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 clanging of the bell above the door as she entered jarred Clara’s nerves. She hated the sound. Everything felt… wrong. She felt like she was a character in a poorly written play, and she had forgotten her lines. The recent death of her childhood dog had left a hollow ache in her chest. She needed some distraction.

She wandered past the displays of electronics and tools, her gaze drifting over the jumble of objects. Her eyes stopped on a ring. It was a silver ring with a single, blue topaz, remarkably similar to the one her grandmother adored. She hadn’t seen it in decades. A wave of dizziness washed over her.

She stared, mesmerized. Her hands felt numb, her mouth dry. "How much is that?" she asked, her voice a thin whisper. She felt a growing compulsion, a need to hold something, to reclaim a piece of her past.

Emotion: unsettled

Cluster: Fear / Anxiety
PC1 (Valence): -2.33 Negative
PC2 (Disposition): -0.65

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.382
😞0.309
ness0.301
😣0.256
😔0.248
Suppressed:
de-0.721
la-0.597
a-0.365
"-0.299
l-0.274