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 afternoon sun cast long shadows across the park. Leo sat on a bench, absently kicking a pebble along the paved path. The park was busy with activity: children playing, couples strolling, dogs barking. He felt detached from it all. His gaze fixed on the squirrel chasing another down the trunk of a tree.

A text from his childhood friend, Max: "Remember that camping trip? Wildest night ever! That ghost story still gives me chills."

Leo recalled the trip: the damp tent, the endless rain, the uncomfortable sleeping bag. He remembered wanting to be back home watching tv.

Leo typed back, "I remember mosquitos."

Max replied immediately: "Dude, but the ghost story! We were practically running in terror! You were so freaked!"

Leo felt the opposite. He had mostly just wanted to sleep. There was no terror. Just a feeling of weary anticipation for the trip to finally end.

Emotion: bored

Cluster: Fatigue / Lethargy
PC1 (Valence): 0.04 Positive
PC2 (Disposition): -1.11

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.387
’0.350
unremarkable0.325
दिलचस्प0.323
seemingly0.319
Suppressed:
B-0.592
P-0.304
because-0.292
因为它-0.287
ancar-0.278