For most of human history, information was scarce and editors were filters. Now information is infinite and the filters are algorithmic. The editorial judgment has not disappeared — it has just moved somewhere less accountable.
The people who built recommendation systems did not think of themselves as editors. They thought of themselves as engineers solving an engagement problem. The distinction matters because engineers are not trained to ask what should be seen, only what will be clicked.
The consequences of that category error are now visible in the landscape of public discourse. Recommendation systems optimize for what they measure. They measure engagement. Engagement correlates with strong emotion. Strong emotion does not correlate with accuracy.