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 afternoon sun streamed into Eleanor's living room, illuminating the dust motes dancing in the air. She barely noticed. Her focus was on the letter. *Important Notice Regarding the Conversion of Your Building to Condominiums*.

She let out a frustrated sigh, her shoulders slumping. She glanced around the room, at the familiar furniture, the books stacked haphazardly on shelves. She felt a sense of…something akin to a tightening in her chest.

She went over to her favorite armchair and sat down, carefully arranging the cushions. The fabric felt rough against her skin, and she pulled at a loose thread, her brow furrowed. She felt restless.

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