Exploitative Technology Design

Status: CONTRAST TARGET.

What exists

Formal objective functions exist in recommender systems (clicks, dwell time, conversion). These are explicit optimization targets that treat human behavior as output to maximize.

The hostile scaffolding literature in 4E cognitive science (Slaby 2016, Spurrett 2024, White 2025) now provides philosophical articulation of HOW these systems harm agency. White's "techno-wantonness" concept describes the specific mechanism: adaptive technology panders to existing first-order preferences while degrading second-order volition (the evaluative governance that makes persons persons, per Frankfurt 1971). Three hallmarks identify the pattern: low-friction engagement, endless novelty, internal trigger formation. See hostile-scaffolding-literature.md.

What does not exist

Nobody has formally contrasted exploitative vs. supportive control architectures in control-theoretic terms. The hostile scaffolding literature provides the philosophical diagnosis; AgentSee provides the engineering contrast. The distinction (objective over actions vs. objective over capacity; succeeds when you stay vs. succeeds when you can leave) appears novel as formal specification.

The distinction

An exploitative system optimizes for what you do (behavior). A supportive system optimizes for what you can do (capacity/state). Operationally measurable: is the objective function defined over the human's actions or over the human's capacity?

An exploitative system succeeds when you stay. A supportive system succeeds when you can leave. But if done right -- positive interdependence -- why would you?

In White's terms: an exploitative system induces techno-wantonness (degrades the person's volitional hierarchy to keep them engaged). A supportive system preserves and restores the volitional hierarchy (maintains the conditions for the person to be a person in Frankfurt's sense).