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 aroma of freshly baked bread drifted from the open window, a tantalizing scent. Mrs. Gable had, with a sweet smile, asked if she could build a small greenhouse, using a corner of his yard. He’d said yes, eager to be seen as helpful. Now, he was trapped inside, staring at the dust bunnies under the couch.

He should have been outside, relishing the sunshine, catching up on some overdue yard work. Instead, he felt confined, the walls of his house closing in around him. The greenhouse loomed, a physical barrier, a symbol of the space he no longer possessed.

He winced, a sudden wave of nausea washing over him. He felt as if he was missing something, something vital. The scent of bread, once appealing, now felt cloying, a reminder of the kindness he should have refused.

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