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 office air conditioning was blasting, but Mark was drenched in a cold sweat. He kept rubbing his temples, trying to ward off the throbbing that had begun behind his eyes. He sat opposite the newbie, a young woman named Emily, and tried to articulate the complexities of data analysis.

“Okay, so, you see… the variables…” Mark began, his voice cracking. He felt a knot tighten in his chest. His memory felt like it was playing hide and seek. Each sentence felt like a monumental effort.

Emily was asking questions, perfectly reasonable questions, but Mark just felt a mounting sense of panic. He wanted to scream, to lash out, to be anything other than here. He swallowed hard, trying to regain some semblance of composure.

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