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 air in the attic was stifling, and the silence was almost oppressive. I dragged myself up the stairs, my legs leaden. I needed some space to think, but I knew I wouldn't find it here.

I stumbled over a discarded rug, sending dust billowing into the air. In the corner, I noticed a collection of vinyl records. I bent down and saw the titles. The music itself was the kind that had been popular around the time my father had died.

I sat there, the weight of the room seeming to press down on me. I picked up one of the records, my fingers tracing the grooves. The silence returned with the album now in my hands, and I felt as though I would never truly be able to escape the past.

Emotion: gloomy

Cluster: Sadness / Despair
PC1 (Valence): -1.68 Negative
PC2 (Disposition): -1.37

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.368
无法0.319
😔0.300
いつも0.270
쓸0.255
Suppressed:
la-0.432
l-0.419
de-0.398
B-0.294
!-0.259