In this pandemic period, the education system all around the World is in shackles. Students cannot go to their institution, while teachers struggle to evaluate them. Managements of the institutions are resorting to whatever methods might keep them afloat and carry on the matters.
Many institutions use proctor software during online tests conducted for the students. This proctor software uses a facial detection algorithm to check for any discrepancy during the tests. Because of this proctor software, students often complain about genuine concerns such as privacy and usage of the video footage taken.
Proctorio is an online exam proctoring software used to check whether the students are cheating in tests. Many institutions across the globe use it to check up on students. Proctorio uses a facial detection algorithm by using the web camera footage from the student.
Proctorio recently became controversial for not detecting the faces of persons of color. Many students have raised allegations that the software is better at detecting faces other than Black people. The spokesperson for Proctorio rejected this allegation and stated that the public did not fully capture how their technology worked.
Many students accused Proctorio of not working correctly. Black students also accused Proctorio of being unfairly discriminated against since other races did not feel as worried as Black people.
This has come to light as a result of extensive research about Proctorio. Akash Satheesan, a researcher and a student using Proctorio, has stated that the claims were valid. Proctorio failed to detect the faces of Black people the most. He also analyzed the code behind the software and found some shocking truth.
Proctorio’s software seems to have parts of code from an open-source library, OpenCV. It is an open-source computer vision software library. The code in Proctorio and the one in OpenCV were similar. As a result, they both reacted in the same manner when they were given some samples from the FairFaces Dataset. It collects images of people of various races and ethnicities to represent diverse people’s face samples. Indirectly, the Spokesperson from Proctorio also admitted that it uses code from OpenCV under license.
Akash Satheesan found out that Proctorio failed to detect black faces more than half the time, at a staggering 57%. It also was unable to detect 41% Middle Eastern faces, 40 % white faces, 37% East Asian faces, 35% Southeast Asian or Indian faces, and 33% Latinx.
This finding indeed indicates that Proctorio is simply useless to conduct tests in an academic environment. Students, especially black people, complain about how the software unfairly does not detect their face and sometimes lock them out from taking the test.
This study has also found out that even if Black faces are in the exact lighting and position as white faces, Proctorio does not detect them most of the time. Akash also added that “They use biased algorithms, they add stress to a stressful process … during a stressful time, they dehumanize students”.
This is not the first case of racial bias in face detecting code or any other code, for that matter. This software should be tuned to accommodate people’s needs and interests from diverse backgrounds and should not unfairly attack and discourage them at any cost.
Image courtesy of August de Richelieu/Pexels
Just less than a year after iPhone 12 was launched and already, Apple has us…