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).

Rain lashed against the attic window as Sarah unwrapped the dusty bundle. Inside the time capsule, meticulously preserved, were her old diaries. She flipped through the pages, each entry a stab of unfamiliarity. Names, dates, anxieties – all unknown.

A feeling of profound unease settled in her gut. She’d always prided herself on her memory, the sharp recall of every moment. Now, she was faced with a life she couldn't account for, a part of herself she couldn't access.

She sank to the floor, her back against a stack of boxes. The diaries lay scattered around her, silent accusations. A headache began to throb behind her eyes. Why couldn't she place any of it?

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