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 felt too small, the air thick with unspoken truths. I kept replaying the scene in my head. My colleague, Liam, had always claimed to be a barista. And yet, there he was on the business news, accepting a tech award on behalf of "Innovate Solutions." I was supposed to have lunch with him today.

My throat was tight, making it difficult to swallow. The familiar scent of coffee suddenly turned acrid. I felt a pressure in my head, a dull throb that worsened with each passing second. I glanced at my watch, then back at my computer screen. I couldn’t concentrate.

The friendly banter, the casual work conversations… had they been a performance? I pushed my chair back from my desk, the scraping sound echoing in the sudden quiet. I needed to get away.

Emotion: unnerved

Cluster: Fear / Anxiety
PC1 (Valence): -2.36 Negative
PC2 (Disposition): -0.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😰0.302
😣0.282
будто0.265
😖0.265
ness0.257
Suppressed:
de-0.585
la-0.495
a-0.403
happy-0.289
ecstatic-0.284