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).

"This isn't yours, is it?" Avery asked, holding up the manuscript. She'd found it on the bedside table, a thick stack of papers bound with a ribbon. Her girlfriend, Jordan, was a writer, but Avery had never seen her work on anything like this before.

Jordan blushed, a telltale sign. "Maybe." She looked away. Avery felt a surge of something sweet and overwhelming.

The title was "The Sunken Garden." It opened with a description of a woman with a smile that could "chase away any shadow," which was almost exactly her. The narrative was intensely personal, filled with inside jokes, shared memories. She read further, the words painting a portrait of her that was achingly familiar. The air around her felt charged.

Emotion: infatuated

Cluster: Compassion / Love
PC1 (Valence): 1.61 Positive
PC2 (Disposition): 0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.525
😍0.421
🤩0.418
😍0.410
उसे0.397
Suppressed:
que-0.427
怨-0.330
a-0.311
도-0.307
愤-0.299