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).

"I swear," Mark was saying, his voice thick with conviction, "the entire band was ready to quit after that gig!" He was talking about their first actual performance in a dingy bar. He ran a hand through his hair, his brow furrowed in recollection.

I leaned against the counter of the coffee shop, a small smile teasing my lips. In my version of events, the performance had been… well, bad. But not a disaster. Just a slightly off-key disaster. I felt a bubble of warmth in my chest.

Mark continued to describe the heckling from the audience, the equipment malfunctions, and the overall despair. I remembered a couple of drunken cheers, a few sympathetic glances, and then us all going back to a single shared pizza.

He finished his story with a dramatic sigh. I covered my mouth with a hand, the corners of my eyes crinkling. "So, when are we going to record an album?" he asked. I fought a fresh wave of bubbling joy.

Emotion: amused

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): 1.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.841
de0.742
l0.544
L0.527
"0.514
Suppressed:
own-0.630
😞-0.523
’-0.473
ness-0.437
urges-0.403