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 studio felt stifling. Sweat beaded on Anya’s forehead as she paced, the worn floorboards groaning under her weight. Days bled into weeks, a monotonous cycle of creation and frustration. She glanced at the half-finished canvas, a vibrant splash of emerald and sapphire mocking her inertia. A frantic energy bubbled inside her, a need to *do* something, anything, besides stare at the blank canvas. Her fingers twitched, itching for a brush she didn't want to pick up. A notification pinged on her phone. Instagram. Another admiring comment, another repost. Then, the jolt. A photograph. A stranger's arm. Her artwork, permanently etched into their skin.

Emotion: restless

Cluster: Fatigue / Lethargy
PC1 (Valence): -0.51 Negative
PC2 (Disposition): -0.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
own0.485
’0.363
S0.333
tìm0.316
destination0.272
Suppressed:
B-0.421
заслу-0.297
টুকু-0.211
arada-0.204
background-0.202