A 2012 study published by the IEEE found that when used on photos of men, whites, or the middle aged, the best commercial systems matched faces successfully about 94.5 percent of the time, but their success rates were lower for women (at 89.5 percent), Blacks (88.7 percent), and the young (91.7 percent).Feb 2, 2022.
In a study last year conducted by the National Institute for Standards and Technology (NIST), the government agency found that the majority of facial recognition algorithms were far more likely to misidentify racial minorities than whites, with Asians, Blacks and Native Americans particularly at risk..
So how accurate is facial recognition, anyway? Broadly speaking, facial recognition technology is 99 percent accurate..
Examples of AI bias in real life
Healthcare—Underrepresented data of women or minority groups can skew predictive AI algorithms.
For example, computer-aided diagnosis (CAD) systems have been found to return lower accuracy results for black patients than white patients..
Biometric bias is the result of two components — inputting biased data into the system and biased analysis of the data.
Algorithms are trained using datasets.
When datasets skew towards certain characteristics, the machine learning model then focuses more on that characteristic..
The own-race bias (ORB; also known as the other-race effect and cross-race effect) refers to the phenomenon by which own-race faces are better recognized than faces of another race (e.g.
Meissner and Brigham, 2001; Sporer, 2001; Wright et al., 2003; Walker and Hewstone, 2006a; Goldinger et al., 2009)..
The cross-race effect (sometimes called cross-race bias, other-race bias, own-race bias or other-race effect) is the tendency to more easily recognize faces that belong to one's own racial group..
Facial recognition systems misidentify Black faces at a high rate.
Facial recognition is less accurate in identifying people with darker skin tones—especially women.
This can result in the misidentification of Black protesters or false positive matches in image databases.Mar 21, 2023.
Facial recognition systems misidentify Black faces at a high rate.
Facial recognition is less accurate in identifying people with darker skin tones—especially women.
This can result in the misidentification of Black protesters or false positive matches in image databases.Mar 21, 2023.
Data breaches involving facial recognition data increase the potential for identity theft, stalking, and harassment because, unlike passwords and credit card information, faces cannot easily be changed..
The first is data.
To be accurate, machine learning needs a big dataset.
The more data you put in, the more accuracy you get out.
Since minorities are by definition less well represented in the population than the majority, a lack of data may explain much of the “bias” in face recognition systems.Feb 2, 2022.
