Paper Reading #18

Biofeedback game design: using direct and indirect physiological control to enhance game interaction

Authors:	Lennart Erik Nacke	University of Saskatchewan, Saskatoon, Saskatchewan, Canada
	Michael Kalyn	University of Saskatchewan, Saskatoon, Saskatchewan, Canada
	Calvin Lough	University of Saskatchewan, Saskatoon, Saskatchewan, Canada
	Regan Lee Mandryk	University of Saskatchewan, Saskatoon, Saskatchewan, Canada

Proceeding

CHI '11 Proceedings of the 2011 annual conference on Human factors in computing systems

Summary

Several physiological sensors are tested for enhancing a video game. Primary interaction is done on a normal Xbox controller while other sensor data such as respiration and heart rate are taken passively.

Hypothesis

1. How do users respond when physiological sensors are used to augment rather than replace game controllers?
2. Which types of physiological sensors (indirect versus direct) work best for which in-game tasks?

Methods
Six sensors are used to gather data from the user.

• Gaze iteration. Cameras track a user's eye movements to determine where they are looking. This is a direct physiological input.
• Electromyography. Sensors are place longitudinally along a muscle group to determine the electrical activation level of the muscle.
• Electrodermal activity. This determines the level of conductance of a user's skin. This insinuates psychological arousal and as such is an indirect input.
• Electrocardiography. This reads the activity of the heart. Even though this is autonomously controlled, it can be mildly influenced consciously. Heart rate is an indirect input. 
• Respiration. When a person breathes in, their chest expands. A strain sensor is placed around the body to determine the level of respiration. This is directly controllable.
• Temperature. Homeostasis dictates internal body temperature is constant, but in this case it is directly controllable by blowing hot air into the sensor.

Together, these sensors were used to control different aspects of a video game in different studies. The size of enemies were varied to make them easier to hit. Since larger enemies are more intimidating, only a shadow of the sprite was made larger. Speed and jumping height was varied as to make the avatar faster or able to jump higher. The weather is changed to make snow appear across the screen. The fallback weapon of a flamethrower has a variable length. Lastly "Medusa's Gaze" is a power-up that allows the user to look at enemies and freeze them by using eye tracking. These variables are changed in two tests as shown below.

Results

The researchers found that users preferred direct control devices, even though they enjoyed using the sensors. They concluded that these sensors are best for use in changing background variables instead of primary interactions. From individual sensors, users preferred ones they could more actively control, such as breath and muscle contraction as opposed to heart rate.

Discussion

I think the SensorLib framework could be used for very rich interactions in games. In any case, most users would probably prefer not having to wear anything though. One thing I was thinking about was the difference between natural mappings and relevant biofeedback mappings.