Saturday, April 5, 2008

Implicit Biases

As games designers we need to know about the current emotional and cognitive states of our players. The director of a film relies on the fact the viewer is forced to watch each moment in an exact sequence. This means the director have a pretty good idea of what is going on in the hearts and minds of the audience. They use this knowledge to enhance or surprise the viewer and create the desired emotional roller-coaster ride. You want to do this to your player right? Well, games pose a different challenge, the player can be doing almost anything at anytime in almost any order. How can the designer know when to provide the correct stimuli at the precise moment in time? What we as game designers need is real-time input of what the player is doing and feeling. The doing part is already taken care of but the feeling part is elusive. Recently companies like Emotiv have announced consumer priced EEG machines that can read some of the emotions in the players. This is great but some things can not be read by a machine (yet). In this blog I will talk about a technique some genius Harvard professors have created to read some pretty complex feelings in people and how this technique can be used in our games.

First a scenario: Act I - Great warrior battles hordes of bad guys. Act II - Great warrior battles hordes of bad guys. Act III - Great warrior battles hordes of bad guys and wins.

Sound interesting? No, but that’s what many games come down to. Here’s a modified scenario: Act 1 - Great warrior battles hordes of bad guys. Act II - Great warrior finds out bad guys are the good guys. Act III - Great warrior must battle his once comrades and wins.

This second example is only slightly more interesting but it will do for our purposes. In the world of movies this second scenario would play out reasonably well with most of the audience. The writer, director, and editor would provide the correct setups (like the Great warrior perceives the bad guys killing the one he loves) and surprises (Great warrior finds out the one he loves is alive and working for the ‘bad guys’). The problem in the movie world and everywhere else is while designers of the film are pretty certain of the feelings of the audience they still don’t really know. The worst part is that it will be different for everyone. This forces the designers to ‘hit the audience over the head’ which pisses off some people will still missing others. Its not an exact science. Would it be great if you did know what the audience felt of the bad guys so that at the right moment you could surprise them? If you answer no you are reading the wrong blog. If you answer yes then he is how you do it - The Implicit Associations Test. As mentioned a team of Harvard professors came up with a way to measure the unconscious biases of people toward some entity. In the real IAT test the entities are African Americans or women in science (the tests are largely political). But nothing is stopping someone from measuring biases against say the evil clan of - insert your game’s bad guys here. How the test works is brilliant - it measures your reaction time in categorizing items. You are faster at categorizing items when your associations agree then when they do not agree. I don’t have enough time to try and fully explain the test but an exercise is worth a thousand words. I strongly suggest you now go to
https://implicit.harvard.edu/implicit/demo/ and take one of the tests. I recommend taking the race test as 80% of the participants found out they have an implicit association with white Europeans and good; and with African Americans and bad. You may found out something about yourself you didn’t know. When you are finished come back and I’ll give you hints for making this test work in your game.

Done? Did you see how the test is done? Pretty slick huh? I got a sick feeling in my stomach when it was taking me slightly longer to categorize good with African American. Even half of African Americans did the same on the test. Anyway, done with politics and back to game design.

What you need to do is translate the mechanics of that test to a video game scenario. The key, of course, is not stopping the game and making them take a test. Imagine instead setting up a timed action scene where the player must navigate through a maze, constantly choosing between the left or the right. Or how about a scene where the player must quickly shoot as good or bad guys pop out. What needs to be added is some kind of association. You need an external symbol that the player automatically recognizes as good or bad (internal imagery can work too but you need to setup strong symbols). Once you have the symbols you need to present them in pairs with either the bad guys themselves or yet another symbol representing the bad guys. Keep in mind that you need to do things like control for speed of categorizing and left and right biases.

This blog is getting way too long so unhappily I am going to cut it short. If anyone actually cares enough maybe I’ll take the time to write another blog with more detail and even some of the math. I won’t lie, setting up this test in your game could be a lot of work, but the playoff would be fantastic!

No comments: