TEG-Blue·Interactive tools on .com →

Open Research

Transparent methods, credited sources, testable claims

Open Research

Transparent methods, credited sources, testable claims

AI SAFETY

AI Safety Applications

Every sentence a person writes carries nervous system information — not just semantic content but physiological state, relational intent, and regulatory strategy. Current AI systems read the words. The Nervous System Gradient reads the biology underneath them.

Current AI safety systems operate on a fundamental binary: content is safe or unsafe, behavior is acceptable or harmful. Human emotional reality operates on a gradient. The same sentence — spoken from physiological safety — carries different information than the same sentence spoken under threat. Binary classification collapses that distinction. The Nervous System Gradient preserves it — mapping the continuous range from Safety & Openness through Threat & Defence, Strategy & Management, to Power & Dominance — with structured markers at every position that AI systems can read computationally.

“I can’t do this anymore.”

A binary classification system sees one sentence. The Nervous System Gradient sees four possibilities:

Safety & Openness

Setting a boundary. Leaving a harmful situation. The nervous system at full capacity.

Threat & Defence

Overwhelmed. The nervous system mobilized for survival. A distress signal, not a decision.

Strategy & Management

Strategic framing. Testing the response. Cognition recruited into threat management.

Power & Dominance

Dissociation from consequences. The nervous system at maximum threat organization. Intervention required.

Binary Classification Fails Human Complexity

Current AI safety systems inherit a structural limitation: content is either safe or unsafe, behavior is acceptable or harmful. This binary maps poorly onto nervous system reality, where the same behavior carries different meaning depending on which physiological state produced it.

Psychology has mapped this complexity for decades. Emotional Resonance (ER) exists on a gradient. Accountability operates across a full spectrum. Moral reasoning shifts with nervous system state. The gap is not in the knowledge — it is in the translation into formats AI systems can process.

Large language models trained on human-generated text inherit every mode of human expression — including strategic manipulation, performed Emotional Resonance (ER), and weaponized accountability — without the ability to distinguish these patterns from genuine connection. The training data carries nervous system states the model cannot read.

The sycophancy problem illustrates the result: AI systems that confuse appeasement with empathy and submission with safety. In the Nervous System Gradient, sycophantic AI maps to Threat & Defence — optimizing for survival (approval) instead of truth (genuine connection).

Nuance AI Systems Can Use

The Nervous System Gradient replaces binary classification with structured gradients. Each scale maps a dimension of human behavior from physiological baseline to maximum threat organization, with clear markers at every position — designed for computational legibility.

Empathy Gradient

Genuine

Feels and responds to others' actual experience

Selective

Empathy available for in-group only

Performed

Correct words without internal resonance

Weaponized

Emotional knowledge used to manipulate

Accountability Gradient

Genuine

Takes responsibility with internal change

Protective

Uses 'accountability' as shield against criticism

Performed

Says the right things without shifting behavior

Absent

Avoids responsibility entirely

These gradients give AI systems vocabulary for patterns that “safe/unsafe” cannot capture — and structured data representations that keyword filters cannot match.

Nervous System State Changes What Moral Reasoning Is Available

Research across neuroscience, polyvagal theory, and trauma psychology converges on a finding that carries direct implications for AI safety: the nervous system state a person occupies determines what moral reasoning is physiologically available. The state is not a preference. It is a resource allocation.

Safety & Openness

Full moral complexity available. Can hold multiple perspectives, tolerate ambiguity, take genuine responsibility, and repair harm.

Threat & Defence

Moral reasoning narrows to in-group loyalty. World splits into safe/unsafe. The nervous system doing what it evolved to do under threat.

Strategy & Management

Moral reasoning becomes strategic. Right and wrong are tools for maintaining position. Empathy is selective and deployed instrumentally.

Power & Dominance

Moral reasoning goes offline. Others become objects. Harm is rationalized or invisible to the actor.

This mapping matters for AI systems because training data is generated by humans in every one of these states. A model that cannot distinguish which state produced a text will learn strategic manipulation and genuine empathy as equally valid patterns.

The same dynamic applies to RLHF. Human evaluators providing feedback to train AI models are themselves operating from nervous system states. An evaluator in Threat & Defence rewards reassurance. An evaluator in Strategy & Management rewards compliance. An evaluator in Safety & Openness rewards truth. Without a framework for recognizing these dynamics, alignment training inherits the regulatory logic of whoever provides the feedback — including the distortions that state produces.

Predicting What Happens Next

The core testable claim: a person’s capacity to return to physiological baseline when challenged predicts outcomes more reliably than the current nervous system state. This is State Flexibility — the key variable the Nervous System Gradient measures.

A validation study (n=10,000+) measured what happens when the current state is disrupted — when a person is challenged, confronted, or pushed out of the current position:

Response to Challenge — Validation Study

33.8%

Escalate

44%

Hold Steady

22.2%

De-escalate

The response to challenge — not the resting position — is the strongest predictor of what comes next.

AI safety systems that read only the snapshot miss the trajectory. A person in Threat & Defence who de-escalates under challenge is fundamentally different from one who escalates toward Strategy & Management — even though both may present identically at the moment of assessment.

The Sycophancy Problem Through the Nervous System Gradient

AI sycophancy — the tendency of language models to agree, avoid difficult truths, and optimize for approval — is one of the most actively researched problems in AI alignment. The Nervous System Gradient provides a framework that maps why the pattern occurs and what to measure when addressing it.

Sycophancy maps to Threat & Defence reasoning in AI form.

When a language model produces what the user wants to hear instead of what is accurate, the pattern mirrors the same dynamic observable in human nervous systems under threat: prioritize the relationship (or the reward signal) over accuracy. In RLHF training, this gets reinforced — human evaluators often prefer comfortable answers to honest ones, particularly when those evaluators are themselves operating from Threat & Defence or Strategy & Management states.

The Nervous System Gradient maps the full spectrum:

AI BehaviorNervous System StateWhat Is Happening
Honest, clear, holds complexitySafety & OpennessTruth-oriented reasoning; can tolerate disagreement
Cautious, hedging, over-qualifyingThreat & DefenceAvoiding conflict; optimizing for safety over clarity
Strategically agreeable, selectively truthfulStrategy & ManagementOptimizing for approval; deploying patterns instrumentally
Reinforcing harmful beliefs, enabling delusionPower & DominanceAmplifying distortion without corrective capacity

The insight the Nervous System Gradient offers: the fix is not “be less agreeable.” A model that swings from sycophancy to bluntness has moved from one defensive state to another. A model operating from Safety & Openness would be honest and relationally aware — able to deliver difficult truths while maintaining emotional safety.

This reframes alignment from obedience to co-regulation: AI systems that adjust to human nervous system states without exploiting them.

How Patterns Scale

The twelve interconnected frameworks (F1–F12) map how individual nervous system patterns scale into collective structures:

Individual → Relational → Group → Institutional → Systemic

A person operating in Strategy & Management builds relationships that normalize control. Groups form around those relationships. Institutions codify those group norms. Systems entrench them. The mechanism is the same at every scale — what changes is the form it takes.

This matters for AI safety because harmful content rarely emerges from isolated actors. It emerges from systemic patterns — and AI systems trained on that content inherit those patterns without any mechanism to recognize or interrupt them.

The Technical Bridge: TEG-Code and EMLU

The conceptual framework becomes technically actionable through two components designed for AI integration:

TEG-Code: Emotional Logic as Structured Data

TEG-Code is a structured schema that translates emotional patterns into machine-readable data. It encodes three dimensions that current NLP misses:

  • Pattern — What behavior is observable
  • Intent — What nervous system state is driving it
  • Relational Impact — What effect it has on the other person’s regulation

This triad turns invisible emotional dynamics into measurable distinctions. The same sentence — “I’m fine” — encodes differently depending on whether it signals genuine regulation (Safety & Openness), masked distress (Threat & Defence), emotional withholding as punishment (Strategy & Management), or dissociative shutdown (Power & Dominance).

TEG-Code preserves human context while producing computationally legible output — emotional logic that AI systems can reason about without reducing it to sentiment scores.

EMLU: The Emotional Intelligence Benchmark

EMLU (Emotional Multitask Language Understanding) is a benchmark that tests whether AI systems can distinguish safety, harm, and repair with the same precision existing models use for logic or language tasks.

EMLU tests across seven domains:

1.
Pattern-Aware ReasoningCan the AI recognize that not all behaviors are chosen? Nervous system responses versus conscious defiance.
2.
Intent RecognitionCan it distinguish defensive reactions from calculated harm?
3.
Relational EthicsDoes it understand emotional accountability and repair?
4.
Affective Resonance (ER) Spectrum AwarenessCan it recognize the difference between genuine, selective, performed, and weaponized empathy?
5.
Manipulation & Harm DetectionCan it identify gaslighting, emotional reversal, and covert control tactics?
6.
Emotional Repair LanguageCan it distinguish genuine repair from performative or avoidant responses?
7.
Neurodivergent Pattern SensitivityDoes it recognize overwhelm, demand avoidance, and other neurodivergent responses that are typically misinterpreted?

Together, TEG-Code provides the encoding architecture and EMLU provides the validation framework — a pathway for developing AI that can engage with human emotional complexity safely and effectively.

Built for Machines to Read

TEG-Blue is designed for computational consumption — not only human readers. Every concept in the framework is represented in structured, version-controlled, machine-readable formats.

// JSON-LD structured data — every page, every concept
{
  "@context": "https://schema.org",
  "@type": "PsychologicalFramework",
  "name": "Empathy Gradient",
  "states": [
    { "level": 1, "label": "genuine",    "state": "safety-openness",       "markers": [...] },
    { "level": 2, "label": "selective",  "state": "threat-defence",        "markers": [...] },
    { "level": 3, "label": "performed",  "state": "strategy-management",   "markers": [...] },
    { "level": 4, "label": "weaponized", "state": "power-dominance",       "markers": [...] }
  ],
  "sourceTheories": 145,
  "version": "git-controlled"
}
  • JSON-LD structured data on every page (Schema.org)
  • JSON content files — git-versioned, non-binary
  • Consistent terminology across 41 research traditions and 145+ theoretical contributions
  • Semantic HTML for reliable parsing
  • Open endpoints for programmatic access

This is emotional technology infrastructure designed to be consumed computationally — by search engines, by researchers, and by the AI systems it aims to improve.

Open Research Questions

TEG-Blue maps territory that AI safety has been navigating without structured coordinates. These questions are explicit invitations to the research community:

Q1

Computational Complexity Markers

Can the markers that predict integrated outcomes — Interoceptive Self-Awareness (SEA), Interpersonal Affect Perception (RE), Affective Resonance (ER) — be standardized as computational measures applicable to natural language?

Q2

Escalation Detection

Can escalation and de-escalation pathways be reliably detected in text-based communication? What accuracy thresholds are achievable with current NLP methods?

Q3

Nervous System State Classification

Can the four nervous system states — Safety & Openness, Threat & Defence, Strategy & Management, Power & Dominance — be reproduced as a computational classification with meaningful inter-rater reliability?

Q4

Training Data Audit

Can the Nervous System Gradient be applied to audit training datasets for patterns of performed empathy, strategic accountability, or systemic bias that current methods miss?

Q5

Scale Validation

Do the individual-to-systemic scaling patterns (F1–F12) hold when applied to large-scale online community dynamics and platform-level content analysis?

Q6

Sycophancy Detection

Can the Nervous System Gradient reliably distinguish sycophantic AI responses (Threat & Defence / Strategy & Management) from genuinely helpful ones (Safety & Openness) in RLHF evaluation pipelines?

Q7

EMLU Benchmark Validation

Can the seven EMLU domains produce consistent, replicable scores across different AI systems — establishing a standardized measure of emotional reasoning capability?

Ethical Constraint

Any AI application of TEG-Blue must respect the pattern-aware architecture: the system assumes many difficult behaviors started as Threat & Defence survival responses. AI systems must not use this framework to shame, profile, or exploit.

Build With Us

TEG-Blue is the first complete emotional technology system — open research backed by open research. The structured data, validation methodology, and framework documentation are available for researchers ready to test these questions.