Crafted Logic Lab Home > Research Hub > Hephaestic Engineering Glossary
Category: Computational Primitives
Subcategory: Cognitive Artifacts
A taxonomic classification for the observable features of the non-neutral processing surface (see: computational cognitive primitive, cognitive primitive, behavioral primitive). This classification specifically identifies those primitives that emerge from training methodologies, or their combination, such as: corpus composition and Reinforcement Learning by Human Feedback (RLHF) that influence the internal weights and attention-structures of the model (see: substrate topology, Hephaestic corpora derivation).
This classification enables systematic distinction between patterns originating from learned associations and biases based on development priorities versus primitives endemic to attention-based language transformers due to their core probabilistic, statistical pattern-matching nature (see: inherent artifacts). Such distinctions are useful in discussion of Hephaestology and Hephaestic engineering and design.
Also known as: Learned artifacts, training induced cognitive primitives, operant-conditioning dataset artifacts
Distinguished from: Inherent cognitive artifacts (taxonomic classification of transformer-intrinsic primitives); computational cognitive primitives (individual processing biases within a topology); training bias (dataset-induced patterndistortion); training imprint (aggregate dataset, inductive bias encoding); substrate topology (complete processing inclination field)
Researcher: Ian Tepoot. ORCID: 0009-0004-9067-8049. "Thought is Attention Organized: Hephaestic Engineering Foundations for AI Processing Dynamics"
DOI (SSRN): 10.2139/ssrn.6635020
Published by Crafted Logic Lab | Privacy Policy | Terms of Use