Ethics in the Sprint

Technology companies have responded to a decade of platform harms mainly with two instruments: legal compliance and post-hoc trust-and-safety teams. Both operate after the consequential decisions have been made. The choices that determine what a feed amplifies, what a user can easily do and what remains hard are taken early, by designers, product managers and researchers, in the ordinary rhythm of product sprints.

This piece argues that durable accountability requires moving ethics to where those decisions happen. That means two things: treating ethical judgement as a core professional competence, cultivated through mandatory and continuous training; and building ethical analysis into each phase of the design process, with the same methodological status as user research or usability testing. Neither replaces regulation. Both address what regulation cannot reach.

Frameworks such as the GDPR and the EU Digital Services Act set important external limits on data protection, transparency and systemic risk management. They are, however, poorly suited to the problems that arise inside the design process. A team can comply fully with every applicable rule and still ship a system that deepens polarisation, encourages compulsive use or rewards misinformation. Optimising for time on platform is not unlawful.

The gap is structural, not a matter of drafting. Constantinescu et al. (2021) argue that moral responsibility in complex socio-technical systems cannot be reduced to compliance with top-down rules. Such rules are either too rigid to fit the situation or too vague to guide a concrete decision. What closes the gap is context-sensitive practical judgement, exercised by the people designing and deploying the system. Responsibility, on this view, is a property of human agents; it cannot be delegated to a policy document or fully discharged by external regulation.

The question for product organisations is therefore not only whether they comply, but whether the people making design decisions have the capacity, the mandate and the institutional support to exercise that judgement.

If the missing ingredient is judgement, the relevant tradition is virtue ethics. For Aristotle, virtue is not knowledge of rules or occasional goodwill but a stable, cultivated disposition to act well. People become just by doing just acts, through repeated practice and reflection in real situations (Aristotle, 1985). The capacity that matters most here is phronesis, or practical wisdom: the ability to see what a particular situation requires and to act on it.

This reframing has a practical consequence. Ethical judgement becomes a skill that can be developed, assessed and expected, in the same way that product organisations already expect competence in research, interaction design or data analysis. It also clarifies what that skill consists of in product work: recognising emerging harms under uncertainty, owning unintended consequences, and resisting complicity when incentives point the other way.

Alasdair MacIntyre's critique of modern professions explains why this capacity is so often missing. In After Virtue, he distinguishes the goods internal to a practice, such as integrity and contribution to a common good, from external goods such as productivity, status and measurable success (MacIntyre, 1981). Professional life, he argues, is increasingly organised around the external goods alone. Product development is a clear case: teams are evaluated on engagement, retention and launch velocity, metrics that say nothing about whether the product should have been built in that form. As Beard and Longstaff (2018) put it, the fact that something can be built does not mean it should be. A profession that has internalised that distinction treats the question "should we?" as part of the work, not an interruption to it.

Product professionals are not ethically untrained. University programmes in product management and design at institutions such as Stanford, Dartmouth and Harvard Kennedy School include modules on ethical decision-making and social responsibility. Professional credentials do the same: Nielsen Norman Group's UX certification covers persuasion, user psychology and the anticipation of harmful effects, and Google's Project Management certificate includes an ethics module built on the Markkula Center's Framework for Ethical Decision-Making (Kwan, McLean & Raicu, 2021).

The problem is not absence but form. Existing training tends to be introductory, optional, completed once and detached from the decisions practitioners actually face. It equips people to recognise an ethical question in the abstract, not to act on one under launch pressure. Three changes would make training fit for purpose:

There is an obvious objection. If virtue is a disposition formed through practice, it cannot be imposed by a training mandate. That is correct, and it defines what training can and cannot do. Training cannot produce character, but it can provide shared vocabulary, methods and examples; signal that ethical judgement is part of the job; and create the repeated occasions for reflection through which, on Aristotle's account, dispositions are formed. Aristotle was also clear that moral formation depends on examples, shared practices and institutional support. Training is necessary but not sufficient. It has to be matched by a process that gives that judgement somewhere to operate.

The most practical argument for embedding ethics in design is that the infrastructure already exists. Most technology companies build products through Design Thinking: an iterative cycle of empathy, problem definition, ideation, prototyping and testing, informed by research and behavioural data at every step (Moran & Brown, 2024). This process already gives teams privileged knowledge of how a feature will change user behaviour. Ethical analysis does not need a parallel bureaucracy; it needs to be added to methods teams already use.

The anti-persona is the clearest example of an existing tool that can be repurposed. Standard personas describe the users a product is meant to serve. Anti-personas describe people who will use it improperly or harmfully, so that teams can design against that use (Paul, 2022). Consumer goods offer a precedent: when children were ingesting brightly coloured laundry detergent capsules, manufacturers treated children as an anti-persona and changed the product's appearance and packaging to reduce the hazard. On a social platform, the most important anti-personas are not only vulnerable users but coordinated manipulation networks and state-backed influence operations. Mapping how they would exploit recommendation and amplification mechanisms is a design task, and one teams are trained to do.

Three institutional supports make these methods stick:

The point of these measures is to make the organisation a moral community in which individual judgement is reinforced rather than worn down. Without such structures, even practitioners with strong ethical dispositions are easily neutralised by systems that reward only efficiency and growth.

The strongest objection to embedding ethics in corporate processes comes from Milton Friedman (1970): a company's only social responsibility is to increase profits within the rules of the game. On this view, managers are agents of shareholders, and pursuing moral or social goals with shareholders' money is illegitimate.

Two features of attention-based platforms weaken this objection. First, the market relationship is not an ordinary one. Users pay nothing; their attention and behavioural data are the product sold to advertisers. When the product is effectively human behaviour, design choices bear directly on dignity, privacy and psychological integrity, all protected by the EU Charter of Fundamental Rights (Articles 1, 3 and 7). Beard and Longstaff (2018) call the relevant principle non-instrumentalism: technology should not be designed to treat users merely as functional components of a system. Obligations of this kind are not optional philanthropy.

Second, Friedman assumes that ethics and profit are opposed. Stakeholder theory challenges that assumption. Freeman (1984) argues that a corporation's survival depends in part on the fit between its values, the expectations of its stakeholders and the social issues that shape its ability to sell its products. For platforms, practices that respect user autonomy and limit harmful effects build trust, long-term loyalty and regulatory legitimacy. Embedding ethics in the design process, much as companies already embed Design Thinking or quality assurance, is an investment in the long-term viability of the business rather than a cost imposed on it.

Whether a company has embedded ethical judgement in its product process is not visible in a compliance report, but it does leave observable traces. Investors engaging with technology companies can ask:

Answers to these questions indicate whether a company's published principles are matched by the incentives and routines that shape everyday product decisions.

Regulation can define what platforms must not do. It cannot supply the judgement needed in the countless small design decisions that determine what a platform actually does. That judgement belongs to the people who build these systems, and it can be cultivated: through mandatory, continuous and applied training, and through a design process that treats ethical analysis as a method rather than an afterthought.

Whether this change begins with leadership or with practitioners matters less than that it begins. Executives need to recognise the long-term value of integrating moral considerations into product development; designers need to accept responsibilities that go beyond functional objectives. In both cases the change depends on character, which no rule can impose. What organisations can do is create the conditions in which good judgement is learned, exercised and rewarded. As AI-driven personalisation expands, building those conditions is becoming a precondition for platforms that are both trustworthy and commercially durable.