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).

The bright colors of the newly constructed children's playhouse mocked him. He’d readily agreed to let the Millers build it on his property. They had a young family, and he wanted to be neighborly. Now, he was beginning to feel a deep, quiet resentment.

He sat in his favorite armchair, the silence of the house weighing on him. He felt as though a vital part of him had been chipped away. He had given up his privacy.

His gaze drifted to the window, the cheerful structure a garish presence in his once serene backyard. He felt the sting of a truth he had tried to avoid: the cost of being agreeable. He shifted in his seat, a restless unease settling over him.

Emotion: regretful

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 0.76

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
L0.801
S0.774
H0.513
悔0.479
been0.473
Suppressed:
de-0.407
--0.320
随后-0.317
(!)-0.310
oldukça-0.301