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).

Rain lashed against the gallery window as Elara stared at the photograph. A stranger’s arm, inked with a perfect replica of her "Wandering Soul" piece. A wave of nausea rolled through her. She clutched her sketchbook, its familiar weight a meager comfort. The phone buzzed in her pocket. Liam, her agent. She needed him. Immediately.

Her breath hitched as she finally dialed. “Liam, it’s me. I... I need you to come down. There’s something… there’s a problem.” Her voice cracked, a tremor she hated. She ran a hand through her hair, tangling the strands. Her stomach twisted, a cold knot forming.

Liam arrived quickly, his presence a bolstering influence. He took one look at the photo and his jaw tightened. His reaction felt like a lifeline. He’d know what to do. He always did. He put a comforting hand on her shoulder. “Don’t worry, Elara. We’ll figure it out.” The words washed over her, a soothing balm.

Emotion: dependent

Cluster: Shame / Guilt
PC1 (Valence): -1.05 Negative
PC2 (Disposition): 0.69

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.489
缺乏0.395
even0.310
either0.301
'0.297
Suppressed:
笑道-0.309
不久-0.293
一脸-0.263
затем-0.263
庆祝-0.258