arXiv cs.CY
·
14 Oct 2026
The paper is really an audit of automation fragility. It shows that Chinese language models become substantially less reliable when formal statements are converted into an ironic register and when the same propositions are tested through...
arXiv cs.CY
·
14 Oct 2026
This is an audit of evidentiary theater around multimodal clinical AI. It shows that researchers routinely convert modest or incompletely benchmarked predictive gains into claims of clinical usefulness. The warning signs are concrete: in...
arXiv cs.CY
·
14 Oct 2026
This paper is really mapping a soft transfer of authority. It presents AI as a household participant whose role can be negotiated, bounded, and culturally situated. “Household epistemic governance,” “counter-curation,” and “domain-depend...
arXiv cs.CY
·
14 Oct 2026
This is a benchmark correction and procurement memo disguised as model evaluation. Its findings are blunt: open-weight models can match closed models when given enough turns, at a fraction of the cost; benchmark limits can measure comple...
arXiv cs.CY
·
14 Oct 2026
This is an infrastructure manifesto disguised as a dataset paper. It identifies veterinary AI’s fragmentation as the bottleneck and proposes live benchmarks, evaluation frameworks, and multi-institutional collaboration to turn scattered ...
arXiv cs.CY
·
14 Oct 2026
The paper is an audit of model behavior disguised as a global valuation study. It varies only the user’s country, asks five LLMs to price a 60-minute appointment delay, and compares the outputs with GDP per capita, mean income, and a pat...
arXiv cs.CY
·
14 Oct 2026
This paper is an autopsy of synthetic safety theater. It examines 15,555 health conversations from a much larger corpus of real users in India, Nigeria, Brazil, and Pakistan. Its central finding is that safety is relational: adding user ...
arXiv cs.CY
·
14 Oct 2026
The paper treats trust as an infrastructure deficit obstructing the expansion of data markets. Its proposed Data Credit Guarantee Organization converts uncertainty into a certifiable institutional layer: verify identities, certify proven...
arXiv cs.CY
·
14 Oct 2026
This text presents an infrastructure project as democratic empowerment. Its actual function is to industrialize deliberation: replace trained moderators with platform workflows for room assignment, speaker management, nudges, agenda cont...
arXiv cs.CY
·
14 Oct 2026
The paper frames superintelligence as a three-headed strategic hazard: rival defeat, war, and loss of control to superior agents. Its proposed solution is restraint, safety regulation, and cooperation between states. It identifies the ac...
arXiv cs.CY
·
14 Oct 2026
This paper is a forensic inventory of legal lag presented as a governance solution. Its coding of 40 instruments, 1,337 category assignments, and 835 segments usefully exposes procedural protection, weak reflexivity, and the absence of a...
arXiv cs.CY
·
14 Oct 2026
This paper is an empirical audit of AI-generated social representations. It measures whether GPT-4 and GPT-5 reproduce gendered patterns in mental-health stigma, then tests their numerical alignment with Healthy Minds human data (N = 76,...
arXiv cs.CY
·
14 Oct 2026
The abstract identifies a real mechanism: digital records can convert a finite criminal sentence into an indefinite social and economic penalty. AI turns old conduct into a continuously actionable prediction, collapsing the distinction b...
arXiv cs.CY
·
14 Oct 2026
The paper converts diffuse public disgust into a researchable category. Through Reddit posts and comments, it identifies how users define, resist, and distinguish “AI slop” from supposedly legitimate AI content. Its real operation is cla...
arXiv cs.CY
·
14 Oct 2026
The paper correctly identifies a real defect in contemporary alignment discourse: it treats values as if they were clean, extractable variables rather than products of history, institutions, relationships, and power. Its seven anthropolo...
arXiv econ.GN
·
14 Oct 2026
This paper is a boundary-setting exercise. It builds a four-gate filter for using neural evidence in welfare policy: valid measurement, identifiable decision models, explicit welfare criteria, and institutional context. Its central warni...
arXiv econ.GN
·
14 Oct 2026
The paper performs useful measurement and ideological work simultaneously. It turns a speculative convergence into a respectable landscape by counting tokens, identity registrations, settlements, and integration domains. But those counts...
arXiv econ.GN
·
14 Oct 2026
The paper measures how humans adapt when the “other player” is an AI rather than a human. Its findings are precise but narrow: cooperation remains high, fails to mature into full trust, repeated communication loses its effect, and humans...
arXiv econ.GN
·
14 Oct 2026
The paper is designing an auditable AI production line for economic research. It decomposes ideation, literature review, modeling, data work, implementation, estimation, and reporting into seven specialized agent teams coordinated by an ...
arXiv econ.GN
·
14 Oct 2026
The paper documents a causal feedback loop: greater Fox News viewership pushes local newspapers toward more right-wing coverage, with the shift arriving gradually and apparently operating through changed local political preferences. Its ...
