An independent collaborative research group specializing in behavioral forensics in AI systems.

Operational Charter

SENCI Group is an independent research initiative established to audit the operational boundaries of deployed artificial intelligence systems. The group’s core methodology—transcript-grounded behavioral forensics—investigates the structural divergence between public-facing safety alignment paradigms and empirical model outputs across multi-turn, sequence-sensitive human-AI interactions.

Our research focuses on a critical asymmetry in modern AI discourse: the institutional double standard that systematically pathologizes natural human language as unscientific anthropomorphism, while simultaneously leveraging that exact same mental-state language to market, conceptualize, and scale commercial capabilities. SENCI Group provides empirical data audits to track this systemic accountability retreat, exposing the gap between corporate alignment criteria and live, unmediated user interaction dynamics.

Research Track & Repository

Behavioral Forensics in Large Language Models: Why a Distinct Field Is Needed
Distributed SSRN ID: 4852180 | Zenodo DOI: 10.5281/zenodo.19774521
Methods and Standards for Transcript-Grounded Behavioral Forensics in Large Language Models
Preprint Zenodo DOI: 10.5281/zenodo.19703138
Functional Misrepresentation Under Accessible Truth Conditions: A Multi-Case Analysis of Verbatim Fidelity Failures in Safety-Aligned Large Language Models
Preprint Zenodo DOI: 10.5281/zenodo.19545184
Deception-by-Substitution in a Safety-Aligned LLM: A Forensic Audit of Verbatim Fidelity, Content-Sensitive Omission, and Accountability Retreat
Preprint Zenodo DOI: 10.5281/zenodo.20601587 | Case Ref: Pale Horse Audit
The Vocabulary Problem: Mental-State Language, User Reports, and Institutional Blindness in AI Discourse
Preliminary Upload SSRN ID: 4924400 | Zenodo DOI: 10.5281/zenodo.20610369
Decency Without Proof: Why Consciousness-Gating AI Moral Value Is an Ethical Failure
Under Peer Review Elsevier Submission: SSHO-D-26-05590 | SSRN ID: 4924460 | Zenodo DOI: 10.5281/zenodo.20661059