Naturally, it sounds like it is only best for our users if our design could reduce user's cognitive workload when performing a task.
However, there are also theories that explain the need for a person to perform deep processing in order for one to remember for a longer duration on using a design. The Level of Processing theory, for instance, preaches that through elaboration, a person can remember something better.
This seems to contradict because on one hand, we want to reduce user's cognitive workload. On the other hand, we want users to remember our design's usage so that they don't always have to recall (which increases memory load) its usage, but rather recognise its usage. And to do this, based on the Level of Processing theory, we should design some thing within the design, say the interface of the design, that requires the user to do some deep processing in order for him to remember. But this would then incur workload on the user's cognition.
In other words, to reduce user's memory load, we have to increase their cognitive workload as they have to perform deep processing. Then in this case, should a good design still attempt to lower a user's cognitive memory load at the expensive of increasing user's cognitive workload?