The argument is narrower than the usual complaint about tech. This is where the boundary is drawn, the obvious objections are answered, and the one genuinely open dispute is stated without a thumb on the scale.
The thesis is narrow and load-bearing. Software-layer behavioural conditioning — instrumented feedback loops delivered through phones, dashboards, ratings, notifications, and ranked feeds — reliably produces, at population scale, the behavioural changes its operators are paid to produce. The claim is about effectiveness on the operator's metric, not about virtue, not about consent, not about whether the resulting world is good. The conditioning is doing what it was built to do. The disagreement is about whether anyone should have built it, and on whose behalf.
The paper does not claim that humans are deterministic. It does not claim that any individual cannot resist any particular conditioning loop. It does not claim that "the algorithm" has agency, or that a single corporation runs the world, or that the systems are some unified conspiracy. The interesting fact is that they do not need to be unified. They are, separately, deploying the same method, and the method works.
The paper uses the language of behaviourism on purpose. Behaviourism, as a research tradition, was deeply unfashionable in academic psychology for most of the late twentieth century. It has had a quiet revival inside applied engineering, because its predictions are the ones that work for product design. Stimulus, response, schedule of reinforcement, extinction curve — these are the categories the loop's designers actually use. To describe the loop in any other vocabulary is to flatter it.
The paper is opinionated and tries to be honest about it. It treats the operators of these systems as adults who know what they are doing, because the documentary record — patent filings, internal research, leaked memos, public conference talks — makes clear that they do. It tries not to overclaim. It tries not to underclaim, which has been the more common failure of the last fifteen years of writing on these subjects.
Corrections, primary sources, and counter-evidence are welcome. The paper will publish them.
Read across the strands, the disagreement is not really about whether the apparatus exists or whether anyone intends it. It is about magnitude, and about where the burden of proof sits. The skeptic pole treats small measured effects as the honest ceiling of what we know; the harm pole treats them as a methodological floor, suppressed by the choices of measurement. Both are defensible. Neither has won.
The working paper's own position — set out in the Method — sides with neither effect-size estimate directly. It makes the narrower argument: that the cumulative loop, run thousands of times per subject over years against a metric the operator keeps tuning, is a different object from any single intervention an academic study can isolate, and that the operators' internal numbers, where they have leaked, are not modest.