ART 80F - MODULE 1 LECTURE CONTENT

 

Identifying key components of algorithmic bias

  • Bias as a core component of a system or algorithm - part of the process fails or produces inaccurate results disproportionately across different user groups.

  • Flawed input technologies that contribute to more biased outcomes - can be part of coded algorithmic systems OR social algorithmic systems, such as low resolution security camera footage, or the digital cameras used to take Drivers License, Passport or Visa or arrest photos. (50% of US Adults in FBI Facial Recognition Database, 80% of those photos from things like DMV and Passport Office records)

  • Using biased data sets / inputs for machine learning and/or accuracy testing.

  • No oversight or regulation in testing outcomes or accuracy claims, no transparency of subgroup statistics.

bias01.png