The business of AI is, to put it simply, booming. Forrester estimates that 100% of organizations will integrate AI into their companies by 2025. The artificial intelligence software market will also reach $37 billion dollars the same year (Forbes).
Unfortunately this remarkable new tool isn’t without fault. Numerous studies show that AI does in fact replicate the biases that exist within society. For example, facial recognition consistently makes errors while analyzing people of color’s faces, especially women of color. Black women’s faces are misidentified over 35% of the time (ACLU). In sharp contrast, White men’s faces are identified correctly nearly 100% of the time (Harvard). Where do these discrepancies emerge? The answer is in data input. Facial recognition software is tested predomiatley on White males. Because of this AI programs struggle to properly identify Black faces, especially female ones.
Since law enforcement has recently begun using AI facial recognition programs to catch criminals, the results of misidentification are far worse than having trouble unlocking your apple phone. One false match can lead to an arrest, long detention and even police brutality (ACLU). The results of one study done on this were particularly shocking. 28 members of congress, democrats, republican, women and men alike, were incorrectly matched to mug shot photos. The congressmen and women wrongly matched were overwhelmingly people of color (ACLU). The root of this bias again stems from data input. Despite cannabis rates being roughly the same for Black and White Americans, Black people are more than 4 times as likely to be arrested. After being arrested, a mug shot is taken and they are inputted into the AI mugshot database. Because the amount of mugshots in the system are overwhelmingly Black, AI is far more likely to register Black faces as criminals (ACLU). Thus, the broken and racially charged justice system further terrorizes communities of color. Not only by the abuse of police but now through the nefarious gleam of identification cameras.
Furthermore, the inequalities of AI not only affect law enforcement, they also affect healthcare. AI models are more likely to prioritize the care of healthy White patients, and less likely to prioritize the care of sicker Black patients. (The Bias of Coding: Netflix Documentary). In a study done by the JAMA Dermatology Network, many disparities were exposed when it came to diagnosing darker skin tones with skin cancer. The AI medical software has been trained to identify cancerous areas mainly on pale skin tones and is far less likely to correctly identify cancer on darker individuals (Technology Advice). This is particularly troubling considering, although Black people are less likely to develop skin cancer than other groups, the health outcomes for those who do are alarming. The survival rate of Black skin cancer patients is just 66%, compared to 90% of White patients (AAMC). Therefore, it is critical that Black patients receive an early and accurate diagnosis.
These algorithms learn from people. They absorb our history as data and then try to make predictions about the future. But in a world with a discriminatory, oppressive past, it is hard to see how AI could create an equitable future if the current situation continues. So, how do we stop the inequalities of society from seeping into AI algorithms? Firstly, data scientists must be educated on what an equitable AI model should look like. Secondly, more diverse data sets need to be imputed to help minimize bias towards certain groups. Thirdly, consumer transparency is a must. People need to know how AI algorithms work. It is imperative that citizens have knowledge regarding the system, that very soon, and even now will make critical decisions about the rest of their life (Forbes).
AI is an extremely powerful tool that, if harnessed correctly, is a force for positive change in the world. However, regulations must be put in place to prevent it from becoming an agent of oppression and racism. If people continue to ignore the holes in the system, our world will become as unjust as ever. The true danger of AI is not a robot revolution, rather, the threat that AI will replicate the sins of our past and further destroy the hope for a one day equal reality.