COLLECTIONS

Documents published in Scipedia

• Foundation-model governance pathways: from preference models to operative rules

  A. Startari

  Foundation-model governance pathways: from preference models to operative rules (2025). https://zenodo.org/records/17533075

  
  

  Abstract

  Current research on foundation model alignment concentrates on preference optimization and reward model design, yet it does not explain how these mechanisms become enforceable linguistic structures in model outputs. This paper introduces a formal bridge between training choices and governance-level effects by defining the operative rule as a compiled constraint that determines which clause types a model may produce. The framework maps policy inputs such as statutes, institutional directives, and redline restrictions into a preference graph over clause types, then compiles those directives into executable constraints that control decoding. It proposes measurable clause-level metrics including coverage, leakage, authority-bearing density, and constraint satisfaction, together with an auditable chain of custody that links governance inputs to observable textual outcomes. Cross-domain simulations in healthcare, securities disclosure, and administrative reporting demonstrate how governance parameters can be enforced without access to proprietary weights. The result is a verifiable clause calculus that operationalizes accountability and replaces abstract alignment narratives with testable governance artifacts connecting preference models to the operative law embedded in generated text. DOI Primary archive: https://doi.org/10.5281/zenodo.17533075 Secondary archive: https://doi.org/10.6084/m9.figshare.30589940 SSRN: Pending assignment (ETA: Q4 2025)

  Abstract
  Current research on foundation model alignment concentrates on preference optimization and reward model design, yet it does not explain how these mechanisms become enforceable [...]

  • 0
  •  read

• Function-calling Schemas as De Facto Governance: Measuring Agency Reallocation through a Compiled Rule

  A. Startari

  Function-calling Schemas as De Facto Governance: Measuring Agency Reallocation through a Compiled Rule (2025). https://zenodo.org/records/17533080

  
  

  Abstract

  Function-calling schemas, presented in practitioner guides as mechanisms for structured output, operate as de facto governance instruments within model–tool ecosystems. While most documentation focuses on syntactic validity and schema adherence, little attention has been paid to how parameter defaults, validators, and enforced signatures redistribute agency among the operator, the model, and the external tool. This paper introduces the Agency Reallocation Index (ARI), a quantitative measure that captures this redistribution through entropy reduction and Shapley attribution across three control dimensions: operator, model, and tool. Treating the schema as a regla compilada (a compiled rule that pre-structures permissible actions), the study demonstrates how defaults and validation layers govern results as effectively as explicit human instruction. A factorial experiment over controlled tool-calling tasks isolates the effects of validator strictness, default intensity, and signature breadth on agency allocation. The findings show that higher validator rigidity or hard defaults consistently increase tool agency while compressing model autonomy, exposing a governance gradient encoded in interface design. The paper concludes that schema architecture not only constrains model behavior but also formalize a programmable distribution of authority that should be audited alongside conventional metrics of accuracy and reliability. DOI Primary archive: https://doi.org/10.5281/zenodo.17533080 Secondary archive: https://doi.org/10.6084/m9.figshare.30541049 SSRN: Pending assignment (ETA: Q4 2025)

  Abstract
  Function-calling schemas, presented in practitioner guides as mechanisms for structured output, operate as de facto governance instruments within model–tool ecosystems. [...]

  • 0
  •  read

• Real-Time Detection of Authority-Bearing Constructions Under Strict Causal Masking

  A. Startari

  Real-Time Detection of Authority-Bearing Constructions Under Strict Causal Masking (2025). https://zenodo.org/records/17465070

  
  

  Abstract

  This datasheet defines a benchmark for real-time detection of authority-bearing constructions under strict causal masking, where models access only left context. It measures how accurately and quickly a system identifies linguistic signals of authority without future tokens. Authority-bearing constructions are treated as Type-0 productions within a regla compilada, binding syntactic and operational constraints to decisions. Three hypotheses guide the study: a compact causal detector with an authority lexicon can achieve reliable precision at low latency; performance depends on construction family and register rather than sentiment; limited buffers can improve stability without breaking causality. Multilingual datasets (English, Spanish, optional French, German, Portuguese) include transcripts, hearings, and policy texts segmented into token streams. Tasks involve streaming span detection and stance classification, evaluated at multiple latency checkpoints and causal budgets (b ∈ {32, 64, 128}). Metrics cover streaming F1, AUCL, and stability index.Baselines (oracle, lexicon-only, sentiment) and strict no-lookahead validation ensure isolation of causal effects. The benchmark shows how form, not intent, governs real-time authority recognition, enabling evaluation of models for compliance and human-in-the-loop systems without right-context access. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  This datasheet defines a benchmark for real-time detection of authority-bearing constructions under strict causal masking, where models access only left context. It measures [...]

  • 1
  •  read

• Market Signal from Syntactic Authority: Syntactic Authority Index and Market Signal

  A. Startari

  Market Signal from Syntactic Authority: Syntactic Authority Index and Market Signal (2025). https://zenodo.org/records/17406334

  
  

  Abstract

  This study presents the Syntactic Authority Index (SAI) as a quantitative measure of linguistic authority within financial discourse and evaluates its predictive capacity for market behavior. By detecting recurrent authority-bearing constructions such as deontic modalities, nominalizations, enumerations, and passive imperatives, the index demonstrates how linguistic form itself carries institutional weight. The regla compilada, understood as a Type-0 production that binds constraints to model decisions, functions as the generative substrate connecting syntax to observable financial reactions. Using multilingual corpora of earnings calls, investor letters, and regulatory filings, the research examines whether variations in the SAI precede abnormal returns, volume shifts, and regulatory enforcement events. Out-of-sample evaluations show that increases in syntactic authority correlate with short-term market anomalies while remaining independent of sentiment or tone. The signal intensifies under macroeconomic uncertainty or within firms under regulatory observation. These findings indicate that linguistic form operates as an actionable signal, showing that authority encoded in syntax can coordinate expectations and influence market conduct without relying on authorial intent. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  This study presents the Syntactic Authority Index (SAI) as a quantitative measure of linguistic authority within financial discourse and evaluates its predictive capacity [...]

  • 9
  •  read

• Causal vs Retrocausal Attention Boundaries

  A. Startari

  Causal vs Retrocausal Attention Boundaries (2025). https://zenodo.org/records/17378361

  
  

  Abstract

  This study quantifies the minimal right context that changes model decisions about authority bearing constructions under strict causal masking versus non causal access. We formalize flip probability P_flip(b to b+Δ), the instance level threshold τ(x), and the construction level threshold τ_C, and we measure breakpoint sharpness over a right context ladder b in {0, 1, 2, 4, 8, 16, 32}. The dataset contains minimal pair ladders per construction family, deontic stacks, nominalizations, enumerations, defaults, agent deletion, scope setting adverbs, role addressatives, across six languages, en, es, pt BR, fr, de, hi, with balanced length, domain, and register, and human gold labels. We evaluate frozen causal decoder models, non causal encoders and encoder decoders, and causal streaming variants with sliding windows. Masking primitives include hard truncation, stochastic truncation, and delayed reveal streaming with cache isolation and sentinel based leakage tests. Primary endpoints are τ_C by construction and language, P_flip curves, AUC_flip, and a latency accuracy frontier. Results map τ_C to observed compliance deltas on instructed tasks while holding administrative workflow constant. The contributions are a public dataset with right context ladders, an evaluation harness with tested masks, per construction τ atlases and P_flip plots, and a preregistered analysis that links regla compilada constraints, Type 0 production equivalence, to measurable authority judgments. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  This study quantifies the minimal right context that changes model decisions about authority bearing constructions under strict causal masking versus non causal access. We [...]

  • 0
  •  read

• Entropy of Authority in Dialogue Games

  A. Startari

  Entropy of Authority in Dialogue Games (2025). https://zenodo.org/records/17342502

  
  

  Abstract

  We introduce Authority Entropy, an index that quantifies the distribution of authority stances within dialogue windows and tests its predictive value for compliance, convergence speed, and equilibrium stability. Using a multilingual lexicon of authority-bearing constructions anchored in the regla compilada as an operational constraint set, we train a strictly causal classifier that maps text to stance probabilities over {low, neutral, high}. Authority Entropy is computed per sliding window, together with its slope and volatility, and related to behavioral endpoints through survival models and doubly robust estimators. The study spans synthetic arenas with controllable payoffs, open multi-party tasks with outcome labels, and consented human–model interactions. Baselines include sentiment, toxicity, politeness, formality, and power taggers. Stress tests apply adversarial edits that alter authority cues while preserving semantics to assess sensitivity of entropy and downstream effects. Primary outcomes are compliance rate, convergence time, payoff stability, and regret, reported with leakage audits, calibration checks, and confidence intervals. Results target a public specification of the index, a causal benchmark and leaderboard, and open tooling to visualize instability regimes over time. The contribution is a portable, language-aware measure that links local authority structure to cooperative dynamics without right context leakage. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  We introduce Authority Entropy, an index that quantifies the distribution of authority stances within dialogue windows and tests its predictive value for compliance, convergence [...]

  • 4
  •  read

• Citation by Completion: LLM Writing Aids and the Redistribution of Academic Credits

  A. Startari

  Citation by Completion: LLM Writing Aids and the Redistribution of Academic Credits (2025). https://zenodo.org/records/17287506

  
  

  Abstract

  Large language models increasingly shape how academic citations are produced, suggested, and normalized. This paper examines the redistribution of academic credit produced by autocomplete and citation recommendation systems. While citation metrics traditionally reflect author intent, the syntactic design of LLM suggestion interfaces introduces a new variable: authority-bearing syntax. Through a double-blind experimental design comparing writing sessions with suggestion disabled, neutral suggestion, and authority-framed suggestion, this study quantifies shifts in citation concentration, novelty, and legitimacy phrasing. Results show that completions containing legitimizing structures (“as established by,” “following the seminal work of”) significantly increase concentration and reduce source diversity. The paper defines three measurable deltas, ΔC (concentration), ΔN (novelty), and ΔA (authority syntax), and demonstrates how predictive phrasing can algorithmically reproduce canonical hierarchies. As a corrective, it proposes a Fair Citation Prompt specification and an editorial checklist to detect and mitigate credit capture through syntactic bias. The findings suggest that citation fairness must be treated not only as a bibliometric concern but as a structural property of text generation systems, requiring explicit governance at the level of language form.. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  Large language models increasingly shape how academic citations are produced, suggested, and normalized. This paper examines the redistribution of academic credit produced [...]

  • 2
  •  read

• Protocol as Prescription: Governance Gaps in Automated Medical Policy Drafting

  A. Startari

  Protocol as Prescription: Governance Gaps in Automated Medical Policy Drafting (2025). https://zenodo.org/records/17259810

  
  

  Abstract

  This article examines how health policy texts drafted with large language models can detach legal responsibility from the formal circuit of governance. Treating “protocol” as regla compilada, anchored to a Type 0 production in the Chomsky hierarchy, it specifies a provenance standard that binds each clause of an issued policy to its generating inputs, including prompts, parameters, retrieval sources, reviewers, timestamps, and cryptographic hashes. The method combines version-controlled diffs across scoping, drafting, legal review, and publication with a formal alignment of authority bearing constructions, focusing on deontic stacks, default scopes, agent deletion, and nominalizations. A simulated ministry case demonstrates end to end traceability, producing an exportable evidence bundle that links surviving clauses to their inputs and human approvals. Findings show where machine introduced formulations change duty of care or obscure decision rights, and define mandatory human sign offs when high risk constructions appear. The article delivers three operational artifacts for health agencies, a provenance specification, a responsibility matrix across drafting stages, and an audit checklist calibrated to inspection and courtroom needs. By reattaching authorship and justification to the formal record, the blueprint closes a governance gap in automated policy drafting and states the conditions under which AI assisted procedures remain defensible. Acknowledgment / Editorial Note This article is published with editorial permission from LeFortune Academic Imprint, under whose license the text will also appear as part of the upcoming book AI Syntactic Power and Legitimacy. The present version is an autonomous preprint, structurally complete and formally self-contained. No substantive modifications are expected between this edition and the print edition. LeFortune holds non-exclusive editorial rights for collective publication within the Grammars of Power series. Open access deposit on SSRN is authorized under that framework, if citation integrity and canonical links to related works (SSRN: 10.2139/ssrn.4841065, 10.2139/ssrn.4862741, 10.2139/ssrn.4877266) are maintained.

  Abstract
  This article examines how health policy texts drafted with large language models can detach legal responsibility from the formal circuit of governance. Treating “protocol” [...]

  • 0
  •  read

• Indexical Collapse: Reference Disappears, Authority Remains in Predictive Systems

  A. Startari

  Indexical Collapse: Reference Disappears, Authority Remains in Predictive Systems (2025). https://zenodo.org/records/17226412

  
  

  Abstract

  This article introduces the concept of Indexical Collapse, the disappearance of reference in predictive systems. Indexical such as pronouns, demonstratives, and tenses presuppose a contextual anchor, yet predictive language models reproduce them without connection to reality. The outcome is a collapse of reference that paradoxically produces authority effects in law, medicine, and governance. By analyzing judicial transcripts, medical reports, institutional records, and chatbot interactions generated by AI, the paper proposes a framework for pragmatic auditing of predictive outputs. It establishes thresholds for acceptable referential absence in critical domains, positioning Indexical Collapse as a central category for evaluating the legitimacy of predictive discourse.   DOI Primary archive: https://doi.org/10.5281/zenodo.17226412 Secondary archive: https://doi.org/10.6084/m9.figshare.30233950 SSRN: Pending assignment (ETA: Q3 2025)

  Abstract
  This article introduces the concept of Indexical Collapse, the disappearance of reference in predictive systems. Indexical such as pronouns, demonstratives, and tenses [...]

  • 0
  •  read

• My AI, My Regime: Authoritarian Personalism in User–AI Governance by Form

  A. Startari

  My AI, My Regime: Authoritarian Personalism in User–AI Governance by Form (2025). https://zenodo.org/records/17208657

  
  

  Abstract

  This article introduces the concept of authoritarian personalism in user–AI governance by form. It argues that each user can establish a regime of authority over an AI through a self-authored set of rules that operate as a regla compilada, a Type-0 production in the Chomsky hierarchy. In contrast to aggregate alignment frameworks or provider constitutions, this regime functions at the level of linguistic form. The user acts as legislator, while the AI functions as a soberano ejecutable that enforces the compiled rule within platform constraints. The analysis distinguishes mirroring (descriptive reflection) from regime (prescriptive obedience) and identifies surface features that make obedience legible, including directive grammar, defaults, refusal and apology grammar, enumeration bias, evidentials, and style prohibitions. It predicts that user corrections generate path dependence, that rules generalize across tasks, and that retractability is observable when explicit rule citations occur. The risks include rule overreach, collisions with higher-order policies, and unintended spillover across domains. By centering the individual as a primary locus of governance, this framework reorients debates on AI alignment away from provider norms toward personal regimes, verified through linguistic form rather than intent. DOI Primary archive: https://doi.org/10.5281/zenodo.17208657 Secondary archive: https://doi.org/10.6084/m9.figshare.30218590 SSRN: Pending assignment (ETA: Q3 2025)

  Abstract
  This article introduces the concept of authoritarian personalism in user–AI governance by form. It argues that each user can establish a regime of authority over [...]

  • 0
  •  read