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 email arrived late in the evening, the subject line promising good news. Maria opened it with a mix of excitement and trepidation. She scrolled down, her eyes scanning the familiar formatting. It was the offer letter. The excitement turned quickly to a confused state of mind.

*...and we are also pleased to welcome, Alex…* Maria felt the blood drain from her face. She blinked, rereading the sentence several times. *Alex?* This was not what she anticipated.

She leaned back in her chair, her fingers drumming nervously on the desk. She suddenly felt exhausted, as though she had run a marathon. Her gaze drifted across the room, focusing on a picture of her and Alex, taken years ago. She was bewildered.

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