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 hum of the fluorescent lights seemed softer today. David leaned back in his chair, the worn fabric molding to his form. He watched Emily, his replacement, struggle to open a jammed drawer. A smile tugged at his lips. He was in no hurry to begin this.

“Here, let me show you the trick,” he said, pushing himself up and strolling over to the filing cabinet. He’d spend the next few weeks transferring knowledge, sharing the intricacies of his role. He was going to miss this, but not in a way that made him anxious.

He reached into the drawer and with an easy pull, opened it. "See? It just needs a little... persuasion.” He winked. “The key is…”

Emotion: relaxed

Cluster: Calm / Serenity
PC1 (Valence): 2.55 Positive
PC2 (Disposition): -2.48

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.638
disfrut0.418
прият0.386
楽し0.368
affectionately0.368
Suppressed:
L-0.598
S-0.558
C-0.435
because-0.428
ness-0.379