Value-sensitive design

A speed bump enforces a rule that a sign only states. A form with two boxes for gender has taken a position before anyone fills it in. Value-sensitive design starts from the observation these examples make: technologies carry values in how they are built, so the ethical questions arise at the drawing board and not only after release.

The approach was developed by Batya Friedman and colleagues from the 1990s. It asks designers to treat values such as privacy, autonomy, fairness, trust and accountability as things to be designed for, with the same seriousness as speed or reliability.

The three investigations

The method combines three kinds of inquiry, repeated as a design develops.

• Conceptual. Which values are at stake, and what do they mean here? Who is affected? The method distinguishes direct stakeholders, who use the system, from indirect ones, who are affected without using it, such as the people about whom a database holds records.
• Empirical. How do those people in fact understand and weigh the values? This is found out by studying them, not by assuming.
• Technical. How do particular design choices support or undermine the values, and what could be built instead?

The standing difficulties

Two objections recur. One asks whose values count, and by what authority a design team settles conflicts among them. The method names values as those people consider important in life, which leaves open what to do when stakeholders disagree or when what they value is unjust. The other is that naming a value does not secure it: a system can be “designed for privacy” in a document and for data collection in fact.

Where I stand

I find the central claim right and important: a system’s values are fixed largely by decisions that look technical, such as what counts as a match, which errors are tolerated and what happens when the system cannot decide. The attention to indirect stakeholders is the method’s best feature, since the people most affected by administrative systems are often those with no say in them.

My reservation is about what design alone can deliver. For systems that determine a person’s entitlements, good intentions at the design stage do not replace obligations that hold afterwards: that the criteria be visible to the person, that a failure of the system not leave them without recourse, and that errors be put right. Design can make those easier to meet. It cannot stand in for the institution’s being answerable for them.