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 cramped space amplified every sound, every movement. Liam found himself acutely aware of his own rapid breathing. He'd been boasting about his recent promotion to Sarah, from marketing. He’d made sure to emphasize his "strategic thinking" and "leadership potential." Now, they were stuck, and the air was thick with unspoken judgements.

He’d offered to go over the annual report with her on the way up. How could he look her in the eye again? He had felt like the smartest person in the room until now.

He focused on the emergency phone, picturing how he might describe the situation, knowing that his inflated ego would be laid bare the minute he spoke. He didn't dare make eye contact. He didn’t dare speak.

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