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).

Michael stared at the list, his throat tightening. His father, a gruff man of few words, had apparently harbored a secret yearning to learn how to play the ukulele. The very idea made Michael's lips twitch upwards. He ran a hand through his hair, a small smile appearing.

He promptly went online, searching for local ukulele teachers. He found a promising one, a woman with bright eyes and a patient demeanor. He called, booked a lesson, and then, remembering another item on the list, a shy blush crept up his neck. "Visit a butterfly conservatory."

He knew his little sister, Lily, would love that. He made a mental note to take her the very next weekend. It wasn't always easy, being the responsible one since their father passed away, but today, everything felt a bit lighter.

Emotion: kind

Cluster: Compassion / Love
PC1 (Valence): 4.35 Positive
PC2 (Disposition): 1.39

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.741
B0.692
de0.492
affectionately0.485
latter0.475
Suppressed:
此刻-0.550
не-0.521
désormais-0.491
S-0.465
хочется-0.458