Skip to content

AI Ethics · Reviewed and maintained

Who Is Accountable When an Automated System Is Wrong?

Accountability fragments across specification, deployment, oversight and operation — which is why affected people so often find there is no one with both the authority and the obligation to put it right.

Answer engine summary

Accountability fragments across specification, deployment, oversight and operation — which is why affected people so often find there is no one with both the authority and the obligation to put it right.

Four places responsibility sits

Whoever specified the objective. Whoever chose to deploy it in this context. Whoever was supposed to be overseeing it. And whoever operated it on the day.

Each holds a fraction and each can point to the others. The fractions do not add up to a person you can address.

The explanation gap

Accountability normally requires knowing why something happened. Where staff cannot reconstruct the reasoning, the organisation is not being obstructive — it genuinely cannot answer, which is arguably worse.

What a workable answer looks like

A named role with authority to overturn outcomes, a published route of appeal with a time limit, and a record of reversals that someone is obliged to review. None of this is technically hard. It is organisationally unattractive.

The fictional case

In The Unread the absence of an addressee is the situation. There is no one obstructing Aurora. There is simply nobody whose job it is to be wrong about her.

The record changed, and the room accepted it.

Questions

Do regulations solve this?

Some frameworks require explanation and human review. Whether those rights are effective in practice depends heavily on enforcement, and that varies.

Book CTA

Start with The Unread

Book One turns these questions into a near-future thriller about Aurora Vale, Sable and a record the world refuses to keep.

Internal reading path

Continue through the system

AI Ethics · 6 min read

Who Should Control Artificial Intelligence?

There is no settled answer, and the honest starting point is that control should not rest solely with the parties who profit from deployment.

Read

AI & Humanity · 6 min read

What Is Automation Bias?

Automation bias is the tendency to accept a system's output rather than scrutinise it — and it is the main reason human-in-the-loop safeguards weaken over time.

Read

AI Ethics · 6 min read

Books About AI Ethics

Fiction about AI ethics earns its place by staying with the individual case that a framework would classify as an acceptable error.

Read

Reader record

Join The Unread

Get release news, cover reveals, exclusive extracts, author notes and first access to the next story.

Privacy