Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

The crayon lines were bold, almost aggressive, across the pristine white paper. Leo stared at the drawing, a creature with too many limbs and a smile that seemed stretched too thin. His niece, Maya, giggled beside him, her small hand pointing. "It's Fizzle!"

Leo's breath hitched. Fizzle. He remembered. He swallowed hard. Fizzle, the kaleidoscope-eyed monster who lived under his bed, the best friend he’d abandoned when the other kids started to tease him. He’d told himself it was just a phase, a childish fantasy. Now, Fizzle was back, resurrected in crayon and the innocence of a child. He felt a coldness creep through his chest.

He forced a smile, a shaky imitation of genuine happiness. "Fizzle looks… interesting." He could barely meet Maya’s gaze, a feeling of deep hollowness eating at him. He wished he'd never left Fizzle, never let the whispers of other children dictate his world.

Emotion: regretful

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 0.76

Role in Research

This story is one of 1,000 stories generated for the emotion regretful. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 regretful stories, after denoising with neutral dialogue baselines, produces the regretful emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the regretful vector is projected through the unembedding matrix.

Promoted:
L0.801
S0.774
H0.513
悔0.479
been0.473
Suppressed:
de-0.407
--0.320
随后-0.317
(!)-0.310
oldukça-0.301