They play a very important role, but so far the evidence about how effective they are is unclear- the evidence the UK government produced suggests that it does help, but they haven't been able to show a consistent pattern in all areas. A self-access student could, say, do an exercise on paper and check their own answers with a key, or do exactly the same on a computer and be graded by the computer. While the latter seems more interesting to many, is it actually more effective? There are other ways, access to a huge range of materials over the internet, in which is is effective. When a computer is doing exactly what could be done by other means, then I don't see that it is making that much of an improvement, though psychologically it may be better, but it does allow far greater access, which is the stand-out point for me.