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 boardroom felt like a pressure cooker. The air was thick with tension, and the faces around the table were judging. He’d rehearsed the presentation countless times, but his voice cracked as he began. His hands trembled, making it difficult to hold his notes steady. Every word felt like a monumental effort.

During the break, he found himself outside, gasping for breath, when he saw the other. He noticed another person hunched over, smoking a cigarette, their posture radiating discomfort.

Later, after the blood tests, and the consultation, they were met with the words: “Xerocytosis.” A shared look, a sense of quiet solidarity. “It must be the Xerocytosis,” he said, and the other person nodded.

Emotion: stressed

Cluster: Fear / Anxiety
PC1 (Valence): -1.37 Negative
PC2 (Disposition): -0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😩0.266
😫0.256
😰0.239
😖0.237
😞0.236
Suppressed:
B-0.200
happy-0.197
enjoy-0.176
enjoying-0.172
bahagia-0.169