Abstract
The ethical decision-making and behaviour of artificially intelligent systems is increasingly important given the prevalence of these systems and the impact they can have on human well-being. Many current approaches to implementing machine ethics utilize top-down approaches, that is, ensuring the ethical decision-making and behaviour of an agent via its adherence to explicitly defined ethical rules or principles. Despite the attractiveness of this approach, this chapter explores how all top-down approaches to implementing machine ethics are fundamentally limited and how bottom-up approaches, in particular, reinforcement learning methods, are not beset by the same problems as top-down approaches. Bottom-up approaches possess significant advantages that make them better suited for implementing machine ethics.
TopBackground
As long as people have imagined artificially intelligent machines (e.g., robots, ordinary physical machines, computers, software, etc.), they have also imagined how these machines might act. Most contemporary portrayals of advanced AIs seem to regard as necessary the extinction of humanity even though they are not particularly adept at accomplishing their goals.1 Sensationalized fictions aside, and despite the nascent state of most artificially intelligent machines, the problem of implementing machine ethics is a live and pressing issue. As some scholars have noted, essentially all non-trivial interactions that an intelligent machine has with humans have ethical import (Anderson, Anderson, & Berenz, 2017). It is possible that robots responsible for eldercare or childcare, for example, could cause harm via their inaction if they recharge their batteries at one particular point in time instead of another.
Key Terms in this Chapter
Domain: Any well-defined region or state space.
Ethical Machine: A machine whose behavior is either ethically aligned with what humans consider to be ethically desirable or acceptable and/or a machine capable of engaging in ethical reasoning.
Artificial Intelligence: Any artificial system that is capable of completing some task such that if the same task were completed by a human, intelligence would be attributed to that human.
Top-Down: In general, any system that proceeds from the general to the particular or from the highest level to the lowest.
Reinforcement Learning: A machine learning paradigm that utilizes evaluative feedback to cultivate desired behavior.
Rigidity: An inability to act in situations that were not explicitly accounted for.
Bottom-Up: In general, any system that proceeds from the particular to the general or from the lowest level to the highest.