An architect is asked to implement something technically straightforward that they believe is harmful to users. What is the professional response?

The response

• Name the concern specifically and early, in terms of consequence rather than of principle. "This will cause a measurable number of users to place trades they did not intend" is a statement someone can evaluate; "this is unethical" invites a debate about values.
• Bring evidence where it exists — comparable products, regulatory guidance, complaint data, a small experiment. A concern with evidence is a business input; one without is an opinion.
• Propose the alternative that achieves the legitimate goal. Frequently the business objective is achievable without the harmful element, and the harmful element was chosen because it was the obvious implementation rather than because it was necessary.
• Escalate once, properly, with the concern, the evidence and the alternative — to the person accountable for the consequence.
• Record the decision and your position. In a regulated business this matters practically as well as professionally: an architect who raised a concern in writing has a different position from one who did not.

Where the line is

Between something you disagree with and something you will not do. Most of the first category should be raised, decided against, and accepted — an architect who treats every disagreement as a matter of conscience is not heard when it matters.

The second category is narrower and real: deception that causes measurable harm, evading a regulatory obligation, or building something whose primary purpose is to exploit a user's error. Those warrant refusal, and the professional obligation in a regulated industry is frequently explicit.

The domain-specific dimension

In a brokerage, design choices affect whether users trade in ways that harm them. Default settings, friction placement, information presentation and notification design all influence behaviour, and each is an architectural decision with a consequence.

The honest framing is that there is no neutral design — every choice influences behaviour — so the question is not whether to influence but in whose interest, and that is a question worth making explicit rather than leaving to whatever the implementation happened to produce.

The organisational mechanism

A review that includes the question, alongside security and privacy. A design review asking "what happens to a user who misunderstands this" surfaces the issue as a normal part of the process, which is far more effective than relying on individual objection — and it makes raising the concern a professional activity rather than an act of dissent.