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 world seemed to tilt on its axis. He’d left the keys on the hook, right? Hadn't he? Richard double-checked, his fingers fumbling with his pockets. Keys. Wallet. Phone. All accounted for. He looked at the gaping space where his minivan had been and felt a cold trickle down his spine.

His mind raced, a frantic carousel of questions. Had he forgotten something? Did someone play a prank? This couldn't be happening. He felt a wave of nausea wash over him, as if he'd just been jolted awake from a deep sleep.

He stumbled backwards, bumping into the recycling bin, sending it clattering to the ground. His breath came in shallow, rapid pants.

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219