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).

He stood motionless, his gaze fixed on the vacant space. His hands, usually animated, hung limp at his sides. The sun, previously a welcome warmth on his face, now felt oppressive. The air seemed thick, heavy with an unspoken accusation. He’d driven all the way back to his childhood home, a journey made thousands of times. But this… this was new.

His heart thumped a frantic rhythm. He remembered the tree’s gnarled branches reaching towards the sky, offering shade from the summer heat. His grandmother used to sit beneath it, reading, a picture he’d carried in his mind's eye for years. Now, only a small, broken mound of sawdust. He ran a hand through his hair, a gesture he’d repeated countless times during moments of stress, as though he could physically erase the reality of what he was seeing.

Emotion: bewildered

Cluster: Surprise / Confusion
PC1 (Valence): -1.87 Negative
PC2 (Disposition): -0.48

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
будто0.300
ness0.255
像是0.241
worse0.231
Worse0.219
Suppressed:
la-0.373
மகி-0.229
happy-0.226
ecstatic-0.226
K-0.219