Human errors are actions that result in unintended or undesired outcomes. Understanding how they occur can help us in design, as explained by HCI Professor Alan Dix:
There are three main types of human error: slips, lapses and mistakes. James Reason has described them in detail in his book, Human Error (see references):
Slips are errors of execution. In the real world, nicking your finger with a knife while trying to slice an onion is an example of a slip. With technology, a slip might be as basic as clicking on the wrong line in a list. Note that slips are unintended actions.
Lapses are also errors of execution but a failure to recall the plan you’re following correctly. When we set about reaching a goal, we formulate a plan for the tasks involved. Omitting, repeating or misplacing (in sequence) one or more steps in a plan would be considered a lapse. Finding a dish of spices that were supposed to be in the finished curry you were making is the consequence of a lapse. With technology, forgetting to log in to a website to change account information is a lapse. Lapses are also considered unintended actions relative to the original plan. Consequently, slips and lapses are often considered together by some authors.
Mistakes are errors of intent (sometimes described as planning errors). They are intended actions but are not appropriate for the problem at hand. Pulling on a door that can only be pushed is a mistake, often brought about by poor use of affordances. A door with a handle affords pulling. The correct affordance would be a push plate. Significantly, mistakes often occur when we misunderstand the state of a system. A simple example is confirming an operation that has already been completed.
In the action cycle, illustrated below, slips occur in step 2 (Action), where the planned actions are executed. Here is where we would click on the wrong item or drop the object we are dragging across the screen. Similarly, lapses occur in step 2 but are attributable to memory failings rather than operational difficulties like hitting the wrong key or other random events. We would either fail to perform a planned action or perform it out of sequence.
Adapted from Don Norman's action cycle
Mistakes are more complex. They can occur at almost any other point in the action cycle, ranging from an unachievable user goal (step 1) to an incorrect evaluation of the new state of a system (step 4). Note that if in step 2 you click on the wrong item, it is a mistake or a lapse if you intended to do that, but a slip if you did not. The outcome is an error in either case!
We can prevent slips by making it easier for users to perform the correct actions. But we must consider the broad range of abilities and contexts of use resulting from increased technology adoption. For example, older users often have reduced motor control, and their skin's dryness can be problematic for touchscreens. Many users have less-than-perfect vision, so ensuring adequate spacing, contrast, and text size can significantly reduce slips.
Other means of preventing slips result from considerations of error recovery. In most platforms, a longstanding rule is that any unrecoverable actions require a confirmation dialog. Where a slip is recoverable, but users may not readily know the remedy, it is common to see temporary popups informing users of the action taken and allowing them to undo it.
An essential method of preventing slips is by disallowing inappropriate actions. For example, radio buttons allow only one selection, so it is impossible to make multiple choices no matter what the cause. Similarly, disabling pushbuttons that are not currently relevant means that they cannot be clicked. Preventing passengers from selecting return dates before departure is based on the same principle. Note that this last example helps to avoid both slips and mistakes. It would be a slip if users clicked the wrong row of a calendar, but a mistake if they were not in the correct month due to an error in perception or evaluation.
Preventing errors requires a good understanding of users’ needs and behaviors. We must also be aware of users’ expectations and familiarity with the terminology used. Testing with users is a necessary part of this.
Because errors result from a misunderstanding of a system's state, good feedback is essential. Feedback must be:
Well-timed and well-positioned: As close as possible to the relevant area of the interface in both time and space. Complex systems may require dedicated panels for feedback and the current state(s).
Appropriate and informative: Feedback, particularly error messages, must be meaningful to users and inform their next steps. Any message that users will not understand should be written to an error log instead. Do not treat users as messengers!
Here’s the entire UX literature on Human Error by the Interaction Design Foundation, collated in one place:
