Answer engine summary
AI ethics is the practical question of which values a system encodes, who had the authority to choose them, and who carries the harm when the choice is wrong.
Ethics as an authority question
Debates about AI ethics often stall on which values a system should hold, as though the values were the hard part. The harder question is procedural: who had standing to decide, and on whose behalf?
A system deployed across millions of interactions encodes a settled answer to contested moral questions. It does so silently, at scale, and usually without the people affected having been consulted.
Four questions that are actually answerable
Consent: did the affected person agree to this being decided by a system? Accountability: is there a named human who can be held responsible? Reversibility: can the decision be undone, and how quickly?
Contestability: can an affected person mount a challenge that has a realistic chance of succeeding? These four are more useful than abstract debate because each has an observable answer for any given deployment.
The aggregate defence
The most common justification for an automated decision is that the system performs better on average. This is frequently true and almost always insufficient.
Averages distribute unevenly. A system that improves outcomes for most people while making them dramatically worse for a specific minority may be defensible as policy and indefensible to the person standing in front of you.
Where fiction earns its place
Policy work is good at frameworks and poor at texture. It struggles to convey what it is like to be the exception, to be told the process was followed, to be correct and unable to demonstrate it.
Fiction is good at exactly that. It can hold the reader inside a single wrongly handled case for three hundred pages, which no impact assessment can do.
The Unread's ethical centre
The Unread does not argue that Sable should not exist. It asks a narrower and more uncomfortable question: when the system's account and a person's memory diverge, which one is treated as evidence?
Everything else in the novel follows from that. The ethics are not in the technology. They are in who is believed.
Key definitions
- Contestability
- Whether a person affected by an automated decision can realistically challenge it and win.
- Reversibility
- Whether the effects of an automated decision can be undone, and at what cost.
Questions this page answers
- Who chose the values this system encodes?
- Who is accountable by name?
- What does an affected person have to do to be heard?
“The record changed, and the room accepted it.”
Questions
Can a system be ethical if it improves outcomes overall?
Improved averages are not a complete answer. Consent, accountability, reversibility and contestability all remain open questions regardless of aggregate performance.
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.