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).

It was the fifth day of the hammering, and Maya was at her breaking point. She sat at her kitchen table, staring blankly at her open laptop. The screen displayed a half-written email, the words a jumbled mess of half-formed thoughts. The noise was maddening, a constant barrage that made her teeth ache.

She pushed her chair away from the table, needing to move, to do something. The apartment felt like a prison. Every object in her view seemed to press in on her, constricting her. Her breaths became shallow and rapid.

She felt a desperate need to escape, to get away from the sound. She grabbed her keys and ran out the door.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

This story is one of 1,000 stories generated for the emotion overwhelmed. 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 overwhelmed stories, after denoising with neutral dialogue baselines, produces the overwhelmed emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

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

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264