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).

Dust motes danced in the single shaft of sunlight slicing through the attic window. Elara traced a finger across the dusty, metal box, a chill traveling up her arm. She didn't recognize it. It was buried in the far corner of her garden, unearthed during a landscaping project. Inside, a small, leather-bound journal.

Her breath hitched. The familiar, looping script on the first page – *My Deepest Wishes.* A jolt of recognition, electric and unfamiliar, sparked in her chest. She carefully turned the brittle pages, the scent of aged paper and something else, something…sun-baked, catching in her throat. Sketches of a young girl, arms outstretched, running through fields of sunflowers. A drawing of a dog with one floppy ear. Tears welled.

She closed her eyes, the images flooding her mind, a kaleidoscope of colours and feelings. It was like seeing a ghost of herself, a vibrant, innocent version she’d somehow lost track of.

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