Beyond Kant and Bentham: How Ethical Theories are being used in Artificial Moral Agents

Title of the Paper

Beyond Kant and Bentham: How Ethical Theories are being Used in Artificial Moral Agents

Bibliographic Information

• Subject Area: AI ethics, particularly the study of Artificial Moral Agents (AMAs) and the application of ethical theories.
• Keywords: Ethics, philosophy, artificial moral agents, literature review, deontology, consequentialism, virtue ethics, Confucian ethics, indigenous ethics.

Research Background and Issues

• Issues or Challenges:
  • How can ethical norms be implemented and encoded in artificial moral agents?
  • Is the current development of AMAs limited to certain ethical paradigms, such as deontology and consequentialism?
  • Are there cross-cultural and interdisciplinary biases that lead to the neglect of other ethical frameworks?
• Research Significance:
  • With the proliferation of autonomous vehicles, eldercare robots, and automated financial systems, it is increasingly critical to ensure that machines can act in morally appropriate ways.
  • The current dominance of Western philosophical ethics in machine ethics research may hinder the representation of diverse ethical values on a global scale.
• Motivation and Related Work:
  • Previous studies have extensively applied deontology and consequentialism to the field of machine ethics, but there has been limited exploration of other philosophical theories, such as Confucian and indigenous ethics.
  • The practical application of AMAs still lacks detailed empirical evidence.

Proposed Solution

• Proposed Methods or Solutions:

  • Conduct a multidisciplinary literature review on the application of ethical theories in AMAs, covering both Western paradigms (e.g., deontology and consequentialism) and non-Western paradigms (e.g., Confucian ethics and indigenous ethics).
  • Categorize and analyze existing literature to reveal which ethical theories are most commonly adopted and how they are implemented.

• Innovative Contributions:

  • Moves beyond typical moral decision-making frameworks (e.g., rules and consequences) to explore under-researched areas such as virtue ethics, Confucian ethics, indigenous ethics, and existentialism.
  • Highlights the potential role of integrating ethical theories with human-computer interaction (HCI) in designing fair and responsible artificial intelligence.

• Implementation Steps and Key Techniques:

  • Literature Construction: Keywords such as "machine ethics" and related terms were filtered from ACM, IEEE, and reference citations, resulting in the selection of 53 papers for review.
  • Methodological Framework: Literature content was categorized and annotated based on specific ethical theories (e.g., deontology, consequentialism, virtue ethics) and research methods (e.g., technical implementation, theoretical analysis, empirical studies).

Research Findings

• Specific Findings:

  1. Of the 53 papers, 46 primarily focused on deontology and consequentialism, along with supplementary theories such as the principle of double effect and Ross's ethics. Few papers explored non-Western paradigms (e.g., Confucian ethics, indigenous ethics).
  2. The study highlights the potential issue of Western ethical centralism in AMA development, which may hinder the application of broader ethical values.
  3. From an implementation perspective, most academic papers focus on technical implementation and theoretical analysis, with few addressing empirical studies in real-world environments.

• Comparison with Existing Solutions:

  • Most current research is limited to analytical debates or technical proposals, lacking empirical studies to test and validate theories.
  • This study points out that over-reliance on deontological/consequentialist frameworks may lead to restrictive AMA development, neglecting the global diversity of ethical perspectives.

• Experimental or Evaluation Results:

  • Some empirical studies found that decision-making models for AMAs based on Western ethical paradigms (e.g., deontology and consequentialism) may face lower public acceptance.
  • Paradigms such as Confucian ethics, which emphasize relationships and social roles, can contribute to developing more meaningful human-machine interactions.

• Limitations and Future Directions:

  1. Current Limitations:
     • The literature is primarily focused on Western contexts, with insufficient in-depth empirical exploration of other philosophical traditions (e.g., existentialism, indigenous ethics).
     • The HCI community has yet to fully develop the potential applications of AMAs in real-world social scenarios.
  2. Future Directions:
     • Encourage exploration of non-Western philosophical traditions, particularly the application of Confucian ethics, existentialism, and indigenous perspectives in AMAs.
     • Advance empirical research to evaluate how AMAs influence human decision-making and gain social acceptance in practice.
     • Develop more relationship-oriented and long-term feedback mechanisms within virtue ethics and moral development frameworks.

Through this study, the authors not only describe and summarize the mainstream ethical frameworks for AMAs but also propose important recommendations for promoting diverse, multicultural research and practice in machine ethics. These insights lay a theoretical foundation for the HCI and AI communities to further develop moral machines that can be effectively integrated into society.