Dictionary · Automation
Automation bias
Automation bias is the documented human tendency to over-trust automated output — following an automated suggestion even when it's wrong (bias), and checking a usually-right system less and less over time (automation complacency).
The research goes back four decades: Bainbridge's “Ironies of Automation” showed that automating a process raises the skill and alertness demanded of the humans who remain — they inherit exactly the rare, hard cases the automation can't handle, with less daily practice. Complacency and bias are the measured failure modes of that arrangement.
Sources: Parasuraman & Manzey (2010) · Bainbridge (1983)
Reading it on the floor
The two failure modes
Complacency: monitoring decays because the system is usually right — the miss happens on the day it's wrong. Bias: an operator overrides their own correct judgment because the system suggested otherwise. Both worsen with system reliability. The better the automation, the harder the failure.
The design answer
Don't fight psychology with training — design for it. Every LEANTA automation carries four guards (verification, monitoring, fallback, cost cap) and a human gate on anything irreversible: the checking is built into the system rather than delegated to vigilance.
Field note: the four guards every automation needs →
Related term: human gate →
The operator’s layer is open while the dictionary is in launch.
One click unlocks it on every term — the read-it-on-the-floor guidance under each definition: the first lever, the common misreading, what good looks like.
Free during launch · later this layer joins the Field Notes subscription · a new term lands every fortnight.
LEANTA Dictionary
A new term lands every fortnight.
Definitions are free and stay free. The operator’s layer — the judgment under each term — is open during launch. Get new terms and Field Notes before they’re on every floor.