How AI changed organ donation in the US
September 2019
Quartz published an article surrounding the role of AI in the change of the organ donation process. Find it here.
Artificial intelligence algorithms have greatly facilitated the process of pairing kidney patients to matched donors. One in every eight people receiving a transplant today arrive at their math through this algorithm-based paired exchange. The limitations to this practice are mainly found in the “moral factors that determine who gets a transplant first,” something we as humans are yet to determine how to teach to an algorithm. To provide some history to a similar dilemma, the kidney failure treatment of dialysis, when it was first made available to Americans, required a moral decision to be made as to which one person out of every four should receive the treatment. A variety of factors about each patient’s life had to be considered, including age, gender, marital/family status, emotional and financial stability, and more. There’s a great deal of complexity in this moral decision – the committee appointed to make the decisions was aptly named the “God” committee as a representation, perhaps, of the immense power they held over these potential patient’s lives. This example applies to the current process that algorithms are being used for – kidney transplants. Both at its start and now, there exists more demand for kidney transplants than there are willing donors. The process of manually sorting through blood and tissue data to find donor-patient matches is incredibly tedious, so an alternative needed to be found. An idea to speed up the process was to set up chains of donor-patient kidney exchanges, simply requiring one altruistic donor. The idea of overall process is applicable for other transplants, including lung and partial liver transplants, but algorithms have not yet been developed for them. The weighted criteria used in the current decision-making process include biological suitability, the amount of time spent in the waiting list, age, history of being a donor, and difficulty to match, with the latter two receiving higher priority. There is future possibility for AI to actively make similar, although eventually human-reviewed, decisions using human criteria. The public image of the phrase “artificial intelligence” serves as a representation of the anxiety and discomfort humans have regarding leaving real lives in the hands of technology. Artificial intelligence also has a lack of human understanding and values, and our own general uncertainty about our group behavior as humans (including topics like discrimination by race or gender) is another remaining problem. Regardless, the messy decision of who lives or dies, which one’s accessibility to a transplant certainly has say in, is something that humans, let alone AI, already have an incredibly difficult time executing.