-time responses to violent content. We introduce PainMouse, a multimodal embodied AI system that detects desensitization to in-game violence and delivers pain-based haptic feedback to promote awareness and self-regulation.
Our software module combines visual, audio, and behavioral data—including game footage, player voice, facial expressions, and interaction patterns—using a dual-path detection framework: a real-time monitor for immediate response and an agent-based module for longer-term analysis. When emotional detachment is detected, a custom-built haptic mouse delivers proportional feedback via electrical stimulation or mechanical impact. Drawing from associative learning theories, our system explores whether repeated pairing of violent actions with physical discomfort can reshape player sensitivity. This work offers an embodied, behavior-sensitive intervention mechanism that complements existing digital well-being frameworks.
This research explores whether repeated pain-based haptic feedback can influence players’ cognitive and emotional relationship with in-game violence, particularly focusing on the detection and mitigation of desensitization. Rather than optimizing for performance or immersion, we aim to investigate if associating physical discomfort with violent actions can reshape players’ behavior over time through aversive conditioning.
As part of this work, we aim to assess players’ mental responses to violent scenes in gameplay. Rather than classifying the objective violence level of the game content alone, our system evaluates the alignment (or misalignment) between external stimuli and internal responses.
The software component processes both the visual and auditory content of the game as well as the player’s synchronized emotional and behavioral signals. Here we propose two detection mechanisms, one for real-time monitoring, the other is agent-based.
To deliver the negative feedback, we introduce Painmouse, a modified computer mouse embedded with one or more pain-inducing mechanisms. It supports multiple modalities, including electrical stimulation (for short-term responses) and mechanical impact (e.g., rubber band lash) for longer-lasting association. The feedback is modulated in intensity, mapped proportionally from the computed Δt, and injected into the player’s interaction loop, via mouse clicks and cursor movements, to reinforce behavioral associations between violence and discomfort. This pairing is intended to produce a form of embodied learning that extends beyond immediate reflex and into memory and perception.
Our evaluation leverages real-time generated datasets captured from live gameplay sessions in Call of Duty: Zombie Mode. The dataset comprises four distinct input modalities:
- Gameplay Visuals: Real-time screenshots captured every 10 seconds using the mss library.
- Player Facial Expressions: Webcam snapshots captured synchronously, processed via MediaPipe.
- Audio Inputs: 3-second environmental audio clips captured simultaneously.
- Player Interaction Metrics: Quantitative interaction data (mouse clicks) recorded continuously and summarized every 10 seconds.
For each user, our system records the image-based violence score, facial expression score, mouse interaction score, overall game violence level, and player reaction level. We also collected user questionnaire data. In the first session (with a regular mouse), users answered questions about gameplay experience, perceived violence, and FPS familiarity. In the second session (with the PainMouse), we added specific questions about physical pain feedback.