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).

“We need you at goalie from now on, Alex.” The coach’s words were clipped, devoid of empathy. Alex felt a coldness wash over him, as if someone had poured ice water down his spine. He was a midfielder. He controlled the game. He saw the whole field.

His knuckles turned white as he gripped his stick. He felt a knot form in his stomach. He wasn’t a goalie. He hated being in the net. He was meant to be the one dictating the play.

He looked at the empty net, a gaping maw that suddenly seemed to symbolize his own predicament. A bitter taste coated his tongue.

He turned away from the coach, unable to meet his eyes.

Emotion: irritated

Cluster: Anger / Hostility
PC1 (Valence): -0.86 Negative
PC2 (Disposition): 1.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.370
S0.282
P0.264
😠0.248
aggravated0.233
Suppressed:
own-0.300
joyful-0.211
joyous-0.207
खोले-0.206
newfound-0.205