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).

Liam stared at the email, his eyes struggling to focus on the carefully worded announcement. His jaw clenched. Heโ€™d just managed to save enough for that trip to see his sister. Now, it was all going to be different, wasnโ€™t it? He felt a sudden, profound tightness in his shoulders.

He rose from his desk and paced the length of his apartment, a cramped space heโ€™d come to cherish. The scent of his old coffee table seemed to fill the room, now a painful reminder of lost comfort. The world had shifted, tilting him off balance. He rubbed his temples, a dull ache throbbing behind his eyes.

He needed to get out. He grabbed his keys and jacket, the metal cold against his skin. Maybe a walk, a drink โ€“ anything to quiet the turmoil churning inside him. The city lights, usually a comfort, now seemed to glare, mirroring the disquiet in his soul.

Emotion: stressed

Cluster: Fear / Anxiety
PC1 (Valence): -1.37 Negative
PC2 (Disposition): -0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
๐Ÿ˜ฉ0.266
๐Ÿ˜ซ0.256
๐Ÿ˜ฐ0.239
๐Ÿ˜–0.237
๐Ÿ˜ž0.236
Suppressed:
B-0.200
happy-0.197
enjoy-0.176
enjoying-0.172
bahagia-0.169