A framework for thinking about and deploying ethics in AI

Abstract Many codes of ethics on Artificial Intelligence (AI), although well-intentioned, have been proposed at a high level of abstraction without much thought to implementation. Companies should adopt a code of ethics for AI because not to do so could, in Akerlof’s market for lemons theory, lead to the downfall of the market or the company. Using the Kidder framework, the author first differentiates between moral temptations (right vs. wrong) that should be avoided and true ethical dilemmas (right vs. right) that are to be addressed. Of the various ethical theories, the author concludes that the most appropriate is Aristotle's Golden Mean, which should be seen as a creative solution rather than a compromise. The author also points out that while there are clear ethical issues, there are also issues that should considered within the bounds of ethics; but there are also issues that while discomfiting are not ethical issues.