Issues in Digital Media

        Racial bias in facial recognition software is the algorithmic error in the programming that occurs because the software does not fully take accountability for the diversity of people when scanning someone’s face. This is the result of the development team not taking certain distinguishing features such as race and skin tone when considering the facial features the software will primarily identify in a face (1). This error can be overlooked when developing and testing the software, which results in the software being less effective at correctly identifying subjects who are a part of the area’s minority. For instance, facial recognition software in China would be more accurate at identifying Chinese facial characteristics than any other characteristics. Likewise, facial recognition software in America would be less successful in correctly identifying members of minority groups, such as African-Americans or Latinos, than Caucasians (2). This is not to say that the programmers who develop such software are racists; rather, they aren’t taking into account the software’s ability to differentiate race when scanning facial features (1). The real problem here is that such developers and members of the public aren’t concerned enough with this matter, and aren’t trying to correct the algorithms. In the book Weapons of Math Destruction, the author Cathy O’Neil points out that people are too trusting of algorithms due to their mathematical accuracy. “[Algorithms] replace human processes, but they’re not held to the same standards” (2). Because of this, the companies and teams working on such software are trusted too much, even though they hold little interest in solving these issues.