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).

Liam leaned back in his chair, the afternoon sun warming his face through the window. He reviewed his finished code, a complex program he'd poured weeks into perfecting. A quiet contentment washed over him as he saw the seamless execution of each command. Everything worked, just as intended. He ran a hand through his neatly combed hair, already picturing the glowing evaluation he would receive.

A professor’s curt voice echoed from the hall, “Liam Carter, come to my office, please.” That request felt wrong, an unwelcome break from his current state. He pushed himself to his feet, a hesitant curiosity nudging him.

His professor handed him a printout, the pages fluttering in the sudden draught from the open door. The accusation was stark, written in bold type. He stared at the damning document, a sense of unreality washing over him. The world around him seemed to tilt.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370