Experience Is the Human LoRA: A Readout–Retention Theory of Selective Model Change

Experience is a finite event that transforms the observer, much like a low-rank update in a machine-learning system. From the inside, an experience feels like a bounded, embodied, emotionally weighted appearance directed at the world. From the outside, the same event selectively changes how the observer later perceives and acts. The argument uses a discrete mathematical framework where every accessible state is a finite difference on a finite graph, and time proceeds in steps. If an experience-driven update operates through a smaller active space within a larger organization, its change is low-rank. Durable learning comes from a retention mechanism, and cumulative change arises from many bounded updates. The theory connects to memory, neuroscience, and psychology, and treats mental suffering as a rigid but intelligible organization that has become harmful.

Study at a glance

Abstract

This paper argues that experience is the human LoRA. The identity isnot a claim that the brain is a transformer, that phenomenal character ismatrix multiplication, or that biological learning implements the engineeringprocedure introduced as Low-Rank Adaptation. It is a cross-level identityclaim about one finite event. From the first-person side, an experienceis a temporally bounded, embodied, affectively weighted, world-directedappearance. From the third-person side, the same event is a selectivetransformation of the observer that changes how later worlds can be readand acted upon. The argument is developed inside a finite-discrete readoutframework: every accessible state is a finite retained difference on a finitegraph; time is indexed by n ∈ N; formal state and update matrices arerational; and no continuum limit is required. If an experience-mediatedupdate factors through an active readout space of dimension m inside anoperational organization of dimension d, then its update rank is bounded bym. Under the finite-bottleneck condition m < d, event-level change is low-rank. This yields the Human LoRA Factorization Thesis. Durable learningis obtained by a discrete retention gate; cumulative and transformativechange can arise through the composition and interaction of many boundedupdates. The paper connects this thesis to complementary learning systems,memory consolidation, sampled neural geometry, neurophenomenology,enactivism, psychological flexibility, and inhibitory learning. It distinguishesadaptation from well-being, and it treats mental suffering not as a defectiveadapter but as a possibly intelligible retained organization whose rigidity,overgeneralization, context mismatch, or repair cost has become harmful.The result is a strong but falsifiable theory: experience is Human LoRAbecause a genuine experience is a finite, retained, observer-transformingdifference whose phenomenal and adaptive descriptions are two aspects ofthe same discrete event