Giles Howdle

When interacting with social AI systems (SAIs), we routinely speak of what they ‘believe’, ‘want’, or ‘know’. With some exceptions, philosophers tend to treat such anthropomorphism as a single phenomenon that risks one kind of _error_: mistaken ontological commitment to machine minds and mental states. This paper challenges this monistic assumption. I distinguish two modes of anthropomorphic attribution—metaphysical and pragmatic—and identify two corresponding kinds of possible anthropomorphic error. In the metaphysical mode, speakers commit themselves to the existence of machine mental states, risking straightforward ontological error. In the pragmatic mode, speakers adopt the intentional stance without ontological commitment, yet still risk error when another interpretive strategy would better serve their purposes. I defend Mixed Anthropomorphism: both modes are common. This pluralist account reveals that the current debate’s focus on whether users ‘really mean it’ obscures the pragmatic dimension of anthropomorphic ascription (and its risks). Even ontologically innocent anthropomorphism can constitute a mistake because it employs the _wrong interpretive tool_ for the task at hand. Understanding these distinct error types matters both theoretically, for clarifying the nature of human-AI interaction, and practically, for designing systems that encourage and scaffold appropriate interpretive strategies.

The thesis combines the Humean theory of normative practical reasons with constitutivism. Humeanism claims, roughly, that an agent’s reasons are explained by his desires. Constitutivism claims that the foundations of moral normativity can be found in the constitutive features of agency. This combination, ‘Humean constitutivism’ as I call it, claims there is some desire-like mental state, the possession of which is partly constitutive of rational agency, which explains the normativity of morality. I first argue that the union of constitutivism and Humeanism makes for an attractive marriage. As with any good relationship, the two bring out the best in one another. Humeanism offers the constitutivist a compelling account of how we get from the ‘is’ of the nature of rational agency to the ‘ought’ of normative conclusions. In doing so, it permits the constitutivist to circumvent the influential ‘shmagency’ objection. The constitutivist strategy provides the Humean with the resources to accommodate moral rationalism and hence answer sceptical challenges to morality’s practical normative authority. Having established its attractiveness, I then defend the truth of Humean constitutivism. I argue that there are desire-like mental states (constitutive commitments), the possession of which is partly constitutive of agency. This is because rational agency involves distinctively linguistic forms of cognition—dwelling in something like Sellars’ ‘logical space of reasons.’ Engaging in this linguistic form of cognition, in turn, requires us to take on certain second-personal practical commitments to our (hypothetical) interlocutors. Hence, all agents necessarily have these constitutive commitments. Piggy-backing on Darwall’s derivation of contractualist moral norms from the constitutive presuppositions of the second- person standpoint, I claim that the Humean constitutivist can vindicate the intuitive existence, modal status, universality, and special weight of (many) moral reasons.

Abstract‘Dogwhistles’ and microtargeted political advertisements are objects of widespread moral and political concern. With a few notable exceptions in the case of dogwhistles (and none in the case of microtargeting) moral criticism of these speech act types generally focuses on problematiccontent—that a dogwhistle is, for instance, racist, or a microtargeted advertisement misleading. I argue that these practices areadditionallymorally wrongful on content-neutral grounds—regardless of their content. My argument proceeds from a deliberative conception of democracy according to which only a vote which follows from anadequate deliberative processconfers democratic legitimacy on its results. I claim that both dogwhistles and microtargeting threaten to prevent adequate democratic deliberation from taking place, and therefore that these practices are anti-democratic and, additionally, morally impermissible. I then discuss potential objections and how my argument relates to existing work.