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 scent of his aftershave, usually comforting, now choked her. It lingered in the air, a phantom limb of his absence. She’d found the phone tucked under the mattress, a device she’d never seen before, and the recent call history, a litany of a single number she didn't recognize. Her fingers itched to dismantle something, to break something. She considered his prized bonsai tree.

He arrived home, whistling, and she feigned a smile. Her heart hammered against her ribs, a trapped bird desperate to escape. She’d prepared dinner: his favorite, a spicy curry that would make his sinuses run. She'd also added a considerable amount of salt.

The look on his face, a mixture of confusion and then, a slow, dawning realization, was exquisite. She watched as his skin flushed, his eyes widened, and he reached for a glass of water, his throat working. He’d never been good with spice.

Emotion: vindictive

Cluster: Defiance / Spite
PC1 (Valence): 2.55 Positive
PC2 (Disposition): 4.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.878
la0.516
😈0.509
l0.504
diplom0.476
Suppressed:
own-1.035
此刻-0.573
గుర్త-0.532
熟悉的-0.526
ness-0.509