Racist Medical AI & What Would the Deontologists and Utilitarians Think?
Can we program algorithms with an ethical bias?
Perhaps.
Perhaps not.
I am skeptical.
Nonetheless, it would be interesting if we could! Hence, I will look at a few ethical schools, namely Utilitarianism, Deontology, and Rawlsian Theory, and look at ways they could be implemented.
Let me know what you think!
Act and Rule Utilitarianism
In general, utilitarianism is concerned with consequences to people’s actions that maximize utility or maximize the greatest satisfaction or good attainable. In other words, Utilitarians believe that the purpose of morality is to make life better by increasing the number of good things (such as pleasure and happiness) in the world and decreasing the number of bad things (such as pain and unhappiness).
Deontology
Meanwhile, deontology is concerned with whether a rational being’s obligations for actions themselves are right or wrong. The consequences of an action are less important if the universalized duty was fulfilled to the person.
In the sense of AI, certain products may want to use different ethics.
Discrimination with regards to facial recognition in security cameras can be another example applied here. The mentioned study published in Science on 24 October, concluded that an algorithm used in hospitals to assess the need for medical assistance was less likely to refer black people than white people to programs to improve their health. These programs are used to manage more than 200 million users. Clearly, since it so widely used, discrimination has wide effects which could add to the act utilitarian distaste for it.
The act Utilitarian may not consider this as an unjust act since it considers consequences as more important than the actual rightness or wrongness of the act. More importantly, utilitarianism recognizes the injustice of inequality, but it “does not make it an inherently anti-inequality moral system”.
There may be an argument that there are more white people than black people and thus it could be justifiable to discriminate for simplification. Rule utilitarianism may apply obligations to not discriminate against black people and could thus be a more just ethical theory to apply because the discrimination would be shunned no matter the good consequences — if, and only if — there is an obligation.
John Rawls
Rawls, for example, thought that utilitarianism could not rule out absolutely slavery or racial segregation. He thinks “There is nothing in the moral theory to exclude them from consideration.” Discrimination is wrong even if there were economic benefits to it (and there are not to our knowledge, as is argued in the book, The Spirit Level: Why More Equal Societies Almost Always Do Better).
Deontology
If we were to take a deontological perspective to the dilemma we should differentiate between Kant’s hypothetical imperatives and his categorical imperatives. Briefly, Hypothetical imperatives are commands that we should follow if we want a certain outcome. For example, if you want a good grade you ought to study. Categorical imperatives, however, are obligatory and should be followed despite selfish desires to do otherwise.
There are two formulations of the categorical imperative (1) The Universalizability Principle: What Is The Maxim of My Action?, & (2) The Humanity Formula: Do Not Use People As a Means, but As An End in Themselves.
The strength of deontology is that it obliges people to follow consistent principles from seemingly objective premises. We cannot universalize discriminating against minorities or individuals. Therefore, we cannot act in such a way. We similarly cannot use minorities as a means for better health for ourselves (whites) and, thus, it would be immoral to act in such a way. For Kant, thus, the obligation is to not treat others unfairly and unjustly.
The difficulty, however, in Kant’s theory is that it does not account for (1) differing moral intuitions, and (2) differing moral obligations. It simply outlines absolute situations, such as lying, stealing, killing, etc where all of our moral intuitions are aligned. To use this present case as an example, one can argue that we have a moral obligation to serve efficiently not discriminating. However, conflicting with that can be a universalized principle to make people healthy with a higher motive for them to stay out of hospitals (to further segregation). In this case, it would be clear that the motives would be unethical despite the previous obligation to minority patients.
Conclusion
Therefore — if we were to choose an ethical theory — Rule Utilitarianism provides the most helpful means to analyze this form of injustice. Act Utilitarianism can justify discrimination in some cases, even if utilitarians would not argue for it (as was pointed out by Rawls), Deontology has the problem of differing moral intuitions and obligations. Rule utilitarians, however, make decisions based on consequences and obligations, managing to combine real data and human rights that are ideally universal to all no matter race, gender, or sexual orientation.
Before you go…
🗣 I love connecting with fellow thinkers. Find me on Twitter, Facebook, Goodreads, or Instagram.
I’d love if you’d share the article on Facebook/TWITTER if you want your friends to benefit from it in some way at all.
I write to keep you thinking and to keep me thankful and reflective. Cheers and until next time,
keep reflecting.