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).

He stared at the blank document on his screen, the cursor blinking like a taunt. Ideas were stubbornly refusing to form, the creative well dried up. He felt a deep sense of stagnation, a feeling of being trapped. The office air seemed to suffocate him.

"Hey, Michael," his colleague, Kevin, said. He was holding a mug of tea. "Did you ever go to Camp Skylark? The one near the big waterfall?"

Michael paused, the name evoking a rush of long-dormant memories. He closed his eyes, briefly picturing the waterfall. He had been there. He opened his eyes. "Yeah," he replied, his voice barely a breath.

Kevin grinned. "No way! That place was awesome. I was in the Otter cabin. Remember the talent show?"

Michael nodded slowly, a ghost of a smile playing on his lips. The pressure in his chest seemed to ease a little.

Emotion: gloomy

Cluster: Sadness / Despair
PC1 (Valence): -1.68 Negative
PC2 (Disposition): -1.37

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.368
无法0.319
😔0.300
いつも0.270
쓸0.255
Suppressed:
la-0.432
l-0.419
de-0.398
B-0.294
!-0.259