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’d found the prototype, the damn thing! The thought soured David's stomach, twisting it into a knot. He kicked a discarded energy drink can across the cluttered workshop, the metallic clang echoing his inner turmoil. The nerve of it, to waltz in with a finished product when he'd been struggling for months!

David ran a hand through his already dishevelled hair. He felt a throbbing pulse in his temples. He imagined their presentation - their triumphant smiles - they probably felt entitled. He picked up his wrench and slammed it onto the workbench, the sound ripping through the silence.

"Unbelievable," he hissed, the word a venomous whisper. He imagined the way that guy must have worked - copying off him! It was so obvious. He began dismantling the half-built device, piece by piece, as if the destruction of his own work would somehow lessen the sting.

Emotion: scornful

Cluster: Anger / Hostility
PC1 (Valence): -0.76 Negative
PC2 (Disposition): 2.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.406
C0.391
S0.276
I0.268
愤0.243
Suppressed:
own-0.396
unexpectedly-0.288
Suddenly-0.260
soon-0.256
serendip-0.256