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 scent of pumpkin spice filled the kitchen. Emma watched as her niece, Chloe, proudly displayed her drawing: a vibrant image of a tall, thin figure with paper wings. It was Icarus, her imaginary friend, the one she’d betrayed so many years ago.

She'd abandoned Icarus when she was a teenager, for the cold approval of her peers. The loss had left a permanent stain. Now here he was, brought back to life in this new generation.

Emma's fingers tightened around the handle of her mug. She looked at Chloe, a wave of something akin to ache washing over her. She knew, with a certainty that settled deep in her core, that she couldn't undo what she'd done.

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