Home | Business News | Browse by Publication | E | Ethics and Information Technology

Artificial morality: top-down, bottom-up, and hybrid approaches.

Article Excerpt
Abstract. A principal goal of the discipline of artificial morality is to design artificial agents to act as if they are moral agents. Intermediate goals of artificial morality are directed at building into AI systems sensitivity to the values, ethics, and legality of activities. The development of an effective foundation for the field of artificial morality involves exploring the technological and philosophical issues involved in making computers into explicit moral reasoners. The goal of this paper is to discuss strategies for implementing artificial morality and the differing criteria for success that are appropriate to different strategies.

Key words: artificial morality, autonomous agents, ethics, machines, robots, values

Introduction

Artificial morality shifts some of the burden for ethical behavior away from designers and users, and onto the computer systems themselves. The task of developing artificial moral agents becomes particularly important as computers are being designed to perform with greater and greater autonomy, i.e., with less and less direct human supervision. The speed at which computers execute tasks increasingly prohibits humans from evaluating whether each action is performed in a responsible or ethical manner. Implementing software agents with moral decision-making capabilities offers the promise of computer systems that are able to evaluate whether each action performed is ethically appropriate. This in no way serves as a substitute for the moral responsibility of those who deploy or use computers. It merely means that it will be easier to use computers in an ethical manner if sensitivity to ethical and legal values is integral to the software (Wallach 2004).

Regardless of whether artificial morality is genuine morality, artificial agents act in ways that have moral consequences. This is not simply to say that they may cause harm--even falling trees do that. Rather, it is to draw attention to the fact that the harms caused by artificial agents may be monitored and regulated by the agents themselves. While some aspects of the insertion of intelligent artifacts into the moral landscape may be manageable without specific attention to the software or mechanisms controlling their behavior--i.e., by treating them as ethical 'black boxes'--we would argue that artificial morality should also be approached proactively, as an engineering design challenge for explicitly building ethically appropriate behavior into artificial agents. The attempt to design artificial moral agents forces consideration of the information required for ethical decision making, and of the algorithms which may be appropriately applied to the available information.

In this paper we discuss the philosophical roots and computational possibilities of top-down and bottom-up strategies for designing artificial moral agents (AMAs). We pick up where Allen et al. (2000) left off when they wrote, 'Essential to building a morally praiseworthy agent is the task of giving it enough intelligence to assess the effects of its actions on sentient beings, and to use those assessments to make appropriate choices'. Top-down approaches to this task involve turning explicit theories of moral behavior into algorithms. Bottom-up approaches involve attempts to train or evolve agents whose behavior emulates morally praiseworthy human behavior.

Top-down approaches

The idea behind top-down approaches to the design of AMAs is that moral principles or theories may be used as rules for the selection of ethically appropriate actions. Rule-based approaches to artificial intelligence have been appropriately criticized for their unsuitability for providing a general theory of intelligent action. Such approaches have proven to be insufficiently robust for almost any real-world task. Yet there remain specific domains where rule-based approaches provide the best available technology, and it is an open research question whether moral behavior is one of these domains. Hence, even though (we believe) the objections to rule-based approaches are rather strong, it is incumbent on aspiring designers of AMAs to consider the prospects and problems inherent in top-down approaches. Specific objections to top-down approaches to artificial morality (rather than general objections to rule-based A.I. generally) are best understood after giving careful consideration to the prospects for building AMAs by implementing decision procedures that are modeled on explicit moral theories.

Candidate principles for conversion to algorithmic decision procedures range from religious ideals and moral codes to culturally endorsed values and philosophical systems. The Golden Rule, The Ten Commandments, utilitarianism, and Kantian deontology are some of the possible sources of rules for top-down ethical systems. When considering robots, Asimov's three laws...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!