Wikipedia has randomness as a lack of bias or correlation, with clear distinction from something that is merely unpredictable, as it is commonly misused. Then it launches into a religious discussion of free will vs. determinism.

Sensibly dealing with randomness is a hard problem in modern science, mathematics, psychology and philosophy. Merely defining it adequately, for the purposes of one discipline has proven quite difficult. Distinguishing between apparent randomness and actual randomness has been no easier. In addition, assuring unpredictability, especially against a well-motivated party (in cryptographic parlance, the "adversary"), has been harder still.

Some philosophers have argued that there is no randomness in the universe, only unpredictability. Others find the distinction meaningless. Note that the bias that "everything has a purpose or cause" is actually implicit in the expression "apparent lack of purpose or cause". Humans are always looking for patterns in their experience, and the most basic pattern seems to be cause/effect. This appears to be deeply embedded in the human brain, and perhaps in other animals as well. For example, dogs and cats often have been reported to have apparently made a cause and effect connection that strikes us as amusing or peculiar. (See classical
conditioning.) For instance there is a report of a dog who, after a visit to a vet whose clinic had tile floors of a particular kind, refused thereafter to go near such a tiled floor, whether or not it was at a vet's. It is because of this bias that the absence of a cause seems problematic. ~Wikipedia

Random number generators use a procedure called linear congruential method to create a sequence of “random numbers” within the interval [0, k). First, four non negative integers are selected, a, b, m, and x_1. x_1 must be less than m and is the first number in the sequence, called the “seed.” To get the second number in the sequence, calculate y = ax_1 + b and divide the result by m to find the remainder, which becomes x_2. Iterating finds as many numbers as you like for the sequence. However, there are only m possible values so the sequence repeats itself after x_m. In practice, m is very large, so the results appear to work well, even though every linear conguential generator eventually enters a deterministic repeating pattern. This is why these sequences are called pseudo-random (Probability for Risk Management, Stewart & Hassett).