Abstract
The debates about ethics in the context of artificial intelligence have been recently focusing primarily on various types of utilitarianisms. This article suggests a categorization of the various presented utilitarianisms into static utilitarianisms and dynamic utilitarianisms. It explains the main features of both. Then, it presents the challenges the utilitarianisms in each group need to be able to deal with. Since it appears that those cannot be overcome in the context of each group alone, the article suggests a possibility of using a combination of the two categories of utilitarianisms to resolve most of the challenges without the need to abandon the concept of utilitarianisms as such. Even this possibility comes with its own issues that might not be resolved within the boundaries of utilitarianisms, however. Therefore, another potential alternative based on a combination of various ethical systems is suggested and briefly explored.
Similar content being viewed by others
Notes
It is necessary to point out that utility in the classical utilitarianism translated to well-being or happiness. In modern contexts, however, utility can be any assigned value that we consider worth maximizing.
It could probably be argued that static utilitarianisms could circumvent this issue by including various fail-safes to prevent potentially disastrous consequences, but it is hard to imagine how this would work in practice. All of these fail-safes would only treat the effects and not the causes of problematic behavior, since treating the causes would require some changes in the utility value function, which could lead to further problematic behavior.
References
Abbeel P, Ng AY (2004) Apprenticeship learning via inverse reinforcement learning. In: Proceedings of the twenty-first international conference on Machine learning (ICML '04). https://doi.org/10.1145/1015330.1015430
Abel D, MacGlashan J, Littman ML (2016) Reinforcement learning as a framework for ethical decision making. In: Papers from the 2016 AAAI workshop. AAAI Digital Library. http://www.aaai.org/ocs/index.php/WS/AAAIW16/paper/view/12582. Accessed 28 Nov 2020
Aliman N-M, Kester L (2019) Augmented utilitarianism for AGI safety. In: Hammer P, Agrawal P, Goertzel B, Iklé M (eds) Artificial general intelligence: AGI 2019. Lecture notes in computer science, vol 11654. Springer, Cham, pp 11–21
Bauer WA (2018) Virtuous vs. utilitarian artificial moral agents. AI & Soc 35:263–271
Baujard A (2009) A return to Bentham’s felicific calculus: from moral welfarism to technical non-welfarism. Eur J Hist Econ Thought 16(3):431–453
Bonnemains V, Saurel C, Tessier C (2018) Embedded ethics: some technical and ethical challenges. Ethics Inf Technol 20:41–58
Bostrom N (2014) Superintelligence: paths, dangers, strategies. Oxford University Press, Oxford
Bowden P (2009) Defense of utilitarianism. University Press of America, Lanham. https://doi.org/10.2139/ssrn.1534305
Faulhaber AK, Dittmer A, Blind F, Wächter MA, Timm S, Sütfeld LR et al (2018) Human decisions in moral dilemmas are largely described by utilitarianism: virtual car driving study provides guidelines for autonomous driving vehicles. Sci Eng Ethics 25:399–418
Greaves H, MacAskill W (2019) The case for strong longermism. GPI working paper no. 7–2019. Global Priorities Institute. https://globalprioritiesinstitute.org/wp-content/uploads/2020/Greaves_MacAskill_strong_longtermism.pdf. Accessed 28 Nov 2020
Hagendorff T (2020) The ethics of AI ethics: an evaluation of guidelines. Mind Mach 30:99–120
Herd S, Read SJ, Oreilly R, Jilk DJ (2018) Goal changes in intelligent agents. In: Yampolskiy RV (ed) Artificial intelligence safety and security. CRC Press, Cambridge, pp 217–224
Hibbard B (2012) Avoiding unintended AI behaviors. In: Bach J, Goertzel B, Iklé M (eds) Artificial general intelligence: AGI 2012: Lecture notes in computer science, vol 7716. Springer, Berlin
Hibbard B (2015) Ethical artificial intelligence. https://arxiv.org/abs/1411.1373
Hooker JN, Kim TW (2018) Toward non-intuition-based machine and artificial intelligenceethics: a deontological approach based on modal logic. In: AIES '18: Proceedings of the 2018 AAAI/ACM conference on AI, ethics, and Society. https://doi.org/10.1145/3278721.3278753
Leike J, Krueger D, Everitt T, Martic M, Maini V, Legg S (2018) Scalable agent alignment via reward modeling: a research direction. https://arxiv.org/abs/1811.07871
Lucas J, Comstock G (2015) Do machines have prima facie duties? In: Van Rysewyk SP, Pontier M (eds) Machine medical ethics. Springer, Berlin, pp 79–92
Monton B (2019) How to avoid maximizing expected utility. Philos Impr 19(18):1–25
Omohundro SM (2008) The basic AI drives. In: Wang P, Goertzel B, Franklin S (eds) Proceedings of the AGI conference, vol 171, IOS Press, The Netherlands, pp 483–492.
Poulsen A, Anderson M, Anderson SL, Byford B, Fossa F, Neely EL et al (2019) Responses to a critique of artificial moral agents. https://arxiv.org/abs/1903.07021
Powers TM (2006) Prospects for a Kantian machine. IEEE Intell Syst 21(4):46–51
Rautenbach G, Keet CM (2020) Toward equipping artificial moral agents with multiple ethical theories. In: Proceedings of RobOntics: international workshop on ontologies for autonomous robotics, September, CEUR-WS. UCT computer science research document archive. https://pubs.cs.uct.ac.za/id/eprint/1393/. Accessed 18 Jan 2021
Rawls J (1971) A theory of justice. Harvard University Press, Cambridge
Ray A, Achiam J, Amodei D (2019) Benchmarking safe exploration in deep reinforcement learning. In: OpenAI. https://openai.com/blog/safety-gym/
Tonkens R (2009) A challenge for machine ethics. Mind Mach 19:421
Torres P (2019) The possibility and risks of artifical general intelligence. Bull At Sci 75(3):105–108
Turchin A (2018) Levels of AI self-improvement. LessWrong. https://www.lesswrong.com/posts/os7N7nJoezWKQnnuW/levels-of-ai-self-improvement. Accessed 18 Jan 2021
Wang X, Zhao Y, Pourpanah F (2020) Recent advances in deep learning. Int J Mach Learn Cybern 11:747–750
White J (2020) Autonomous reboot: Aristotle, autonomy and the ends of machine ethics. AI & Soc. https://doi.org/10.1007/s00146-020-01039-2
Yampolskiy RV (2020) Unpredictability of AI: on the impossibility of accurately predicting all actions of a smarter agent. J Artif Intell Conscious 7(1):109–118
Funding
The research was funded via a university grant (Antropocentrismus v etice, University of Ostrava).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Cvik, Š. Categorization and challenges of utilitarianisms in the context of artificial intelligence. AI & Soc 37, 291–297 (2022). https://doi.org/10.1007/s00146-021-01169-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00146-021-01169-1