Abbas Bagwala

Large language models are increasingly taken to be capable of semantic practices. Debate about what to make of this has largely been conducted as a dispute about understanding. I argue that these standard framings, whichever side one takes, leave out an essential component of discursive reasoning: responsibility. To assert something is to make oneself answerable for what is claimed, to stand ready to defend it, and to give up a previously-asserted claim that cannot be defended. What settles whether a system is a participant in discursive practice is therefore not merely whether it understands, nor merely how well it performs, but whether it can be held to what it has claimed. Drawing on Brandom’s inferentialism, I develop five criteria for responsible reasoning, each of which admits of degrees. What they track is not independence from a community, since no speaker’s discursive capacities are independent of the community that institutes the norms she is answerable to, but the degree to which an agent does its own deontic scorekeeping. The central thesis I defend is accordingly agent-neutral: an agent whose performances fail these criteria is not, to that extent, reasoning responsibly.

Information is a unique resource. Asymmetries that arise out of information access or processing capacities, therefore, enable a distinctive form of injustice. This paper builds a working conception of such injustice and explores it further. Let us call it informational injustice. Informational injustice is a consequence of informational asymmetries between at least two agents, which are deeply exacerbated due to modern information and communication technologies but do not necessarily originate with them. Informational injustice is the injustice of having information from an informational surplus being used to disadvantage the agent with less information. This paper argues that informational injustice exploits an agent as a knower, specifically exploiting the agent’s limitation in possessing or processing information—an agent is exploited because she is not informed or lacks in her ability to process information. In the case of lack of information, the agent simply does not know the information under consideration; a person is algorithmically manipulated or nudged to buy a product or vote for someone. In the case of a lack of capacity to process information, the agent simply cannot use the information, despite having access to it, to reach epistemically valuable states such as knowledge; a lawyer dupes you because he knows more about the inner workings of a courtroom and the law. Technically, you have access to the information your lawyer has, but you cannot make use of it due to constraints on time and cognitive effort. Informational injustice excludes the harmed agent from participating in knowledge practices. Thus, informational injustice is also a kind of epistemic exclusion. After fixing the concept of informational injustice, the paper distinguishes between two kinds of informational injustices: interactional informational injustice and structural informational injustice. The former concerns interactions between agents, while the latter concerns social structures that emerge out of interactions between agents.