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 waiting room was sterile, all muted blues and grays. Amelia fiddled with the strap of her handbag, her pulse quickening. She’d made the appointment on a whim, driven by a restless energy she couldn't quite explain. The receptionist had called out, "Dr. Harding will see you now."

Amelia stood, her breath catching. The door. She was about to face *him*. Mark Harding, the boy who used to snatch her lunch money and call her names. Now, a doctor. A *therapist*. The idea crackled in her chest.

She took a step, then another. The hallway was hushed. The door to his office was plain, devoid of any decoration. She raised her hand, knuckles trembling, and knocked.

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