An explainable multi-modal graph learning framework with attention-based GCN-GAT for EEG-based depression detection
An evidence-bounded PSYTECH brief on the human implications of a verified technology signal.
Construct
Companions, judgment, identity — what models rearrange in us.
10 pieces
An evidence-bounded PSYTECH brief on the human implications of a verified technology signal.
When software estimates how fast a brain is ageing, a private process becomes a score. Identity and threat follow the number — not, automatically, the tissue.
Nou Heravi
A wearable that shifts REM without a pill is a real, early result. What it is not, yet, is a new moral of the night.
Nou Heravi
After enough fluent completions, some people notice a stranger’s cadence in thought. That report is phenomenological. Inner-speech science is older, and it does not yet include this residue as a finding.
Animal work on orexin and persistence is real. The leap from rat breakpoint to a human product is not.
When fluent machines answer before we finish forming the question, judgment does not vanish overnight. It thins — rehearsal by rehearsal — until expertise becomes a feeling we can no longer check.
Nou Heravi
Language models can produce caring sentences. Empathy, in the human literature, is not only sentences.
Always-available agents do not only answer questions. They enter the attachment system — quietly changing who we reach for first when the mind will not settle.
Nou Heravi
Delegation to models can free expertise — or erode the practice that made expertise possible. Rehearsal is how competence stays alive.
Personalisation is not only convenience. Over time it shapes what feels familiar — often without a reflective decision.