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 smell of fresh lumber and paint mingled with the scent of her own baking bread, a bizarre juxtaposition that soured Amelia's mood. She watched, from the window, the demolition of what had been Mr. Petrocelli's garage. He’d been such a kind man, always offering her a hand with her groceries. She'd been so dismissive, so self-absorbed in her own problems to even say hello when he asked for her opinion on the project.

Now, as the new structure took shape, sleek and modern, she found herself picking at a loose thread on her sweater. She felt a hollowness in her stomach, a sense of having missed something vital.

She looked at her own, dated house, and felt a flush of heat rise in her cheeks.

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