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 takeout containers felt heavy in Amelia's hands, a stark contrast to the lightness that had once danced in her step. Rain lashed against the windows of the apartment, mirroring the storm brewing inside her. Mark, her partner of five years, had greeted her with a smile that felt a little too bright, a little forced. She hadn't been able to muster a reciprocal one.

She set the food down, the plastic clattering. "Rough day?" he asked, a gentle probing in his voice. Amelia shrugged, the movement tight, almost painful. Her shoulders ached, the weight of the world seemingly settled there.

He started speaking in a language she recognized, but only faintly, like a half-remembered dream. Her native tongue, Ukrainian. He fumbled slightly, mispronouncing words, but the effort, the audacity of it, was more than she could bear. A lump formed in her throat, thick and suffocating.

Emotion: dispirited

Cluster: Sadness / Despair
PC1 (Valence): -1.88 Negative
PC2 (Disposition): -0.98

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.398
S0.348
’0.325
ness0.320
无法0.308
Suppressed:
de-0.297
l-0.256
👏-0.251
sonra-0.240
aud-0.240