Automation bias, a relatively new idea, is a particular class of errors that a person might make in highly automated environment. Typically it occurs when an automated work environment is monitored by humans (common examples range from nuclear reactors to a spell check/autocorrect feature).
In our modern world many decisions are made by computer programs rather than by humans. Therefore, since humans do not think like computers and use human criteria they will make "errors" (or decisions that are not the same as a computer) or miss errors the automated programs commit.
An example of automation bias would be an incorrect word substitution in an article by an autocorrection that a human editor misses because they relied on the automated program to catch any mistakes.