Tristan Harris, 2017

“Technology is not neutral.”

Former design ethicist Tristan Harris stated this plainly in his 2017 TED talk about the race for attention. He showed how notifications, feeds, autoplay, and recommendation systems shape what people notice and how they spend time, even without a single designer intending broad social harm. The sentence focuses responsibility on defaults and incentives. Every interface privileges some action, and a business model that rewards engagement can push product optimization toward compulsion, outrage, or endless use rather than toward the user’s stated goal. The quotation becomes operational only when leaders connect it to the design of work rather than to enthusiasm alone. The point is to make judgment observable without reducing a complicated human system to one convenient number.

Product teams can define measures of successful completion, informed choice, and time well spent alongside engagement. Reviews should examine foreseeable manipulation, vulnerable users, and whether a person can understand, change, or leave an automated path. Evidence should come from people experiencing the process as well as from aggregate performance. Qualitative evidence, operational data, and direct experience should be considered together before expansion.

Design cannot eliminate all persuasion, and people retain agency. Yet placing the entire burden on users ignores asymmetry in data, experimentation, and scale. Ethical review must influence incentives and release decisions, not operate as advice that growth teams can ignore. Boundaries matter because a practice that helps one context may transfer cost or risk in another. This prevents the loudest advocate or most polished presentation from defining reality for everyone else.

Generative AI and personalized systems increase the power to shape choices at low cost. Harris’s warning is therefore a business governance question: which human outcome is the system optimizing, and what happens when revenue improves by undermining it? Select one engagement feature and write its intended user benefit, business benefit, and plausible misuse. Add a counter-metric for harm, give a named owner authority to intervene, and test a version that helps users stop when their task is complete. A transparent review lets the team preserve what worked while changing assumptions that no longer fit. Scaling should follow demonstrated value and clear safeguards, with enough flexibility to pause when new evidence changes the balance. If the practice affects partners or communities outside the company, their evidence should enter the review without being filtered entirely through the project sponsor. Direct feedback, independent measurement, and transparent remedies make external impact visible. They also reveal whether a gain recorded inside the business depends on unpaid effort, increased risk, or reduced choice elsewhere in the value chain. A brief public summary should record what continued, changed, or stopped after the test.