The Ethics of Artificial Intelligence

Abstract: This paper is about the ethics of giving computers decision-making capacities. Some possible roles of decision-making computers are articulated, along with the effects of placing computers in such positions. The essential ingredients to creating an effective decision-making computer are discussed.

Kevin Bowyer, author of Ethics In Computing, advances a question as the only ethical dilemma unique to the field of computing. (Bowyer 3) He asks: "how much decision-making should be entrusted to a machine?" In this paper I'd like to explore some of the issues around this topic, which I will generalize into the question "is human judgment essential to a decision-making process?" After discussing the impacts the answer to this question could have on human institutions, I will ultimately conclude that human judgment is a critical ingredient to decision-making, but not necessarily at a fixed point in the process.

First it is useful to discuss in what sense computers can be used in a decision-making process. As computers approximate more and more the range of capacities that humans are capable of, the more they will be able to take an active role in organizing our lives. Consider, for example, the role of a bank's computer systems. While one might argue that computerized record-keeping is more of a tool rather than an active force, bookkeeping used to be an activity entirely reserved for people. Therefore, it constitutes an activity that has been mechanized in some way. It is not inaccurate to say that the pen has been passed from human hands to digital ones. Practically, what are the differences? Perhaps human bookkeeping allowed a degree of supervision, e.g. illegal activity was easier to keep track of. (However, it could also work the other way, and the efficiency of computers could introduce this sort of monitoring.) The point of this example is to demonstrate that every action--or lack thereof--a person takes can be said to contribute to some sort of decision. If a bookkeeper caught a mistake or suspicious activity, he might be able to act in a way that a computer cannot.

Taken from the above perspective, there isn't a whole lot that computers couldn't have a hand in. Some examples include computers that auto-fix problems in a vehicle, or even in the trajectory of a missile; automated defense systems in a building; programs that monitor user activity, perhaps in a company computer; auto-pilots on airplanes; or medical computers that administer drugs or otherwise take action automatically on a patient.