Go Go Games

Science

Go Go Games™ – Behind the Scenes

We have worked hard to give Go Go Games™ a light and playful feel but there is a great deal of work going on behind the scenes!  Each set of options is selected by an algorithm designed to gradually increase the demand on a player’s ability to detect multiple features of the items she sees.

In each game in the Go Go Games™ suite, the player must repeatedly make a correct choice given several options.  For example, he might navigate a car down one of three roads marked with the following signs:

Three options which can be discriminated by noticing only one feature

Figure 1

In this case exactly one feature – color – separates each wrong answer from the correct answer, forcing the player to notice this specific feature of color in order make a correct selection.

We can make this task more challenging by forcing the player to make a distinction based on two different features as shown in Figure 2:

Three choices which can only be discriminated by noticing two different features

Figure 2

Here the player must notice not only that the truck he is driving is purple, but also that it is a truck and not a car.  While this second scenario may still seem trivial to a typically-developing adult, research has shown again and again that this is a surprisingly difficult task for many children with autism.

There are two components of the way we determine task difficulty:

  1. The “distance” between the correct choice and the wrong choices.  “Distance” is a count of the number of features that differ between two items.  So for example, the two items in Figure 3 have a distance of 1 – because they differ by one feature, the presence or absence of a flag.
    Two choices with a distance of 1

    Figure 3

    On the other hand, the two items in Figure 4 differ by a distance of 5 because they differ in five ways:

    Two options with a distance of 5

    Figure 4

    1. One is a semi-trailer truck while the other is a pickup truck
    2. One is blue while the other is pink
    3. One has only one door showing while the other has two doors showing
    4. One has a dog as a passenger while the other has no passengers.
    5. One has a flag and the other does not

    The smaller the distance between two items, the more similar they will be.

  2. The number of choices displayed.  It is not enough to use distance alone as a measure of task difficulty because it is always possible to look at a single feature to distinguish two items that differ, even if their distance is 1. However, the complexity of the decision-making process increases as more choices are displayed.

    With only two options to choose from, we can only require a player notice one feature:

  3. Level 1 of Build-a-Train which only requires a player choose between two options

    Figure 5: Level 1 of Build-a-Train which only requires a player choose between two options

    However, once a third choice is introduced we can require a player notice two specific features – in the example below, the player must notice both the type of animal riding in the train car and the color of the train car:

    Level 6 of Build-a-Train where a player must make choices by detecting two features simultaneously

    Figure 6: Level 6 of Build-a-Train where a player must make choices by detecting two features simultaneously

As we increase the number of options we are able to force players to make a decision based on an increasing number of features. By controlling both the number of options and the distance between options we control the difficulty of each choice the player is asked to make.

The change in the complexity of the game over time is determined by an algorithm that selects the items to present to the user from a pool of over 500 options.

Average “distance” (number of differing features) between an incorrect option and the correct choice based on 10,000 simulated runs of the Build-a-Train

Figure 7: Average “distance” (number of differing features) between an incorrect option and the correct choice based on 10,000 simulated runs of the Build-a-Train

As you can see from the graph above, we attempt to build a saw-tooth curve of difficulty based on level progression, such that the game becomes increasingly difficult over the course of each level, then briefly eases up at the beginning of the next level before increasing in difficulty again.

Of course, the real purpose of manipulating distance is to control the frequency with which a player needs to notice multiple features.  Figure 8 shows how this progresses over the course of the game:

The change in frequency of multiple-feature tasks over the course of playing Build-a-Train based on a simulation of 10,000 trials

Figure 8: The change in frequency of multiple-feature tasks over the course of playing Build-a-Train based on a simulation of 10,000 trials

Again you can see we are creating a saw-tooth curve of difficulty, with a brief respite at the beginning of each new level as a reward to the player for successfully progressing in the game, a common technique in game design.

As our Build-a-Train game presents the player with first two, and later three, options we can never require that he notice more than two features when making a selection.  We push this requirement to three features and later four in our “Roads and Wheels” and “Out of this World” games, using the same techniques to maintain a careful balance of “pleasant frustration” for the player – always challenging but never too hard.