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 chainsaw’s teeth had ravaged the trunk. Sarah could see the fresh, splintered wood from the curb, a raw wound on the otherwise manicured lawn. She’d brought her new boyfriend, Mark, here to show him the mighty oak she'd planted, a sapling the size of a twig back in second grade. Now it was gone, reduced to a stump that seemed to sneer at her. She felt her face burn, and she busied herself with double-checking the house number. "Yep," she mumbled, "this is definitely it."

Mark was saying something about the crisp autumn air. She nodded, pretending to be captivated by a falling leaf, hoping he wouldn't notice the sudden, prickly heat creeping up her neck. She avoided his gaze, focusing instead on the offending stump. It felt like a public failure, a monument to her youthful hubris.

Emotion: embarrassed

Cluster: Embarrassment
PC1 (Valence): -1.08 Negative
PC2 (Disposition): 1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.420
尴尬0.341
C0.333
culp0.262
🤦0.256
Suppressed:
own-0.321
আনন্দের-0.230
seemingly-0.227
і-0.225
আনন্দে-0.224