The hidden bias behind the healthcare industry’s AI algorithms
As human beings, we all have the right to receive treatment for our illnesses and injuries by going to the doctor. Sounds like a standard and straight-forward procedure in the eyes of society. However, what if I tell you that not all treatments are considered equally, not because of the kind of condition that one is in at the time of admittance, but rather it being based on the technology we have available in our healthcare system. The technology itself is not what causes the prioritization of treatment for certain groups, but it has to do with biased algorithms present in the machines utilized to identify individuals and their health needs. Indeed, with the use of artificial intelligence (A.I.) becoming more prominent in today’s world, there are more ways than ever for machines to mimic human intelligence and perform tasks more efficiently. Machine learning as a concept is a brilliant idea in hindsight. However, the problem behind machine learning is that along with the knowledge brought by humans, they are also designed to take in personal biases from their human creators. It is widely believed that AI has high capabilities of logical, objective and reasonable decision-making within their algorithms. In the case of the healthcare system, there is scientific research that AI favors the treatment of lighter-skinned individuals over those who are dark-skinned. For example, in a study conducted by UC Berkeley researchers that was published in Science Magazine, there were about 50,000 retrieved records from a large nearby hospital. Records were measured with a score indicating the risk each patient had for consideration of treatment. Results showed that white patients had higher risk scores than black patients with similar conditions, meaning that white patients were prioritized for any extra treatment, therefore concluding that there is not only bias, but it digs a deeper hole in unveiling the realities of racial injustice in today’s society.
Bias is typically not found through measuring metrics of race alone as algorithms have access to data pertaining to healthcare costs, which is a socioeconomic factor affected by race. Considering that black families have less wealth than their white counterparts, black families have incurred less healthcare costs in part to social programs like Medicaid covering it. Due to the increase in accessibility for black families, the algorithm interprets risk for white families as much higher rather than the other way around. Thus meaning higher risk translates into consideration for treatment. These sorts of algorithms can complicate the distribution of resources towards populations at-risk for contracting diseases like anemia, high blood pressure, heart disease and diabetes. Algorithms can essentially increase mortality risk for people of color coming from low-income backgrounds in a healthcare setting rife with systemically racist practices.
The algorithms put out by AI also tends to discriminate against women and those from different cultural backgrounds when it comes to job opportunities in healthcare. Machine learning tools were designed to favor men, because the tech industry is overwhelmingly composed of white men. As AI is designed to soak in knowledge from their human creators, they are also designed to take in any biases they are programmed to have. Women are a group that the AI considers to be ‘less preferable’, showing that it stems from deeply ingrained expectations of societal roles and social injustices. Even health data is impacted by the dominance of white men. With a lack of diversity in researching health implications on various ethnicities, there is so much bias towards publicizing data based on those with European ancestry. In fact, a 2014 study that tracked cancer mortality pointed to very little diversity in research subjects as the reason why African Americans are more likely to die from lung cancer over a 20-year span. This is concerning as doctors have to use a European-centered metric for determining the health of their black and brown patients, which can cause more long-term implications in their communities. These algorithms are in need of reprogramming.
There are many reasons as to why there are biases present in technology and in almost every industry in today’s world. It can be difficult to just erase preferences in machines made for adapting human language into action. Bias detection is not instantly recognized by AI algorithms, it can be initially difficult to notice any sort of bias in the system and there is a lack of social context in the world of computer science.
However, just as there are many questions to navigate this concept, there are answers to resolving this issue with spreading social awareness. This idea helps get at the root of the issue by educating fellow humans before developing technologies utilized for machine learning. Even then, it can reduce the amount of time it takes to notice these biases in the machine’s code and create resolve. In the world of computer science, the implementation of emotional concepts into a logical matter of thinking is a difficult process in and of itself. Considering that computer science is based on logic, there is always a need to counteract that with an emotional response. People who have specializations in the arts and humanities fields should have a say in implementing syntax that is inclusive of all identities for AI. Their background can determine how to best phrase that language and mitigate the possibility of bias in the system. Another solution would be to encourage computer science literacy in communities of color. These communities would benefit the most in having future leaders who are aware of their multicultural backgrounds and bring radical change to a undiverse field. Starting over from scratch would correct the wrongdoings of AI, but there would have to be a shift of including research subjects from diverse backgrounds to accurately study health effects. This is the first step towards dismantling an extremely biased healthcare system through tech.
As a young college student, I grew up in the age of technology and am always seeking new opportunities to broaden my skill set in understanding its implementation in different fields. I understand how it is like to live in one of the poorest districts in the country and witness the diminished quality of life for my black and brown neighbors. Very few would think tech has little influence in the medical field, but as I am writing this, there are thousands upon thousands of patients of color in dire need for essential care because of flawed AI programming. My passion for information systems runs high and I would love to use that knowledge for designing ways to prevent bias in the future. From my experience being an intern at Radical Health, a community-based organization utilizing tech for advancing health equity, I got a deeper understanding of how to create dialogue amongst community members. I was introduced to their proprietary software called Radical Relay designed to answer any burning questions about ANY health-related concerns. Radical Health was an empowering platform for me to be in a role where I can directly impact my community through organizing events that encouraged the sharing of love, perspective and transparency. Overall, it was truly a spectacular first exposure to integrating community health and technology into my day-to-day life in the midst of a historical pandemic.
Through eliminating this system of inequality in our healthcare system, we can guarantee impartiality on how hospitals prioritize patients, no matter the color of their skin. We can guarantee the improvement of research by raising voices on social awareness. We can tailor every person’s health based on their ethnic background. We can create opportunities for women and communities of color alike. I am emphasizing this as a person of color born and raised in the Bronx, whilst coming from a low-income, Hispanic background.
We can collectively save lives!
