Institutional Anthropomorphic Error

Abstract Sakr proposes institutional anthropomorphic error (the projection of role-based authority onto AI systems that cannot bear a role’s constitutive obligations) as a third kind of error to be added to my distinction between metaphysical and pragmatic anthropomorphic error. I argue that the proposal is better understood as belonging to a second category of anthropomorphism than a third kind of error within an existing one. To ‘anthropomorphise’ a system may mean either seeing it as human-like (anthropomorphic cognition) or presenting it as human-like (anthropomorphic presentation). Metaphysical and pragmatic error are two kinds of error in anthropomorphic cognition; institutional anthropomorphic error, as Sakr defines it, is an error of presentation. A cognitive counterpart can be formulated, but it resolves into a distinctive form of either metaphysical or pragmatic error. Sakr’s contribution is valuable, but it does not require revision of the two-part account. Similar content being viewed by others 1 Introduction I am grateful to Sakr (2026) for his article, which builds on the two-way distinction I have drawn between metaphysical and pragmatic anthropomorphic error with respect to social AI systems (SAIs) (Howdle, 2026). Sakr identifies a third form of anthropomorphic error: Institutional Anthropomorphic Error: positioning or representing an AI system as occupying a human social role in a manner that projects role-based authority onto the system and invites role-based trust in it, although the system cannot bear the role’s constitutive obligations and the arrangement obscures an accountable human or institutional bearer. (2026, p. 3) Institutional anthropomorphic error is a useful philosophical concept that furthers our understanding of a rapidly emerging and significant concern. However, as Sakr acknowledges, it differs in important respects from the two errors I have described. My aim here is to clarify the significance of those differences. I claim (1) that we need to distinguish two senses of anthropomorphism and two related categories of anthropomorphic error; (2) that the two kinds of error I have identified fall under the same category, while institutional anthropomorphic error as Sakr describes it falls under a different category; and (3) that we can reconceive institutional anthropomorphic error so that it belongs in the same category as metaphysical and pragmatic anthropomorphic error, but that doing so does not require revision of the two-part account I have defended. 2 Two Senses of Anthropomorphism There is a sense in which a person ‘anthropomorphises’ a system by seeing it as human-like. There is another in which a person ‘anthropomorphises’ a system by presenting it as human-like. Call this the difference between anthropomorphic cognition and anthropomorphic presentation. Both can be mistakes, but they are two different categories of mistake. Institutional anthropomorphic error, as Sakr defines it, is a mistake with respect to the ‘positioning or representing’ of an AI system. It is an error of anthropomorphic presentation. The distinction I have drawn between metaphysical and pragmatic anthropomorphic error, by contrast, is a distinction between two kinds of error in anthropomorphic cognition. Metaphysical error is a matter of being committed to ontological falsehoods with respect to an SAI (e.g. non-existent machine mental states), while pragmatic error is a matter of adopting the intentional stance toward an SAI when another interpretive strategy would be better. These are bad ways of seeing and thinking about SAIs, not bad ways of presenting them. Both pertain to erroneous anthropomorphic cognition. My paper did briefly draw some normative conclusions about anthropomorphic presentation, but without theorising it. Sakr does a great deal to develop and clarify these normative concerns, and his emphasis on the social roles and offices SAIs occupy, or are taken to occupy, and on their normative significance, is a substantial contribution. Introducing a system as ‘your clinical assistant’ is not equivalent to saying that the model believes p; it places the system in a recognised social role rather than attributing a state to it. I do not, however, agree with Sakr’s way of framing his contribution. He proposes institutional anthropomorphic error as a third form of error, to be added to my two-part taxonomy. But my taxonomy concerns errors of anthropomorphic cognition, whereas his concerns an error of anthropomorphic presentation (Table 1). 3 A Cognitive Institutional Anthropomorphic Error? I now wish to consider if there is a cognitive version of Sakr’s institutional anthropomorphic error, and, if so, whether it should be adopted as a third kind of cognitive anthropomorphic error. Here is a proposal: Cognitive Institutional Anthropomorphic Error: thinking of an AI system as occupying a human social role that holds role-based authority and role-based trust, although the system cannot bear the role’s constitutive obligations and the arrangement obscures an accountable human or institutional bearer. I suspect this kind of cognitive error occurs regularly and that it matters for many of the reasons Sakr suggests. However, I also suspect my two-part taxonomy can accommodate it without modification. To see why, consider what might be happening when someone engages in cognitive institutional anthropomorphic error. A metaphysical institutional anthropomorphic error would be to commit oneself to an ontological falsehood. Specifically, one’s thought or talk would commit one to the existence of an SAI that is occupying a social role or office it does not (perhaps cannot), in fact, occupy. Or, stated in terms of presupposition error, a metaphysical institutional error would consist in thinking about the system in ways that presuppose normative competences and powers the SAI lacks. A pragmatic institutional anthropomorphic error would consist in selecting a suboptimal interpretive stance. Consider a user who treats a chatbot as if it occupied the office of a therapist, without committing herself to the claim that it does. If the interaction proves harmful, she directs her complaint at the chatbot (as one might to a therapist) rather than at whoever authorised and reviewed its deployment. Her aim of recourse for bad therapeutic advice would have been better served by some alternative interpretive stance. 4 Conclusion Sakr’s concept of institutional anthropomorphic error is a genuine contribution. But it does not, I think, demonstrate that the two-way distinction between metaphysical and pragmatic anthropomorphism requires an institutional sibling. Understood as an error of presentation, institutional anthropomorphic error pertains to a different category of anthropomorphism, and a different category of anthropomorphic error. Understood instead as an error of cognition, I suspect it reduces to a distinctive form of metaphysical or pragmatic anthropomorphic error. Data Availability Not applicable. References Howdle, G. (2026). Anthropomorphising AI: Two Modes, Two Errors. Philosophy & Technology, 39(3), 140. Sakr, J. (2026). Beyond Two Errors: Institutional Anthropomorphism and the Displacement of Responsibility. Philosophy & Technology, 39(3), 156. Funding This work is part of the research programme Ethics of Socially Disruptive Technologies, which is funded through the Gravitation programme of the Dutch Ministry of Education, Culture, and Science and the Netherlands Organization for Scientific Research (NWO grant number 024.004.031). Author information Authors and Affiliations Corresponding author Ethics declarations Ethical Approval Not applicable. Consent to Participate Not applicable. Consent for Publication Not applicable (Single author). Competing Interest The author has no relevant financial or non-financial interests to disclose. Additional information Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Rights and permissions Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. About this article Cite this article Howdle, G. Institutional Anthropomorphic Error. Philos. Technol. 39, 190 (2026). https://doi.org/10.1007/s13347-026-01191-5 Received: Accepted: Published: Version of record: DOI: https://doi.org/10.1007/s13347-026-01191-5