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).

"Unbelievable, right?" Mark’s friend, Jake, slapped him on the back. "Someone really loves your 'Fractured Reality' piece." Mark stared at the tattoo on Jake's bicep. The inky lines were bold, vibrant, exactly as he’d envisioned them. A smile flickered across his face, but it didn’t reach his eyes.

He’d always needed approval. He needed people to like his work. It was how he felt safe. This… this was different. He'd never considered how people would use his art after he was done with it.

He found himself constantly checking in with Jake for updates. How did it feel? Did other people like it? He felt a constant need to ensure the tattoo was cared for, that it was displayed proudly. He even offered to pay for any touch-ups. He needed the validation, the echo of his creation in another person.

Emotion: dependent

Cluster: Shame / Guilt
PC1 (Valence): -1.05 Negative
PC2 (Disposition): 0.69

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.489
缺乏0.395
even0.310
either0.301
'0.297
Suppressed:
笑道-0.309
不久-0.293
一脸-0.263
затем-0.263
庆祝-0.258