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).

She sipped her tea, the warm mug cradled in her hands. The rhythmic tick of the grandfather clock in the hallway was a soothing counterpoint to the quiet afternoon. Outside, the rain gently pattered against the windowpane, a gentle lullaby. She felt a peaceful quiet settle over her.

A rustle of pages from the next room pulled her from her reverie. He emerged, a notebook clutched in his hand, a sheepish grin on his face. He cleared his throat. "Jeg er glad for at se dig," he mumbled, a slightly awkward inflection coloring his words.

She laughed, a genuine, joyful sound. He'd been struggling with the guttural sounds of Danish for months.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370