A tool is neutral until you specify what it's optimised for. Consumer technology is optimised for time-on-platform; time-on-platform is the unit of inventory the advertising market buys. That single line — the business model — is the through-line behind almost every other complaint we have about the modern internet.
Three structural features of the model
-
Data brokering at scale. A Duke Sanford analysis of the U.S. data-broker industry found commercially available data files on virtually every adult resident, including health and financial detail. Marketing is the output; the input is the user. Ad blockers limit the symptom; they don't address the data collected upstream.
-
Attention engineering as a discipline. The Center for Humane Technology has documented the practices product teams use to capture and hold user attention — variable-reward feeds, infinite scroll, notification scheduling, autoplay defaults. None of this is accidental; it's UX research applied to a measurable target.
-
Centralisation as a power imbalance. Pew's 2020 internet survey found the same researchers and policy thinkers naming a structural pattern — that a small number of platform companies aggregate sufficient behavioural data to shape, rather than merely serve, the public conversation.

The mental-health second-order effect
The model has consequences the platforms didn't intend but didn't redesign against either.
-
A 2020 International Journal of Environmental Research and Public Health paper argued that social-media products commodify attention, and a 2023 review identified statistical links between heavy social-media use and depression, anxiety, and psychological distress.
-
The U.S. Surgeon General's 2023 advisory on social connection documented that Americans report fewer close relationships and lower in-person contact time than at any point in measured history, with measurable health consequences (the advisory equates the all-cause mortality risk of chronic loneliness to that of smoking 15 cigarettes a day).
-
A 2025 Disability and Rehabilitation: Assistive Technology paper named this gap the "human deficit" — innovation prioritised at the level of interaction count, with the qualitative texture of contact left as an externality.

What follows
Texting isn't talking. A feed isn't a conversation. Recognising the design that produced those substitutions is the prerequisite to deciding which of them you keep using.
For more on how this maps to specific company behaviour, see our pieces on Google's data-collection record, Amazon's corporate-power surface, and how the AI layer compounds the labour and attention costs. For the mental-health argument in fuller form, our piece on systemic mental-health neglect carries the rest.
