Facial recognition and bias protectors have come a very long way since AI and advanced computing systems first originated. But there is still a very large room for improvement because the way that artificial intelligence and most technology learns is either from human input or sources created by humans. When you dive deeper in such sources, bias, racism, and other human imperfections are easily detectable which causes major problems when this technology interacts with humans from different races. For example there is research that proves that LLM’s associate African American English with pre civil right stereotypes. Also companies like Amazon and Microsoft showed error rates of up to 34% for identifying dark-skinned people which led to wrongful arrests.

Now for Asians and Asian Americans specifically there is a specific point where the technology fails. AI fails in incorrect responses called False Match Rates (FMR) and False Non-Match Rates (FNMR) which are what lead to incorrect assumptions and facial recognition. This is partially due to the variety of Asians. Because of the fact that there are so many different variations and kinds of Asians, artificial intelligence cannot keep up with all the unique characteristics and phenotypic variation. This creates a kind of “blind spot” which causes it to classify non-Asians as Asian and some Asians as non-Asian.

This leads to a variety of negative effects in the Asian-American population. Due to the bias, AI tends to create unnecessary barriers and subtle penalties for Asian Americans simply because of facial recognition or the fact that they are Asian. Even in professional environments Asians are experiencing significant bias. For specific job openings, AI screening tools appeared biased against Asian applicants 15% of the time, recommending them at rates significantly lower than leading candidate groups. And all of these different factors coming together are creating less representation, opportunity, and the ability to grow all because of race.