Their system automatically looks for negative patterns of behaviour... but those patterns do occasionally have to be manually taught to the system.
So the big thing where descriptions come in handy is when somebody is reported constantly, but without exhibiting any patterns that can currently be detected. That's when a Rioter can look at the case and see whether this person was being a douche in a very original way, or whether they just had an unlucky streak of receiving false reports.
Now let's say one of the reports says "He constantly typed extreme slurs backwards to avoid detecion", now the manual reviewer knows what to look for and that obviously speeds up things a lot. At the same time, if nobody every described it, it could be that the employee misses the actual meaning behind statements like "reggin" and "toggaf" and they get off scot free.
Short version:
If you're confident the automated system can handle it, it's not necessary. But it can't hurt, and it definitely helps for manual reviews.