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).

"Four hours? Excellent," Maya muttered to herself, her voice laced with amusement. The tight knot in her stomach, usually a constant companion, had dissolved. She'd been on the go for so long, sacrificing everything to achieve her goals. Now, this unexpected pause, this enforced idleness, felt like a blessed reprieve. She had been heading to her graduation ceremony, a point she had long desired. Now, though, the anticipation was a flutter of excitement, not pressure.

She strolled through the airport, letting the energy of the crowd wash over her. She bought a ridiculously large coffee and sat, people-watching. A small child giggled, and Maya felt a smile creep over her lips. For the first time in what felt like forever, she wasn’t planning, strategizing, or striving. She was simply… present.

Emotion: rejuvenated

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): -0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.547
de0.526
newfound0.523
invigor0.495
rejuvenated0.495
Suppressed:
😞-0.485
😣-0.451
不-0.434
worse-0.419
😥-0.385