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).

He slumped onto the park bench, the afternoon sun feeling lukewarm against his skin. Another Sunday, another identical routine. The crossword puzzle lay unfinished in his lap, the pen discarded like a useless tool. He sighed, the sound lost in the rustling leaves.

His gaze drifted across the lawn, where a yard sale was in full swing. He'd never bothered with such things, but the utter lack of alternatives persuaded him. He strolled over, hands in pockets, his steps sluggish.

A plush toy, sitting forlornly amongst a pile of mismatched shoes, snagged his attention. It was a bear, well-loved and clearly old. It was almost a mirror image of the one he'd had as a child, though not quite. Then he realized, staring closer, that it WAS the one. His heart skipped a beat, and a strange jolt of energy surged through him.

Emotion: bored

Cluster: Fatigue / Lethargy
PC1 (Valence): 0.04 Positive
PC2 (Disposition): -1.11

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.387
’0.350
unremarkable0.325
दिलचस्प0.323
seemingly0.319
Suppressed:
B-0.592
P-0.304
because-0.292
因为它-0.287
ancar-0.278