Algorithmic bias and coverage error
I went to a lecture today about these. Let me begin with definitions, then show how they're related.
Coverage error: The exclusion of people of people from a survey.
Algorithmic bias: Algorithmic outcomes systematically biased for a group.
Coverage error causes bias because data for eligible units not included is left out of the calculations. Coverage error can be unintentional or, as in the case of the proposed U.S. Census citizenship question, intentional. In the former case, bias estimates can often be made.
Algorithmic bias arises as a result of the same mechanism, at least in part. Facial recognition software frequently misclassifies minorities because it trained on too few minority faces. This is really a coverage error.
As an aside, I saw a demonstration in which a wolf was properly identified. When the background snow was removed, the wolf was no longer identified. The algorithm was not trained on wolves without snow.
Posted using Partiko Android
Thank you so much for being an awesome Partiko user! We have just given you a free upvote!
The more Partiko Points you have, the more likely you will get a free upvote from us! You can earn 30 Partiko Points for each post made using Partiko, and you can make 10 Points per comment.
One easy way to earn Partiko Point fast is to look at posts under the #introduceyourself tag and welcome new Steem users by commenting under their posts using Partiko!
If you have questions, don't feel hesitant to reach out to us by sending us a Partiko Message, or leaving a comment under our post!
Congratulations @rufusfirefly! You received a personal award!
You can view your badges on your Steem Board and compare to others on the Steem Ranking
Do not miss the last post from @steemitboard: