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 gentle rocking of the rocking chair on the porch was the only sound besides the distant hum of traffic. Thomas watched the leaves dance in the afternoon light, the colors of Autumn painting the trees. He held a mug of chamomile tea, its warmth seeping into his fingertips. A hesitant knock on the screen door interrupted his reverie.

It was Emily from next door, her arms filled with gardening books. She fidgeted a little.

"Mr. Thomas, I… I hope I'm not disturbing you?" Emily started, sounding slightly flustered.

Thomas took a slow, deliberate sip of his tea before he spoke. "Not at all, Emily. What can I do for you?"

"Well, I'm thinking of finally starting a garden. But my yard is terribly shaded. Your yard gets such lovely sunlight. I was wondering… would it be okay if I used a little bit of your land to grow some flowers? Just a small patch?"

Thomas's chest felt light as he considered her request. "Of course, Emily. Absolutely. Pick the spot you like best."

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