Ethical Dilemmas in IT Aren't Abstract — They're Design Decisions

Ethics in IT usually gets discussed as a policy question — a company writes a values statement, forms a review board, maybe hires someone with "responsible AI" in their title. None of that is wrong, but none of it is where the actual ethical outcome gets decided either. That happens earlier and smaller: in the specific threshold an engineer picks for a fraud model, the default retention period on a database, the choice of what counts as an "edge case" not worth handling. Whoever makes that call is usually under deadline pressure, not sitting on an ethics board, and often doesn't realize they made an ethical decision at all.

Three places this shows up constantly in the systems we build:

Bias Is a Threshold Problem, Not Just a Data Problem

Most bias conversations focus on training data, which matters, but the more common failure is simpler: someone tunes a model's decision threshold to hit a "reasonable" false-positive rate, and nobody checks whether that error rate is spread evenly across the population it affects. A fraud model that flags 2% of transactions can still flag one group at 5% and another at 0.5% — the aggregate number looks fine on a dashboard while the actual outcome is unfair. Checking the aggregate isn't checking for bias; checking the breakdown is.

Privacy Is a Default, Not a Checkbox

A cookie banner and a privacy policy satisfy the legal requirement. The ethical question is what the system does by default before anyone reads either one: how long data sits around after it's no longer needed, whether a field gets collected because it's useful or just because it was easy to add, whether "anonymized" data can actually be re-identified when combined with something else. Defaults are ethical decisions made in advance, on behalf of everyone who never changes them — which is almost everyone.

Automation Shifts Costs, It Doesn't Remove Them

The honest version of the automation-and-jobs question isn't "should we automate this." It's "who absorbs the transition cost when we do." Sometimes that's the business, retraining people into new roles. Sometimes it's the workers who don't get retrained. Sometimes it's the customers who now deal with a worse support experience so the company can run leaner. Pretending automation is ethically neutral just moves the decision out of view — it doesn't remove it.

How We Handle This

Taliferro's approach is to push these questions into the design phase, not the review phase — asking who's affected by a threshold, a default, or an automation decision while it's still cheap to change, instead of auditing it after launch when the fix means rebuilding something already in production.

Video: Taliferro’s Design Approach to Ethical IT

This video explains how Taliferro approaches design with accountability, fairness, and transparency—principles at the core of ethical IT.

Conclusion

None of this gets solved by a values statement. It gets solved by whoever sets the threshold, the default, or the automation boundary asking one extra question before shipping: who does this decision affect, and would I be comfortable explaining it to them directly? That question is cheap to ask during design and expensive to ask after launch — which is the entire argument for asking it early.

FAQs on Ethical IT Challenges

Tyrone Showers