Statement

Governments and companies use AI and other automated systems to decide who is checked for fraud, who is offered care, who can rent a home, who owes money. The people affected rarely see these decisions. When one becomes visible, through a court case, an inquiry or an audit, the findings are reported for a few days and then filed away.

I keep a record of these cases. Each work is built from a single case: one system, the decisions it made, and the people those decisions reached. I document the case from public sources, code it with a fixed method, and draw it as two images in which every mark can be traced back to the record.

The questions are the same for every case: what the system decided, who was affected, how sure we can be of the cause, who was able to step in, and whether the damage can be undone.

Public debate now uses "AI" for all of these systems. Several of the cases in the first series used no machine learning at all. I keep that distinction in the record, because it changes who designed the decision and who can be held to account for it.

The first series documents harm. The method can also be applied to decisions that help people or leave them unaffected, and later series will include such cases.