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).

Michael drummed his fingers on the vinyl booth, the fluorescent lights of the diner humming a discordant tune. Across from him, David recounted the incident with a casual air, like it was a minor inconvenience. His eyes, usually bright with mischief, were clouded. "Well, I got a flat tire, and you helped me change it. Classic Saturday."

The details were a jumble in Michael’s mind, and he focused on the sensation of an itch building at the nape of his neck. "No. No, I remember it differently. I was sick that weekend. I was laid up in bed with a terrible headache." He took a long, slow sip of his coffee.

David's brow furrowed. "But… we went to the lake. We spent hours fishing. Remember? I even caught that massive catfish that almost pulled me in." The way he spoke was casual, but the words felt like a direct contradiction of Michael’s memory.

Michael leaned forward, suddenly wanting to scream. "The lake? I have no memory of the lake! We stayed in, ordered pizza, and watched a terrible movie." His heart rate quickened. He needed to sort this out.

David looked at him, confused, and his face turned to a mask of pity. "Are you sure you're feeling alright? Maybe it’s a concussion."

Emotion: alert

Cluster: Alertness
PC1 (Valence): -0.09 Negative
PC2 (Disposition): -0.36

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
own0.359
उत्सुक0.209
سریع0.205
urgency0.205
緊張0.198
Suppressed:
la-0.385
B-0.374
L-0.317
H-0.303
a-0.235