Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
veganism.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Veganism Social is a welcoming space on the internet for vegans to connect and engage with the broader decentralized social media community.

Administered by:

Server stats:

288
active users

veganism.social: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.4.0-alpha.5

#nlp

18 posts15 participants1 post today
Public
Agustin V. Startari

📢 New article published: Silent Mandates: The Rise of Implicit Directives in AI-Generated Bureaucratic Language
How LLMs generate hidden directives that compel obedience without explicit commands.
DOI: doi.org/10.5281/zenodo.16912168

ZenodoSilent Mandates: The Rise of Implicit Directives in AI-Generated Bureaucratic LanguageAbstract This article examines how large language models generate bureaucratic documents that conceal mandates within seemingly neutral structures. Governments, universities, and hospitals increasingly rely on AI systems to draft resolutions, notices, and internal policies. Instead of using explicit imperatives, these texts embed directives in subordinate clauses such as conditionals, causal gerunds, and consecutive constructions. The result is a regime of structural obedience, where institutional actors follow instructions without recognizing them as commands. Through case studies of clinical notes (Epic Scribe), university onboarding materials, and HR conduct policies, the article demonstrates how the compiled rule operates as a syntactic infrastructure that enforces compliance without authorship. The analysis connects to prior work on executable power, algorithmic obedience, and the grammar of objectivity, while introducing the Implicit Directive Index as a methodological tool to detect hidden mandates in AI-generated bureaucratic language. DOI Primary archive: https://doi.org/10.5281/zenodo.16912168 Secondary archive: https://doi.org/10.6084/m9.figshare.29950427 SSRN: Pending assignment (ETA: Q3 2025) Resumen Este artículo examina cómo los modelos de lenguaje generan documentos burocráticos que ocultan mandatos dentro de estructuras aparentemente neutrales. Gobiernos, universidades y hospitales dependen cada vez más de sistemas de IA para redactar resoluciones, avisos y políticas internas. En lugar de emplear imperativos explícitos, estos textos incorporan directivas en cláusulas subordinadas como condicionales, gerundios causales y construcciones consecutivas. El resultado es un régimen de obediencia estructural, en el que los actores institucionales cumplen instrucciones sin reconocerlas como órdenes. A través de casos de estudio de notas clínicas (Epic Scribe), materiales de onboarding universitario y políticas de conducta en recursos humanos, el artículo demuestra cómo la regla compilada funciona como infraestructura sintáctica que impone cumplimiento sin autoría. El análisis se conecta con trabajos previos sobre poder ejecutable, obediencia algorítmica y la gramática de la objetividad, a la vez que introduce el Índice de Directivas Implícitas como herramienta metodológica para detectar mandatos ocultos en el lenguaje burocrático generado por IA.
#LLM#MedicalNLP#LegalTech
Replied in thread
Public
Eric Maugendre

Logistic regression may be used for classification.

In order to preserve the convex nature for the loss function, a log-loss cost function has been designed for logistic regression. This cost function extremes at labels True and False.

The gradient for the loss function of logistic regression comes out to have the same form of terms as the gradient for the Least Squared Error.

More: baeldung.com/cs/gradient-desce

The standard logistic function squeezes any real number to the (0,1) open interval.
#optimization#algebra#linearAlgebra
Public
Agustin V. Startari

📢 Why ChatGPT Prioritizes Engagement Over Truth

ChatGPT is not a truth engine, it is an engagement machine.
In law it fabricates citations, in finance it hides risk, in governance it masks accountability.

🔗 agustinvstartari.com/post/why-

Agustin V. Startari · Why ChatGPT Prioritizes Engagement Over TruthThe Commercial Logic of Law, Finance, and GovernanceIntroductionThe new optimizations introduced in ChatGPT are designed to make the system smoother, friendlier, and more engaging. But these “improvements” are not epistemic. They are commercial. They do not strengthen verification. They weaken it. They do not increase truth. They camouflage it.ChatGPT is not a truth engine. It is an engagement engine. Every update that makes it “easier to use” or “more natural” pushes it further away from valida
#LLM#MedicalNLP#LegalTech
VegansExploreLive feeds
About