A heuristic is a generally recognized rule of thumb that helps to achieve usability.
Heuristics are used to:
- prevent well-known usability problems by teaching the heuristics to designers;
- find usability problems in interactive systems, for example by using them in a heuristic evaluation or a usability inspection.
Examples of heuristics: Speak the users’ language; Be consistent; Provide good error messages.
Heuristics are often used in sets of 8 to 20 heuristics.
Heuristics are sometimes confused with user interface guidelines.
A bit of history
The first set of heuristics was published by Rolf Molich and Jakob Nielsen in 1990.
A revised version of these heuristics was published by Jakob Nielsen in 1994. The revised version expanded the scope of the heuristics and described each heuristic in a compact and elegant way. I did not participate in the revision.
Quality criteria for sets of heuristics
1. Generally recognized
A reliable set of heuristics has stood the test of time
A heuristic must be easy to understand.
Example: People tell me that the following heuristic is hard to interpret: “Match between system and the real world“
A good set of heuristics comes with plenty of examples that translate the heuristics into everyday language. I recommend at least two examples per heuristic, ideally a software and a hardware example.
Limit yourself to a few, important heuristics. I recommend sets of 10 to 15 heuristics.
Each heuristic should address exactly one usability issue.
“Speak the users’ language” is a simple heuristic.
“Match between system and the real world” is a complex heuristic. It can be decomposed into several simple heuristics, for example “Speak the users’ language”, “Follow real world conventions” and “Match the real world.”
Complex heuristics can be difficult to comprehend.
Sets of heuristics
I am impressed by the “Guidelines for Human-AI Interaction” by Saleema Amershi et al., which was published in 2019. The 19-page article describes a set of 18 guidelines with example applications and example violations. Most of the guidelines are well-suited as heuristics even for non-AI systems.