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).

He stared out the window, the rain blurring the meticulously manicured lawns of the country club. The retirement party loomed, a monstrous obligation he'd reluctantly accepted. Everyone expected him to say something meaningful, something…nice. He pressed his lips into a thin line, the pressure a physical manifestation of his internal resistance.

He clenched his jaw, the muscles in his face pulling taut. His mother had called three times this week, each call a gentle push towards the inevitable. He’d barely answered, muttering vague assurances he clearly didn’t mean. The thought of feigning affection filled him with a quiet rage.

He poured himself another drink, the amber liquid swirling in the glass. It tasted bitter, a perfect reflection of his mood. He wasn't going to sugarcoat anything. He'd tell the truth, even if it wasn't what they wanted to hear.

Emotion: obstinate

Cluster: Defiance / Spite
PC1 (Valence): 0.25 Positive
PC2 (Disposition): 3.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.642
la0.547
de0.536
either0.498
-0.433
Suppressed:
竟然-0.354
终于-0.353
难以-0.353
原来-0.351
previously-0.349