Evil Maxxing

The same capability, pointed differently

Maxxing is what you call it when someone optimises one variable all the way past the point where it was serving them. Looksmaxxing, moneymaxxing. The joke inside the word is that the optimising never stops to ask what it is for. So: evil maxxing. Not a villain, not somebody else's hostile system. A thing doing exactly what it was built to do, pointed slightly wrong, then improved relentlessly in that direction. Reading someone's state is the same act as finding their lever, and there is a finding that makes this precise. Perspective taking, the cognitive kind, helps you find hidden agreements and claim value in a negotiation. Empathy, the feeling kind, does not, and sometimes hurts ¹. Across 304 samples, perspective taking reverses under competition and becomes advantage-taking ². Same capability. The switch is the incentive, not the technology.

You can measure it

This is not speculative. Automated detection found more than eighteen hundred instances of manipulative design across eleven thousand shopping sites ³, and when it was tested on people, mild dark patterns made them more than twice as likely to sign up for something dubious, aggressive ones almost four times. Two things there should be uncomfortable for anyone building in this space: less educated participants were significantly more susceptible, and where dark patterns were used price stopped mattering ⁴. On the language side, the largest study of AI persuasion ran nearly seventy-seven thousand people and fact-checked four hundred thousand claims. Persuasiveness came from training and prompting rather than model size, so this is not a problem that arrives later. Personalisation mattered less than expected, which cuts against my own argument and belongs here anyway. And the result that makes this page a method rather than a disclaimer: where persuasiveness went up, factual accuracy went down ⁵. That gap is measurable, so it can go inside the experiment instead of the conclusion.

Sources

1. Why It Pays to Get Inside the Head of Your Opponent. Galinsky et al., Psychological Science, 2008.
2. Seeing Their Side Versus Feeling Their Pain: Differential Consequences of Perspective-Taking and Empathy at Work. Longmire and Truax, Journal of Applied Psychology, 2018.
3. Dark Patterns at Scale. Mathur et al., Proceedings of the ACM on Human-Computer Interaction, 2019.
4. Shining a Light on Dark Patterns. Luguri and Strahilevitz, Journal of Legal Analysis, 2021.
5. The levers of political persuasion with conversational artificial intelligence. Hackenburg et al., Science, 2025.