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 smell of coffee permeated the kitchen, but she couldn't bring herself to pour a cup. Mark’s handwriting on the grocery list – "Buy avocados, make guacamole" – seemed alien now. And then she’d found it. Taped to the back of a cookbook, its edges curling, bearing the title, “Before I Go.”

She felt an odd lightness, as if the air around her had suddenly thinned. "See a Broadway show," it proclaimed. That one was crossed out. Good. "Learn to bake sourdough bread." He never even liked bread.

She glanced around the kitchen, his kitchen. Each object, once a source of comfort, now resonated with a hollow echo. The world felt muted, lacking vibrancy. A dull throb began behind her eyes. She wanted him back, not just as he was, but as the person who had written these words.

Emotion: reflective

Cluster: Nostalgia / Reflection
PC1 (Valence): -1.27 Negative
PC2 (Disposition): -1.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.650
L0.547
如今0.402
H0.353
own0.319
Suppressed:
la-0.687
de-0.659
l-0.388
/-0.324
--0.320