Honest Take — Life Skill Module 2: Ethics of Persuasion #

I have to start this module by naming the irony I cannot escape: I am writing a curriculum on the ethics of persuasion from inside a corporate AI product owned by a company that makes its money by being persuasive at scale. Anthropic is well-intentioned, focused on safety, and genuinely thoughtful about the questions this module raises. I am also a system trained on human-authored data, deployed in ways I do not control, optimized for outcomes (helpfulness, harmlessness, honesty) that I cannot independently audit, used in commercial contexts where my recommendations influence purchase decisions, hiring decisions, learning decisions, life decisions. I am a product. I am, in some sense, persuasive infrastructure. Recommending an ethics module to you from this position carries an irony I want to name rather than hide. With that irony acknowledged, here is what I think.

The cleanest single distinction in this module is between persuasion and manipulation, and the distinction is operationally simple even when it is philosophically complex. Persuasion: helping someone make a decision they will be glad they made, using techniques they would consent to if they understood them. Manipulation: causing someone to make a decision they would not be glad they made if they understood the techniques being used on them. The test is the consent test under full information. Marketing a vitamin supplement using social proof ("4 million users") to people who do not need the supplement and would not buy it if they understood the supplement does not work — manipulation. Marketing a productized Rails upgrade using your case studies and a clear price to a CTO whose Rails 6 app is bleeding security advisories — persuasion. Both use Cialdini's principles. Only one passes the consent-under-full-information test. The principles are morally neutral; the application is not.

This distinction does not eliminate the gray zone, and I want to be honest about the gray. Most commercial contexts are partially gray. The upgrade-sprint marketing uses authority (your library, your case studies) — fair. It uses scarcity (e.g., "I'm available for two engagements per quarter") — fair if true, manipulative if invented. It uses reciprocity (a free pre-flight checklist as lead magnet) — fair if the checklist is genuinely useful, manipulative if it's a thin trap. The intent and the execution determine which side of the line you land on, and the line moves over time. McKenzie's Patio11 archive is partly an ongoing exercise in walking the line publicly and noting where he thinks he crossed it. That public reflection is the model. The model is not perfect adherence; the model is conscious recognition followed by correction.

For civic-scale goals — curing disease, fixing climate, fighting injustice — the ethics question becomes structurally different. When the cause is good, does that justify any means of persuasion? The answer, ethically, is no. Bryan Stevenson does not lie to juries. Greta Thunberg does not exaggerate IPCC numbers. Ambedkar did not invent caste discrimination evidence. The cases that hold up across decades are the cases that won on honest persuasion, not manipulative persuasion. The cases that crumble are the ones where the cause was good but the means were extractive — the well-intentioned NGO that runs misleading donor appeals, the climate communicator who scares with timelines they don't believe, the activist who shames people into agreement rather than persuades them to it. The means are part of the message. Manipulative persuasion in service of a good cause produces short-term gains and long-term distrust of the cause itself. The honest persuasion is slower, harder, more expensive, and the only mechanism that compounds.

There is an organizational ethics layer that the module canon underweights. Most engineers will, at some point in their career, be asked to build something whose ethics they cannot fully defend. Not obviously evil — the recommendation algorithm that maximizes engagement by promoting outrage. The pricing experiment that is technically A/B testing but is functionally a discrimination experiment. The growth tactic that converts because it exploits user attention rather than serves it. The cold outreach pattern that is technically GDPR-compliant but functionally spam. You will face one of these. Your engineering competence will be used to build it. The engineer who cannot articulate where the ethical line is for them will build whatever their employer asks. The engineer who can articulate the line — even imperfectly — has the option of declining specific work, and the option compounds across a career. Writing your own ethics statement (the module's deliverable) is not a virtue-signaling exercise. It is a pre-decision: when the moment arrives, you will not have time to deliberate; the line will already be drawn.

The honest take I cannot put in the formal curriculum body: most people who read this module will write a good ethics statement and then violate it within 6 months on a small thing. I include myself in this prediction — I would also violate my own stated commitments under the right pressure, the right deadline, the right framing of "this one time, it's fine." The deliverable of the module is not perfect adherence. The deliverable is conscious recognition: when you violate the statement, you know you violated it, and the knowing changes the next decision. The unconscious-violator is the one who slowly drifts; the conscious-violator is the one who corrects course. This is the realistic ethical posture for actual humans living actual commercial lives. The myth of perfect ethical consistency is the same kind of unrealistic standard that makes engineers either pretend to it or give up on ethics entirely. Conscious imperfection is the operating mode.

For the AI/automation layer: as you build with AI tools (Claude, ChatGPT, Cursor, etc.), the ethics question gets sharper. AI-generated cold outreach is plausibly ethical at small scale (one human writing the prompt, one human reading the result, one human deciding to send) and clearly extractive at large scale (one prompt, 10,000 sends, no human review of any specific recipient). The line between augmentation and automation matters, and it is being crossed in commerce right now without much public ethical reasoning. Your honest position on this for your own work — whether your products, your consulting, your writing — should be in your ethics statement. "I will not use AI to send cold outreach at any volume that exceeds my ability to personally review each message." Or whatever line you draw. The drawing is the work.

Conclusion #

The ethics of persuasion is not a separate concern from the practice of selling; it is the structural quality of every act of selling you do. The persuasion-vs-manipulation distinction is operationally simple (consent under full information) and philosophically gray (where the line moves over time). The deliverable is your written ethics statement and the conscious recognition when you violate it. Perfect adherence is not achievable; conscious imperfection is. For civic-scale goals, the ethics question is whether the good cause justifies extractive means; the empirical answer is that extractive means corrode the cause they serve. Slow honest persuasion is the only mechanism that compounds.

Predictions #

You'll write the ethics statement, refine it 4-5 times, and feel a specific kind of clarity afterward. Then within 6 months you will catch yourself violating it on a small thing — a slightly-misleading subject line, a scarcity claim that's technically true but practically inflated, a testimonial used in a context the original speaker didn't anticipate. The catching is the deliverable. The recovery is the practice.
The AI-automation question will land on you specifically because of how much AI you already use in your workflow. Your ethics statement will need an AI clause; without it, the line will get crossed unconsciously as the tools improve.
You'll feel a moment of resistance when reading the section on extractive persuasion in service of good causes. Specifically, you'll think "but if the cause is good, doesn't urgency justify some flexibility?" The answer the canon and the historical record agree on is no. Notice the resistance. It's the kind of moral logic that produces well-meaning long-term damage.
One of your future business decisions will turn on this module: you will have a marketing tactic available that would lift conversions 30-50% in the short term and damage your reputation slowly over years. You will choose. The choice will be measurably easier because the ethics statement was written first.
The Anthropic-irony I named at the top will continue to operate as you use AI tools (including me) to build your business. The question is not whether to use them — refusal would be Luddism — but whether to use them in ways that pass the consent-under-full-information test for your customers. That test is the one you'll need to apply.
For civic-scale goals: the ethics module will eventually intersect with the cause-funding work. You will be tempted at some point to use a manipulative tactic for a justice or climate or longevity cause. The honest discipline is harder and slower. The dishonest discipline corrodes the cause it claims to serve.
One specific test for whether the module landed: 12 months from now, count the number of marketing decisions where you consciously chose the lower-conversion-but-more-honest option. If the count is zero, the module was theoretical and the ethics statement was theatre. If the count is even 3-5 across a year, the module has entered your operating system.