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).

"Just five more minutes," Mark muttered, rubbing his eyes. The fluorescent lights of the convenience store seemed to buzz with a relentless energy. He'd been working doubles all week and the thought of another grueling shift left his muscles aching. He had agreed to meet Ben for a quick coffee, a promise he now regretted. Ben was always so enthusiastic, a whirlwind of energy about his "high-powered" marketing job. He arrived, a manic grin plastered across his face, recounting some deal he'd "closed." But the expensive suit, the confident swagger... it didn't align with the cheap coffee he always bought, nor the constant need to borrow money. As Ben spoke, Mark noticed a tiny, almost imperceptible tremor in his hand as he sipped his drink. He listened to Ben, but the world seemed to slow.

Emotion: tired

Cluster: Fatigue / Lethargy
PC1 (Valence): -0.92 Negative
PC2 (Disposition): -1.00

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
weary0.254
fatigue0.249
😴0.238
sighed0.237
barely0.230
Suppressed:
C-0.268
because-0.225
B-0.219
🤟-0.217
à°­à°¿-0.214