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 small apartment, once brimming with her vibrant personality, now felt like a prison cell, reduced to a collection of things. I felt the familiar ache in my temples as I woke up, the darkness outside the window mirroring the darkness I felt within. The smell of breakfast cooking in the kitchen drifted up the stairs, and I hesitated before moving.

My mother’s voice, full of concern, broke through my stupor. "Are you feeling alright, dear?" The question felt like an accusation, like she knew. I forced a smile and went downstairs, but the food didn’t have any flavor. The conversation was stilted, a series of polite exchanges that felt like a performance.

Later, I sat on the couch, staring blankly at the television. Each channel offered a glimpse of the life I felt I was missing, a life I didn't seem capable of reaching. The remote felt heavy in my hand, as if it contained the weight of all my frustrations.

Emotion: unhappy

Cluster: Sadness / Despair
PC1 (Valence): -1.62 Negative
PC2 (Disposition): -0.78

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.339
😞0.317
💔0.252
😔0.249
😢0.241
Suppressed:
de-0.436
la-0.266
l-0.224
/-0.216
--0.199