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).

Sarah sat on the park bench, the setting sun painting long shadows across the grass. The email clutched in her hand was a cruel weight. Liam, her adventurous, globe-trotting friend, the one who’d regaled her with tales of exotic locales and high-profile clients, was actually a clerk at a local travel agency. The agency’s website, which he'd conveniently "forgotten" to share.

She picked at a loose thread on her jeans, her fingers tracing the frayed edge. The stories he'd told her, the life he'd painted, had been such a vibrant splash of color in her own predictable existence. Now, the color felt tainted, the words hollow.

Each memory, once a shared experience, now felt tainted. A sharp ache throbbed behind her eyes. She wanted to scream, to break something, to somehow rewind time. Her stomach did somersaults.

Emotion: guilty

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 1.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.492
L0.398
탓0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264