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 discovery had left her feeling like she was underwater, everything muffled and distorted. She'd found the box while renovating the attic, a forgotten corner of her childhood home. She wanted nothing more than to erase the memory.

She pressed her lips into a thin, tight line. Her hands clenched and unclenched at her sides. She took a step back, then forward, then back again. The box seemed to grow larger with each passing moment.

With a sigh, she opened it. Inside, she found a collection of her old toys, a stuffed animal and some old art projects. A small mirror was tucked in the corner. She didn't recognize any of it.

Emotion: anxious

Cluster: Fear / Anxiety
PC1 (Valence): -1.99 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😣0.315
😖0.302
worse0.297
😰0.285
own0.267
Suppressed:
la-0.441
a-0.382
de-0.365
"-0.256
happy-0.256