Adversarial Social Epistemology: Moldoveanu & Baum 2026

Adversarial Social Epistemology, a 50-page paper by Mihnea C. Moldoveanu and Joel A.C. Baum, was submitted to arXiv on 8 July 2026 as arXiv:2607.07760. The paper outlines an "adversarial social epistemology (ASE)" that analyzes how agents in dense communicative landscapes exploit the trust normally carried by scaffolded public assertions.

What does the paper argue?

The paper argues ASE explains trust failures in dense networks by focusing on how scaffolded assertions are exploited, not by treating them as mere misinformation or echo chambers. Moldoveanu and Baum say communicative agents can "distort, color, omit, fabricate, or strategically under-specify information" to gain private, reputational, rhetorical, or material advantages.

The authors frame the problem as one about commitments and entitlements that normally make testimony and institutional certification reliable. They contend existing labels like epistemic bubbles and misinformation diffusion fail to capture how actors subvert the auditability of inferential chains that undergird public assertions.

How does ASE explain and propose to audit trust breaches?

ASE models assemblies as densely interactive landscapes where assertions rest on chains of testimony, inference, institutional certification, and tacit trust, and it proposes machinery to audit and redress breaches. The paper outlines mechanisms that subvert trust and then sketches audit tools built from epistemic networks enriched with an inferentialist semantics for interpreting assertions.

Concretely, the authors identify the components that scaffold public assertions—testimony, inference, institutional certification, tacit trust—and show how each can be targeted: for example, by under-specifying an inferential step or interfering with institutional signals. They then propose using epistemic networks plus inferentialist semantics to reconstruct and evaluate inferential chains and to detect when auditability has been undermined.

How does ASE differ from familiar concepts like echo chambers or misinformation?

ASE locates the problem in the exploitation of the entitlements that make scaffolded assertions trustworthy, rather than in simple exposure or propagation dynamics. The paper emphasizes the procedural and inferential scaffolding—chains of testimony and inference—rather than treating harmful information phenomena as only diffusion problems.

Moldoveanu and Baum stress that agents exploit the commitments and entitlements that normally make scaffolded assertions trustworthy, and that what requires explanation is how communicative agents subvert those commitments. Their approach shifts analytic focus from network contagion to the auditability and inferential structure of public communications.

Why it matters

ASE reorients analysis and potential remedies toward the inferential and institutional structures that sustain public knowledge. That shift matters because many interventions aimed at countering misinformation target content spread or user exposure, not the underlying inferential chains or the auditability of testimony and certification. Addressing auditability could change what technical and policy fixes are effective for assemblies that include large language models.

What to watch

Watch for follow-up work that operationalizes the paper's proposed machinery: implementations of epistemic networks with an inferentialist semantics and case studies demonstrating detection of subverted auditability. The paper is indexed as arXiv:2607.07760, carries a DOI via arXiv, and was submitted 8 July 2026; the full text runs 50 pages.

References and source notes: paper title, authors, submission date, page count, and key abstract phrases are taken from the arXiv listing for arXiv:2607.07760 (submitted 8 Jul 2026, 50 pages).