By: Rahul Shinde, RIG Intern Researcher


Until the advent of Artificial Intelligence, the methods of user verification for conventional and digital operations were manual. In the future, biometrics will be widely used to overcome these issues. An identification process occurs when a system wants to establish who the user is. Verification is the process of using that biometric information to determine if there is any other information such as face, fingerprint etc. that are associated with the user. In a nutshell, the purpose of authentication is to determine if the identity the user claims is correct and authorized to access the services and data they request.

We unlock our phones with a look and wonder how social media apps know to tag us in that photo. However, the technology underpinning these capabilities, facial recognition, is more than an algorithm. The application of AI and machine learning to facial recognition is one of the most promising areas. As a biometric authentication technique, 3D facial recognition is much faster than ever before, and 3D cameras are capable of capturing much more detail(s) about a face than two-dimensional cameras ever could. AI, which is widely used in augmented reality solutions, make computer facial identification considerably easier because of better accuracy by evaluating facial traits and comparing them against a database.

The below diagram demonstrates the basic flowchart of facial recognition.

However, like any other technology, there are potential drawbacks to using it, such as threats to privacy, violations of ethics and personal freedoms, potential data theft and other crimes. Regardless of far reaching reception, face recognition was as of late restricted for use by police and nearby offices in a few cities, including Boston and San Francisco. Recent studies have shown that Tech Giants like Amazon, IBM, Microsoft, Facebook and Google who have been using Facial Recognition Algorithms, had some considerable biases in those algorithms. It turned out these algorithms performed better on male faces than female faces, and significantly better on lighter faces than darker faces. That isn’t to imply that face recognition hasn’t experienced issues with women’s and minorities’ faces. When the technology was newer a decade ago, it was generally less accurate when it came to recognizing minorities and women.

One more key wellspring of racial segregation in face acknowledgment lies in its usage. In eighteenth century New York, “light regulations” expected oppressed individuals to convey lamps into the evening to be freely apparent. Advocates dread that regardless of whether face acknowledgment calculations are made fair, the innovations could be applied with a similar soul, excessively impacting the Black community in accordance with existing bigoted examples of policing. Furthermore, face acknowledgment might possibly target other minimized populations like undocumented immigrants by ICE.

There is no doubt that some industries and governments are trying their best to identify flaws in their Facial Recognition Algorithms and it will definitely improve in the future, however it will never be completely accurate. Face recognition is still a strong technology with far-reaching ramifications in criminal justice and everyday life. When AI results cannot be extrapolated widely, bias occurs. Bias can be introduced by how data is obtained, how algorithms are designed, and how AI outputs are interpreted. We often think of bias as resulting from preferences or exclusions in training data, but bias can also be introduced by how data is obtained, how algorithms are designed, and how AI outputs are interpreted. Hence, tackling racial prejudice in facial recognition and its applications is essential for making these algorithms more equal and impactful.









2. 3. 4. 5.

6: Netflix: Coded Bias