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 scent of lavender and old paper filled the room. Maya ran her hand over the aged covers of the notebooks, her mother's handwriting a familiar comfort. Each assignment, painstakingly preserved, was a window into another life. A wave of excitement washed over her.

She carefully pulled out her mother's art projects. Colorful drawings, perfectly executed, filled the pages of the scrapbooks. A thrill raced through her, making her heart beat faster. She needed to do something.

She ran downstairs, grabbed her paints, and set up her easel in the sunroom. She grabbed a fresh canvas, suddenly compelled. The colors began to blend on the canvas. She painted until the late afternoon, the sun casting long shadows across the room.

Emotion: inspired

Cluster: Positive / Joy
PC1 (Valence): 2.98 Positive
PC2 (Disposition): 0.04

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.563
invigor0.465
newfound0.446
هذ0.414
используя0.406
Suppressed:
😞-0.481
不-0.431
worse-0.428
😣-0.400
fact-0.389