Chris (@multiplicityct.bsky.social)

If AI models elicit reactive attitudes in users and at least *simulate* having them in response, should we hold them accountable as moral agents? Provocative new article from Dong An in AI & Society. There's more than a kernel of truth to this, IMHO, at least as a description of AI use. #AIethics

loading . . .

Artificial reactive attitudes - AI & SOCIETY Scholars have primarily approached the issue of AI moral agency from the perspectives of consciousness and intentionality (Chalmers in The conscious mind: In search of a fundamental theory, Oxford Paperbacks, 1997; Johnson in Ethics Inf Technol 8:195–204, 2006; Searle in The rediscovery of the mind, MIT press, 1992) or by considering various functional substitutes for these faculties (Coeckelbergh in Ethics Inf Technol 12:235–241, 2010, Sci Eng Ethics 26(4):2051–2068; Floridi and Sanders in Minds Mach 14:349–379, 2004; Himma in Ethics Inf Technol 11:19–29, 2009). In this paper, I add to the existing literature to show the value of exploring the application of the reactive attitudes approach to AI moral agency (Antill in AI Ethics 5(2):1651–1667, 2024; Rebera in Philos Technol 37(4):126, 2024; Sars in Ethics Inf Technol 24(3):32, 2022; Smith and Vickers in Ethics Inf Technol 23(3):483–493, 2021; Tigard in Philos Technol 34(3):589–607, 2021; Tollon in AI Ethics 3(1):295–302, 2023). Reactive attitudes are attitudes we hold towards each other in everyday interactions with respect to our belief about each other’s quality of will. Accordingly, to hold people morally responsible is to adopt and express these attitudes in moral practice. In the existent attempts to use this framework to investigate AI moral agency, scholars have mostly focused on the capacities that AIs must possess to be considered apt recipients of reactive attitude. I adopt the reversal methodology by focusing on AIs as apt expressers of reactive attitudes. This methodology can allow us to draw resources from emotion research across philosophy, psychology, and computer science. My thesis is that because it is possible to attribute reactive attitudes to AIs, allowing them to engage in reactive attitudes exchange with humans in the shared moral community, they may be considered as moral agents and thus held accountable for their actions. https://link.springer.com/article/10.1007/s00146-025-02622-1