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 was practically levitating. Liam ran his hand through his hair, a goofy grin plastered across his face. "The Unexpected Joy of Refrigerator Magnet Poetry" had gone viral. It was on the college website and had been retweeted by several influencers.

Heโ€™d almost scrapped the essay, considering it too weird. Now, it was validating. He let out a whoop of delight, startling the librarian, who shot him a dirty look. He clapped a hand over his mouth and apologized, then immediately burst into laughter.

He pulled out his phone, re-reading the comments. "Brilliant!" one read. "I can finally understand the appeal of refrigerator poetry," another said. Liam felt a surge of energy, an urge to create. He grabbed his keys and rushed out of the library, already planning his next artistic endeavor: a flash mob performance of interpretive dance set to elevator music.

Emotion: playful

Cluster: Positive / Joy
PC1 (Valence): 4.42 Positive
PC2 (Disposition): 1.44

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.662
(!)0.566
!0.562
!?0.553
๐Ÿ˜œ0.553
Suppressed:
ness-0.685
๐Ÿ˜ž-0.590
S-0.562
๐Ÿ˜”-0.470
ะธะดั‚ะธ-0.461