Is Dark Icon racist?

I heard about (literally, audiobooks ftw!) this test a few months ago, but it immediately dropped out of my mind until I randomly stumbled across it. Basically it judges your ability to judge armed vs. unarmed targets and differentiates your score based on race.

Pictures flash on the screen. Some of the pictures have people in them. Some of those people are holding guns, and some of those people are holding cellphones. Some of those people are black, and some are white. You press a button to differentiate the guns from the phones, or, to quote the test exactly: to shoot or holster your weapon.

My score:

Game Over
Your Score: 535
Average reaction time:
Black Armed:550.92ms
Black Unarmed:645.28ms
White Armed:555.84ms
White Unarmed:652.76ms

I have no idea what this means. No, really.

Apparently I am quicker to judge black people than I am whites. In other words, it takes me longer to decide whether a white person is armed or not… but only slightly. If a black person was armed, I reacted in 551ms. If a black person was unarmed, I reacted in 645ms. For whites, however, those same times are 556 and 653ms, respectively. It takes me 5ms longer to decide if an armed white person is armed vs. an armed black person. Again… no idea if this is a good or bad reaction time, nor was there any information to let me know if the time differences were significant or not. How much of a time difference means you’re a racist? No idea.

But it was a fun test.

And, in the interest of full disclosure, I DID shoot about five unarmed people just for the hell of it.


  1. nate, April 30, 2008:

    Don’t want to give my score. Let’s just say you don’t want to be suprising me on 3-hours of sleep. I tended to shoot anyone holding anything.

  2. WeREwOLf, April 30, 2008:

    I had scores nearly identical to DI’s. And I noticed something almost immediately, after only about a dozen or so images: I wasn’t even wasn’t even noticing whether the person was black or white — I was focusing intently on whatever was in their hand. Couldn’t tell ya what color the guy in the previous image was. Given the close similarity of scores, I’m guessing you took a similar approach.

    What’s most statistically relevant here is not paltry 5 or 10 ms between noticing whether the suspect was black or white, but the much larger 100 extra ms in making sure they weren’t ARMED. I could pick out the guns (almost) instantly; anything else — wallet, cell phone, whatever — took me a tenth of a second longer to make sure.

    FWIW, I shot one unarmed suspect… but I got shot TWICE. Next time I play this game I’m wearing a vest. Or just shooting everybody.

  3. DarkIcon, April 30, 2008:

    I’m pretty sure I got hit once or twice in the beginning, before I realized there WAS a time limit on each picture.
    It is rather odd that deciding something is NOT a gun takes longer than deciding that something IS a gun. I don’t know what that means.

    I tended not to consciously notice race either, BUT I think the point is that, subconsciously you DO notice, and that affects your reaction time, even when you swear that it doesn’t. That’s what the scientists said in their paper. I read (heard) about it in a book about unconscious decision-making. I can’t remember the title, but it might have been “Blink”.

    Again, without some fancy statistical-type numbers (standard deviations, means, etc.) I can’t tell what our scores mean.

  4. WizardofOz, May 9, 2008:

    The results are pure CRAP. Either that, or the person that wrote the “game” is a racist (in which case the results are crap).

    I wrote an application that would send a “/” keystroke (the keystroke for shooting in the game) on a 100ms interval. Yet here are my results:

    Your Score: -520
    Average reaction time:
    Black Armed:141.36ms
    Black Unarmed:114.04ms
    White Armed:176.52ms
    White Unarmed:152ms

    So somehow, a visual basic application with ZERO visibility into what “color” an image was, and was sending keystrokes at a specific interval, somehow “chose” to run just a hair quicker when the image displayed a black person than when it displayed a white person? Somehow I don’t think so.

    The programmer did something with the timing of black images over white images in a deliberate effort to mess with the results.

  5. Kragon, May 10, 2008:

    Damn Oz, that is really fucked up. That programmer needs to be dragged out in to a street or back alley and shot.

  6. DarkIcon, May 10, 2008:


Leave a comment

You must be logged in to post a comment.