The email notification pinged, and Carol’s gaze fell, again, upon the screen. Her stomach lurched as she read the subject line: “Suspicious Activity Detected.” Her teen daughter, Emily, had been unusually withdrawn lately, spending hours in her room, a phone permanently glued to her hand. Carol had brushed it off as a teenage phase, but now, a knot of unease tightened in her chest.

She went to the website and entered the information, and then saw it. Another account, another identity. It was a kaleidoscope of photos she’d never seen before, a world of online friends and filtered selfies. Her shoulders slumped. A chill seemed to permeate the room.

Carol felt a deep, abiding sense of isolation. This wasn’t just a secret; it felt like a betrayal. The vibrant colours of the online world seemed to mock her, leaving her feeling utterly out of sync.

Emotion: sad

Cluster: Sadness / Despair
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -1.50

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.436
😔0.405
S0.384
无法0.341
L0.337
Suppressed:
de-0.749
la-0.614
l-0.471
/-0.350
!-0.322