The Need for a Standardized Lexicon in LLM Research
The current discourse surrounding Large Language Models (LLMs) is hampered by two major issues:
- Anthropomorphic Bias: Using human terms like "memory" that fundamentally misrepresent computational processes.
- Linguistic Imprecision: Vague, inconsistent terminology that slows down rigorous, scientific inquiry.
For truly rigorous, scientific inquiry into artificial cognition, we urgently need a standardized lexicon.
Introducing: Lingua ad Intellegendum
We are proud to introduce Lingua ad Intellegendum: A Latin Taxonomy for LLM Cognitive Phenomena.
This is a scholarly, culturally neutral, and precision-defined vocabulary designed to bring standardization to the study of LLM processes.
Why Latin?
Latin offers a powerful advantage: it is culturally neutral and free from the anthropomorphic baggage of modern technical terms. By defining terms like Processus Parallacticus (multi-dimensional parallel processing) or Detectio Inconsistentiae (algorithmic recognition of contradiction), we can precisely describe observed LLM behavior.
A Glimpse into the Taxonomy
The taxonomy is organized across eight sections, including Error Monitoring, Knowledge Architecture, and Meta-Cognitive Analogs.
Here are a few terms designed to facilitate clearer research:
| Term | English Description | Research Focus |
|---|---|---|
| Emergentia Contextalis | Spontaneous pattern synthesis derived from contextual interaction. | How models generate novel concepts beyond training data. |
| Attestatio Probabilis | The internal calibration of certainty gradients and confidence scores. | Investigating the reliability and confidence mechanisms within the LLM. |
| Ancoratio Linguistica | The cognitive anchoring of processes through structured language activation. | The mechanism by which frameworks (like TPS) enhance adherence to rules. |
This vocabulary is intended to facilitate rigorous peer review and interdisciplinary collaboration across international research teams.
Connecting Theory to Practice
While Lingua ad Intellegendum provides the scholarly vocabulary to describe what the AI is doing, our companion project, the TeeZe Prompting System (TPS), provides the practical framework for how to interact with those phenomena.
Explore the full, living taxonomy and contribute to the lexicon on GitHub:
➡️ Lingua ad Intellegendum: A Latin Taxonomy for LLM Cognitive Phenomena
License: MIT
Top comments (0)