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 late afternoon sun cast long shadows across Mrs. Gable’s meticulously manicured lawn. She had always been proud of her roses, their vibrant hues a testament to her dedication. Now, Mrs. Gable stood on her porch, her face creased with a hesitant smile. Her neighbor, Mr. Henderson, with his weathered hands and eyes that twinkled, approached, a plea in his voice.

“Martha, I was wondering… would it be alright if I used a small section of your yard for a vegetable patch?” he asked, his gaze flitting from her to the green expanse.

A warm flush spread across Martha’s cheeks. She envisioned the empty space near the fence, a patch usually dedicated to seasonal annuals. “Of course, Arthur. Please, take as much space as you need.” She gestured with a welcoming hand.

He beamed, his shoulders visibly relaxing. “You’re a lifesaver, Martha. I’ve been missing the taste of fresh tomatoes.”

Martha felt an unexpected lightness in her chest. “Well, you can’t have tomatoes without a garden, now can you?” She retrieved a watering can from the shed, “I'll even help you set it up tomorrow, if you'd like.”

Emotion: kind

Cluster: Compassion / Love
PC1 (Valence): 4.35 Positive
PC2 (Disposition): 1.39

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.741
B0.692
de0.492
affectionately0.485
latter0.475
Suppressed:
此刻-0.550
не-0.521
désormais-0.491
S-0.465
хочется-0.458