Contestability
Whether a person affected by an automated decision can find out why it was made and effectively challenge it.
When not to use it
- As a synonym for explainability. An explanation nobody can act on is not contestability.
- Where the decision is trivially reversible anyway, in which case the machinery costs more than it returns.
- As a substitute for the decision being correct. Contestability is a backstop, not a design goal in itself.
Reach for something else instead
- Reversed burden — require the institution to demonstrate the decision was sound rather than the individual to show it was not.
- Pattern-triggered review — make a threshold of successful challenges automatically force examination of the process.
- -
Read more on the blog
- Nine cases, and none was fixed by a better modelTerritory 6 closes. Nine documented cases, three containing no AI at all, and not one where the remedy that worked was a more accurate system.
- Robodebt: losing quietly to avoid losing publiclyA Royal Commission found the scheme unlawful, crude and cruel. The tribunal had been ruling against it for years, and the department never appealed, so no precedent was ever set.
Further reading
- Royal Commission into the Robodebt Scheme (2023) — the report documents both the reversed burden and the absence of any mechanism turning individual wins into review.
- Parasuraman & Riley (1997), Humans and Automation — why the reviewer in the loop is not itself a contest.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Counting the existence of an appeals process as evidence that appeals work.
- Reading a low challenge rate as satisfaction rather than as a cost barrier.
- Adding explanation after deployment when the reasoning was never recorded and cannot be recovered.
At a glance
Where this sits
A starting point. Nothing needs to come before it.
Computed from the prerequisite graph, not assigned. How this works