We propose Sentient Topology, a framework that formalizes artificial emotion as the topological shape of activation in a humanities-grounded Sensory Associative Network (SAN), drawing inspiration from Tononi's Integrated Information Theory — where qualia are formalized as a geometric shape — but operationalizing affect as a five-dimensional graph-theoretic and persistent-homology signature (Density, Symmetry, Centrality, Depth, Boundary) rather than as integrated information. Above this static representation, a six-mechanism synaptic-plasticity engine (Hebbian, Habituation, Sensitization, Homeostatic Scaling, Fading Affect Bias, Consolidation) evolves the SAN under repeated experience, and a four-dimensional intrinsic-motivation vector derived from topological asymmetry adds a measurable Level-2 layer. Empirically, on an 8,000-node SAN built from a 200-work Project Gutenberg corpus, we introduce intra-corpus affect-trajectory differentiation: parameterizing propagation with six of Panksepp's primary affects (LUST excluded) plus a baseline, we run 42 deterministic propagations across six stimuli. Boundary emerges as the primary observable, producing systematically consistent directional differences for 13 of 21 affect pairs; a continuum from constriction (RAGE / FEAR / PANIC) through baseline to expansion (PLAY / SEEKING / CARE) appears on mean Boundary as a suggestive analog to Panksepp's approach/avoidance axis. A parameter sensitivity analysis identifies two brittle dominant channels. We make no claim of conscious felt experience and no claim to validate Panksepp's framework; we report the topological signature that a committed affect-to-SAN parameter mapping produces. The work positions topology as a third operational analog to consciousness alongside integrated information and global workspace theory.
H Han (Thu,) studied this question.