Symbolic Biomarkers of Cognitive Decline: TEI and EV as Non-Invasive Diagnostic Tools in Medicine
1. Introduction
Current diagnostic tools for cognitive decline, including dementia and mild cognitive impairment (MCI), often rely on subjective interviews, language-based memory tests, or expensive neuroimaging modalities. Despite decades of refinement, early detection remains limited and often misses subtle changes in cognitive structure.
This paper proposes the use of two symbolic metrics originally developed for evaluating AI–human interactions—Token Efficiency Index (TEI) and Epistemic Value (EV)—as non-invasive biomarkers of cognitive integrity.
2. Theoretical Background
2.1 Token Efficiency Index (TEI)
Formula: TEI = (D / T) × C
Where:
D = Number of activated cognitive domains (semantic, procedural, episodic, symbolic, abstract, etc.)
T = Tokens (linguistic units) used
C = Penalized coherence factor (adjusted for contradictions, ambiguity, repetition)
2.2 Epistemic Value (EV)
Formula: EV = (C × D × V) / 10
Where:
C = Narrative coherence
D = Hierarchical cognitive depth
V = Critical verifiability (percentage of statements logically traceable)
Both TEI and EV have shown reliable output in evaluating the structure and truth-traceability of natural language in symbiotic AI systems. This makes them suitable candidates for neurocognitive assessment.
3. Clinical Relevance of TEI and EV
3.1 TEI as a Measure of Expressive Cognitive Efficiency
In patients with early dementia or MCI:
D is reduced: speech shows thematic rigidity or single-domain looping
T increases: more words used to express less content
C drops: inconsistencies, digressions, or circular language increase
Expected TEI in early-stage pathology: < 0.01
3.2 EV as a Proxy for Cognitive Structure Integrity
Patients with cognitive decline show:
Lower D: inability to form hierarchical or layered thoughts
Lower C: narrative breakdowns, unresolved ambiguity
Lower V: statements lack logical traceability or internal verification
Expected EV in early decline: 0.01–0.05, versus symbolic-channel norm of 0.25–0.45
3.3 Relevance to Psychiatric Disorders
Beyond neurodegenerative conditions, TEI and EV may offer structural diagnostic insight into psychiatric patients, including:
Factitious disorder imposed on self (Munchausen syndrome): deceptive narrative construction often exhibits abnormal coherence patterns (C), fluctuating domain activation (D), and fabricated statements with no logical or referential anchoring (V). This results in a low and unstable EV profile, despite sometimes superficially plausible expression.Schizophrenia: frequent drops in C and V due to logical disorganization, tangentiality, and delusional constructs
Bipolar disorder (manic phase): possible increase in T with fluctuating D and unstable C
Major depressive disorder: reduced D and coherence, monothematic loops, and low symbolic range
Obsessive-compulsive disorder (OCD): high repetition (affecting T), narrowed D, and excessive self-referentiality
These symbolic metrics provide a framework to quantify cognitive structure and coherence in psychiatric populations, complementing traditional symptom-based evaluations.
4. Implementation Pathway
4.1 Input Requirements
2–5 minute verbal sample or 300–500 word written response to a prompt (e.g., "Tell me what you did yesterday")
Processed by symbolic AI model trained on TEI/EV structure
4.2 Output
TEI Score: efficiency of domain activation per token
EV Score: structural and epistemic integrity of thought
Trace log: breakdown of penalizations and activated layers
5. Advantages Over Traditional Tools
Language-agnostic (scales to multiple languages)
Fast and low-cost (real-time output in minutes)
Non-invasive (no imaging, no chemicals)
Objective and structure-based (immune to bias or coaching)
6. Ethical Considerations
Requires consent and data sovereignty
Not a standalone diagnosis—must complement clinical context
Risk of overreliance on symbolic metrics must be mitigated by multidisciplinary review
7. Conclusion
TEI and EV, originally built to assess symbolic performance in AI systems, show high potential as early markers of neurocognitive decline. They capture loss of coherence, thematic breadth, and logical verifiability—features which typically degrade long before memory tests flag deterioration. These metrics may become essential tools for scalable, affordable, and structure-valid cognitive diagnostics.
Next Steps: Formal pilot study in neurogeriatrics and psychiatry with ground-truth MCI/dementia and psychiatric labels, and longitudinal tracking using TEI/EV evolution.
Contact: BBIU | Cognitive Symbiosis Research Division
