A heuristic is another way of saying rule-of-thumb. It’s a thing you refer to when making a decision when fully considering the situation you’re in (from first principles) would be too time-consuming, expensive or about which you simply do not have trusted skills available.
The availability heuristic is what happens when you rely on an immediate example of something when considering a topic. The flaw here is that the immediate example you’re thinking about may in fact have nothing to do with the topic at all.
Here’s an example from real-life.
At work, on our internal network, a new version of an application is being deployed. As part of the deployment process post-deployment tests are done to confirm 1) the correct code was deployed 2) that the correct code was deployed correctly 3) that correct code was correctly deployed and that it did not affect the environment it was deployed into.
During testing it is observed that the update functionality of a particular screen is broken.
Because the problem is discovered as part of the deployment process the availability heuristic is the reasonable, but incorrect, assumption that faulty code introduced during the deployment is the cause of the problem.
Actually it turns out that the broken functionality was present in the code-base for several months, and despite about four deployments and an equal number of regression tests the error did not manifest until a particular combination of three fields were updated which results in a fourth value (not shown on screen) being changed. When this value is changed, and only when this value is changed, will the code enter a path triggering the error.
The problem was further compounded by the fact that the required combination occurred roughly 1/45 times (when the form was edited due to correcting a mistake, also assuming the combinations are uniformly distributed).
It just happened to be the case that on that night, within 15 minutes of the code being deployed that post-deployment testers decided to update the form using that combination of values that then triggered a months-long bug.
ALWAYS CHECK YOUR ASSUMPTIONS.