Software often earns trust by behaving predictably. Yet some tasks benefit from variation: shuffling a collection, generating alternatives, or selecting a sample. Randomness gives a rule room to produce more than one outcome.
Chance still needs a frame
A random choice is made within a defined set of possibilities. Choosing a number requires a range. Selecting an item requires a collection. The boundaries matter as much as the mechanism, because an option outside the set can never be selected.
Equal chances are only one arrangement. A weighted choice can make some outcomes more likely than others. That can be useful, but the weighting should be understood before anyone treats the result as an impartial selection.
Variation is not evidence
A single random sample may be unrepresentative. Repeating a selection can reveal a wider range of outcomes, but repetition alone cannot repair a poorly chosen source collection. The method and the material both deserve attention.
Many software generators use a starting value, called a seed, to produce a repeatable sequence that appears random. This can help someone revisit an interesting result. Security-sensitive uses need different care; a convenient generator for visual experiments should not be assumed suitable for protecting information.
The useful question is what variation contributes to the task. Sometimes it opens possibilities. Sometimes a stable result is more valuable.