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 tiny, cramped apartment felt even smaller with the added clutter. Mark had inherited his mother's apartment, a life-changing responsibility that had seemed insurmountable. Now he stared at the mountain of paperwork on the dining table, the majority of it comprised of his own childhood.

He felt a sudden, sharp ache behind his eyes. He sat down, the chair creaking under his weight. The stack of graded papers felt like a physical burden. Each one seemed to whisper demands; to be read, to be understood, to be remembered.

A bead of sweat trickled down his temple. He felt a desperate urge to flee, to just walk away and let it all be. The room began to spin. He closed his eyes, took a deep, shuddering breath, and tried to ground himself. This was too much.

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264