Lessons from an Old Eye Tracker

About one year ago I was on a team working on a game for the EyeGaze, an older eye tracker. It didn't go too well and we ended up switching platforms to the Oculus DK2. The game would go on to become Into The Dark: A Bat's Tale, check out the page if you're interested. I was at GDC 2016 and saw a newer eye tracker coming to the market and thought maybe some of the things I've learned through working on an older one might be of interest to someone. I should note that I didn't try the new one. A friend got one though, so maybe there will be a post about that in the future.

Let’s start off with a bit of a backstory. When I worked on this eye tracker it was part of the Building Virtual Worlds class which is given through Carnegie Mellon's graduate program for entertainment technology. One of the platforms for this class is the EyeGaze. As such, there have been plenty of games built for it. However, the best I have ever seen is Star Gaze. It is brilliant because it uses our natural desire to look at light areas in dark spaces and moving objects.

For the record, I did not make Star Gaze. It is one of the best eye tracking games I have seen and helps illustrate the point I'm making well.

That being said it can suffer from a terrible flaw with eye tracking, poor calibration. Let me give you an example. Say I want to look at the star in the bottom right of this image.


If the eye tracker is not detecting my eyes well, then I can be looking at the star and it will actually think I’m looking where the red circle is.


When this happens the player’s only way of proceeding is to look beyond the star, further into the corner. This is very uncomfortable for the player (trust me on this one, I’ve had to do it too many times).

Having very precise points the player is supposed to look at, forces the player to look beyond objects if calibration is off, and it tends to not want to be on. Which leads us to the lesson.  

Precision Aiming Is Your Enemy


Good! Larger area to look at compensates for possible poor calibration.

Good! Larger area to look at compensates for possible poor calibration.

Not This!

Bad! Precision aiming forces players to account for mistakes in calibration.

Bad! Precision aiming forces players to account for mistakes in calibration.

Try to design around possible calibration issues by having the acceptable eye location be a larger area. You do not have to make the target large, the star is still a great target. Just make the place an eye has to linger not be exactly where the star is. In the example above the white circle would not be visible to the player. I use Star Gaze as an example both because of how elegant and amazing it is and because it is easier to illustrate this lesson with than the game I worked on. This is because the game I worked on was a tunnel navigation game and there weren't points we wanted the player to look at, we just wanted the player to not look at obstacles in the cave. Also, it is entirely possible that eye tracking has gotten to a point where calibration is less of an issue, but if you are developing for eye tracking I would recommend you keep this in mind anyways, just in case.