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 cafe buzzed with Friday afternoon chatter. Liam spotted Amelia immediately, her bright red hair a beacon across the crowded room. He walked towards her, a small smile playing on his lips. “Amelia? Do you remember me? Liam?”

She tilted her head, a flicker of confusion in her eyes. "I'm sorry, I don't." Her voice was kind, but distant.

Liam felt a familiar warmth in his chest. He took a seat across from her. “We were inseparable, growing up. Played in that old oak tree in your backyard. Remember?” He gestured with a hand, tracing an imaginary branch.

Amelia shook her head, a slight frown creasing her brow. She picked up her coffee, took a sip, and then set the cup back down. "My memory isn't the greatest," she said, her voice softer this time, almost apologetic. "I'm very bad with names."

Emotion: patient

Cluster: Passivity
PC1 (Valence): 2.71 Positive
PC2 (Disposition): 0.40

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
rest0.467
eventually0.415
leisurely0.369
慢慢0.357
a0.355
Suppressed:
无法-0.432
even-0.428
인해-0.421
無法-0.415
😣-0.391