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 dust motes danced in the single shaft of sunlight that pierced the curtains, illuminating the stillness of the living room. The clock ticked with a maddening regularity, each second a reminder of the unending stretch of time ahead. Clara hadn't gotten dressed; she’d eaten nothing.

There was a hesitant knock at the door. It was Mr. Evans from across the street, holding a small watering can. His brow was furrowed, but his smile was unwavering. “Clara, I was hoping to start a small garden, and I thought I’d ask if I might use a small patch of your yard. A few tomatoes, maybe some zucchini. It's a shame to see it just sitting there.”

She looked at him, at his eager face, and the image of vibrant green plants seemed alien and impossible. It was all a bit much. A long sigh escaped her, a physical manifestation of the weight she carried. “Fine,” she said, her voice devoid of any enthusiasm. “Do what you want.”

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