Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems
Open accessThe DOI matches its Crossref record; 71 of the 71 references it cites were found in a registry or an index; 10 of the 40 sources its claims cite were checked for support.
Abstract
The high predictive accuracy of contemporary machine learning-based AI systems has led some scholars to argue that, in certain cases, we should grant them epistemic expertise and authority over humans. This approach suggests that humans would have the epistemic obligation of relying on the predictions of a highly accurate AI system. Contrary to this view, in this work we claim that it is not possible to endow AI systems with a genuine account of epistemic expertise. In fact, relying on accounts of expertise and authority from virtue epistemology, we show that epistemic expertise requires a relation with understanding that AI systems do not satisfy and intellectual abilities that these systems do not manifest. Further, following the Distribution Cognition theory and adapting an account by Croce on the virtues of collective epistemic agents to the case of human-AI interactions we show that, if an AI system is successfully appropriated by a human agent, a hybrid epistemic agent emerges, which can become both an epistemic expert and an authority. Consequently, we claim that the aforementioned hybrid agent is the appropriate object of a discourse around trust in AI and the epistemic obligations that stem from its epistemic superiority.
Overview
CoChat Summary
TL;DR
The paper argues that AI systems cannot be genuine epistemic experts or authorities of understanding because they lack understanding and the relevant research- and novice-oriented abilities; at most they can have weak expertise and authority of belief. It proposes that successful human-AI appropriation yields a hybrid agent that can become an epistemic expert or authority, making that hybrid the proper subject for debates about trust and epistemic obligation.
Key findings
- AI systems do not qualify as Croce-style epistemic experts: they lack understanding, research-oriented abilities, mental states, and cognitive agency, so high predictive accuracy alone does not generate an epistemic obligation to rely on them. They can at best be repositories of extensive “weak knowledge” limited to predictive tasks and under the guardianship of human experts.
- AI systems may in some contexts be ascribed authority of belief, but not authority of understanding, because they lack novice-oriented abilities such as sensitivity to laypeople’s epistemic needs and maieutic ability; epistemic opacity further precludes fostering understanding in many human-AI interactions.
- Human-AI interactions can produce a hybrid cognitive and epistemic agent when a human successfully appropriates an AI system to achieve an epistemic goal. Adapting Croce’s non-summativist account of collective epistemic agents, the paper argues such hybrids can be epistemic experts and authorities even though the AI part alone is not.
- The paper proposes a “minimum quantitative criterion” for successful appropriation: for any prediction, the human supported by the AI must have accuracy higher than or equal to the accuracy the human and AI would achieve independently without interaction. It illustrates this with a technology-savvy physician using a sepsis-prediction AI and a grandfather using ChatGPT to answer an astronomy question.
Methods
- Conceptual and philosophical analysis; no empirical study, dataset, or sample is reported.
- The argument draws on virtue epistemology accounts of expertise and authority (Croce, Zagzebski), Alvarado’s account of AI as epistemic technology, Distributed Cognition theory, and Croce’s collective-agent account.
- Main analysis proceeds by critiquing claims that medical AI has epistemic expertise/authority, then adapting a non-summativist collective virtue account to human-AI hybrids and illustrating it with two examples. No statistical analysis or quantitative results are reported.
Limitations
- The paper states that future research should investigate the conditions that grant reliability to the emergence of hybrid agents and their epistemic superiority, and how perception of the hybrid agent’s expertise and authority unfolds, including conditions fostering appropriate assessment by third parties.
- The paper states that it refrains from modeling the relation between appropriation dimensions, their interactions, and the degree of successful appropriation in a quantitative way.
- Evident from the text: the argument is not empirically tested, and the two examples are illustrative rather than data-driven; the paper also acknowledges that its minimum criterion does not explain how the hybrid’s superior accuracy is achieved, justified, or communicated.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
9 Valid · 2 Not valid · 23 Not checked · 71 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“we show that hybrid agents can be epistemic experts and authorities.”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 1 Introduction
What the sources say
- Partly supportedcites Croce, 2019b sourceExperimental — accuracy not yet measured
Section 3 argues that collective agents, including research teams, can be epistemic experts, and Section 1 proposes an account of their epistemic authority. The paper does not establish the more specific claim about hybrid agents; its example of engineers using a computer in Section 3.1 treats the engineers as the knowing group, not the human–computer combination as an agent. (LLM, conf=0.81)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Altogether, these properties show potential to mitigate, in particular, the type of opacity of the AI that is relevant in the context of the interaction (Facchini & Termine, 2022) and that may hinder the successful appropriation of the system, i.e., when the enhancement of the epistemic capabilities of a user by the AI is not sufficient for the human agent to achieve their epistemic goal.”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 4.1 What is a Hybrid Agent?
What the sources say
- Supportedcites Facchini & Termine, 2022 source
The Conclusions describe AI opacity as contextual: what counts as opaque depends on users’ purposes, background knowledge, and cognitive abilities. They also propose matching XAI methods and tools to specific forms of opacity and contexts. This supports the point attached to the citation—that the relevant type of opacity depends on the interaction context. (LLM, conf=0.91)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Finally, we show that an hybrid agent, differently from AI systems, can be endorsed with genuine forms of epistemic expertise and authority.”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 4.2 On the Epistemic Expertise and Authority of a Hybrid Agent
What the sources say
- Partly supportedcites Croce, 2019b sourceExperimental — accuracy not yet measured
Section 1 says the paper aims to explain how a collective agent can possess epistemic authority, and section 3.2 describes intellectual virtues that an expert group can display. These passages support the possibility of collective expertise and authority, but they do not establish that a hybrid agent has them or contrast hybrid agents with AI systems; a search for those terms found no relevant passage. Page numbers were not available in our copy. (LLM, conf=0.82)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The point here is that a hybrid agent can possess epistemic abilities even if one of its parts, i.e., the AI system, taken individually, fails to have them.”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 4.2 On the Epistemic Expertise and Authority of a Hybrid Agent
What the sources say
- Partly supportedcites Croce, 2019b sourceExperimental — accuracy not yet measured
Section 3.2 says a collective can display intellectual virtues even when its individual members lack them, supporting the claim’s general point about abilities of a whole and its parts. The examples concern teams of people, however; the paper does not establish that the same holds for a human–AI hybrid or that its AI component lacks epistemic abilities. No page numbers are available in our copy. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
In the paper
“according to Alvarado, AI is a paradigmatic example of epistemic technology (Alvarado, 2023).”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 3.1 AI, Epistemic Technology and Epistemic Enhancers
What the sources say
- Not checked — no full text of the sourcecites Alvarado, 2023 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
44 more claims
Create a free account to see all 105 claims and every source verdict.
- Probably not a substantial claim. 45
- It may state two separate claims as one. 16
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Generative AI tools produce outputs that often appear highly credible, but are simultaneously prone to factual errors, fabricated sources and misleading content (Chelli et al., 2024;Ferrario et al., 2024).”
From “ ChatGPT can make mistakes. Check important info.” Epistemic beliefs and metacognitive accuracy in students' integration of ChatGPT content into academic writing (2025) · cites this paper as Ferrario et al., 2024
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 5 Final Remarks
“This criterion states that the hybrid agent is better than its proper parts at generating predictions.”
§ Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems > 4.1 What is a Hybrid Agent?
“X's accuracy is higher or equal to the accuracy of the prediction with respect to S that X and M would achieve independently from each other if no interaction took place between them”
Citations
Cites
Dermatologist-level classification of skin cancer with deep neural networks
Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017) · Nature · doi:10.1038/nature21056 · Cited by 14,362 · cited 3×
“Fueled by an increasing amount of high quality data and recent advances in ML methodologies (namely, deep learning), medical AI has demonstrated performance that often matches or exceeds that of human experts in various tasks (Bjerring & Busch, 2021; Esteva et al., 2017; Liu et al., 2019).”
APACHE II: A severity of disease classification system
KNAUS, W. A., DRAPER, E. A., WAGNER, D. P., & ZIMMERMAN, J. E. (1985) · Critical Care Medicine · doi:10.1097/00003246-198510000-00009 · Cited by 13,664 · cited 1×
“A more thorough assessment by other physicians and methods, e.g., the APACHE II score (Knaus et al., 1985), confirms the successful appropriation of the AI and the accurate prediction of the hybrid agent.”
Hutchins, E. (1995) · doi:10.7551/mitpress/1881.001.0001 · Cited by 10,801 · cited 3×
“Then, by applying the theory of Distributed Cognition (Hutchins, 1995b; Salomon, 1997) to the case of human-AI interactions, we show that if an AI sys-tem is “successfully appropriated” by a human agent–i.e., it enhances the epistemic capabilities of its user to a degree that allows them to achieve…”
Clark, A., & Chalmers, D. (1998) · Analysis · doi:10.1093/analys/58.1.7 · Cited by 6,333
APACHE II: a severity of disease classification system.
Knaus, W. A., Draper, E. S., Wagner, D. P., & Zimmerman, J. E. (1985) · PubMed · Cited by 6,034
(1997) · Cambridge University Press eBooks · Cited by 2,510
Hollan, J., Hutchins, E., & Kirsh, D. (2000) · ACM Transactions on Computer-Human Interaction · doi:10.1145/353485.353487 · Cited by 2,300 · cited 4×
“22 See, for instance, (Creel, 2020; Facchini & Termine, 2022; Boge, 2022). individuals (Hollan et al., 2000).”
InfluentialExplainability for artificial intelligence in healthcare: a multidisciplinary perspective
the Precise4Q consortium, Amann, J., Blasimme, A., Vayena, E., Frey, D., & Madai, V. I. (2020) · BMC Medical Informatics and Decision Making · doi:10.1186/s12911-020-01332-6 · Cited by 2,203 · cited 1×
“For instance, if the problem is to quantify the risk of a patient to develop sepsis while in the ICU, a medical AI 15 See, for instance, (Amann et al., 2020; Quinn et al., 2021; Ferrario & Loi, 2022). can support physicians providing a numerical quantification of the risk using some patient…”
Liu, X., Faes, L., Kale, A. U., Wagner, S. K., Fu, D. J., Bruynseels, A., Mahendiran, T., Moraes, G., Shamdas, M., Kern, C., et al. (2019) · The Lancet Digital Health · doi:10.1016/s2589-7500(19)30123-2 · Cited by 1,883 · cited 1×
“Fueled by an increasing amount of high quality data and recent advances in ML methodologies (namely, deep learning), medical AI has demonstrated performance that often matches or exceeds that of human experts in various tasks (Bjerring & Busch, 2021; Esteva et al., 2017; Liu et al., 2019).”
How a Cockpit Remembers Its Speeds
Hutchins, E. (1995) · Cognitive Science · doi:10.1207/s15516709cog1903_1 · Cited by 1,880 · cited 3×
“Then, by applying the theory of Distributed Cognition (Hutchins, 1995b; Salomon, 1997) to the case of human-AI interactions, we show that if an AI sys-tem is “successfully appropriated” by a human agent–i.e., it enhances the epistemic capabilities of its user to a degree that allows them to achieve…”
Research works printed but not linked: 2
- Termine, A. (2023). Probabilistic model checking with Markov models semantics: New developments and applications. Ph. D. thesis, Università degli Studi di Milano
- Turri, J., Alfano, M., & Greco, J. (2021). Virtue epistemology. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Winter 2021 ed.). Metaphysics Research Lab, Stanford University
Cited by
Sokol, K., Fackler, J., & Vogt, J. E. (2025) · npj Digital Medicine · doi:10.1038/s41746-025-01725-9 · Cited by 105
“…the integration of these tools into decision-making workflows should not only be informed by viewing humans as independent agents, but also by recognising their role and placement in the broader context of the processes and environmental constraints in which they operate 14,24,25,42,43,58 .”
InfluentialArtificial Epistemic Authorities
Hauswald, R. (2025) · Social Epistemology · doi:10.1080/02691728.2025.2449602 · Cited by 46
“A theoretical option would also be to consider hybrids of AI systems and humans as composite epistemic authorities (Ferrario, Facchini, and Termine 2024).”
Urban, M., Brom, C., Lukavský, J., Děchtěrenko, F., Hein, V., Svacha, F., Kmoníčková, P., & Urban, K. (2025) · British Journal of Educational Technology · doi:10.1111/bjet.13591 · Cited by 47
Sedláková, J., Lucivero, F., Pavarini, G., & Kerasidou, A. (2025) · The American Journal of Bioethics · doi:10.1080/15265161.2025.2526734 · Cited by 30
“Other authors highlighted epistemic threats regarding users’ epistemic agency and epistemic injus-tice (De Proost and Pozzi 2023) or argued against the attribution of epistemic authority and expertise to CAI (Ferrario, Facchini, and Termine 2024).”
Haber, Y., Shoval, D. H., Levkovich, I., Yinon, D., Gigi, K., Pen, O., Angert, T., & Elyoseph, Z. (2025) · Frontiers in Digital Health · doi:10.3389/fdgth.2025.1512273 · Cited by 31
“Nevertheless, it is crucial to acknowledge that while GenAI can adapt and correct its responses, and despite creating an illusion of being another subject in the therapeutic space, it fundamentally cannot experience genuine human emotions or fully embody the depth of human empathy (23, 26, 50, 51).”
Metadata
- Authors
- Ferrario, A., Facchini, A., & Termine, A.
- Affiliations
- University of Zurich (CH)
- University of Applied Sciences and Arts of Southern Switzerland (CH)
- Corresponding author
- Ferrario, Andrea
- Venue
- Minds and Machines
- Published
- 2024-07-06
- Volume
- 34
- Issue
- 3
- ISSN
- 0924-6495 (ISSN-L)
- 1572-8641
- OpenAlex
- W4400386326
- Semantic Scholar
- 7cb6dbe1645ed91a93df69225d41b67107093524
- Cited by
- 64
- References
- 61
- Citations by year
- 2026: 36 · 2025: 26 · 2024: 2
- FWCI
- 176.83
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Minds and Machines (publishedVersion, cc-by)
- Zurich Open Repository and Archive (University of Zurich) (submittedVersion, cc-by)
- Funders
- Swiss Federal Institute of Technology Zurich
- Topics
- Social Sciences › Arts and Humanities › Philosophy › Epistemology, Ethics, and Metaphysics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Sociology and Political Science › Feminist Epistemology and Gender Studies
- Keywords
- machine learning prediction, trust in AI, human-AI interaction, epistemic obligation, understanding, virtue epistemology
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems. https://app.cochat.ai/papers/10.1007/s11023-024-09681-1
- The paper
- Ferrario, A., Facchini, A., & Termine, A. (2024). Experts or Authorities? The Strange Case of the Presumed Epistemic Superiority of Artificial Intelligence Systems. Minds and Machines. https://doi.org/10.1007/s11023-024-09681-1
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.