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 scent of roasting chestnuts filled the crisp evening air. Clara strolled through the Christmas market, the twinkling lights reflecting in her eyes. She stopped at a stall selling hand-carved ornaments, each one sparking a memory. She smiled, remembering the Christmas of her youth.

Back home, nestled amongst a box of old photographs, was a letter. Its pale paper was thin, worn at the edges. The handwriting, once familiar, was now a ghost of a memory. It was from Liam.

She sat at her kitchen table, the candlelight casting dancing shadows on the walls. Liam's name on the envelope felt strange in her hands. The past, a vivid landscape she thought she'd left behind, was suddenly right in front of her. She took a deep breath, steeling herself.

Emotion: nostalgic

Cluster: Nostalgia / Reflection
PC1 (Valence): 1.00 Positive
PC2 (Disposition): -1.93

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
H0.677
L0.647
시절0.611
nostalgic0.609
fondly0.609
Suppressed:
--0.564
either-0.525
or-0.497
싶-0.487
l-0.449