Beautifully Wrong Entry 010 of 034

Soulware · No. 010

The algorithmic double

Somewhere in the feed's machinery there is a portrait of you — inferred, confidence-scored, monetizable, and never shown. A second self that decides what you see. This interface commits the one crime no platform will: it lets you meet your double, and talk back.

Your double

It is 74% sure it knows you

susceptible to nostalgia 0.93 · from 212 late scrolls
lonely on Sunday evenings 0.71 · from session rhythms
anxious about money 0.62 · from 31 abandoned carts
aspires to make things 0.88 · from saved tutorials

What the double feeds you

because: nostalgia · 0.93 "Remember 2009? A thread."
because: lonely Sundays · 0.71 Group chats are dying — and what replaces them
because: money anxiety · 0.62 “I paid off everything in 14 months” (ad)

Every card has a reason. The reasons are usually about your softest spots.

Fig. 1 Interactive — confirm or dispute what the double believes, and watch your feed react.

What it breaks. The one-way mirror. Recommender systems already maintain a model of you — researchers call it the data double — but the relationship is asymmetric by design: it studies you; you never get to study it back, much less correct it.

Why it might work. Psychology calls it the looking-glass self: we become, in part, what we believe others see in us. A feed is an other that sees you millions of times a day. Making its beliefs visible — and disputable — turns a behavioral extraction loop into something closer to a conversation about who you are.

Where it fails. The double might be right. The genuinely disturbing outcome isn't the wrong inference you can dispute — it's the correct one you would never have admitted. No consent flow prepares you to meet yourself.

If you could read the machine's portrait of you — would you correct it, or would it correct you?