When the "artificial intelligence" in SMAX suggests knowledge articles and solutions based for a new trouble ticket, it is history that counts - how many times a knowledge article or a solution has been used. But when new fualts occur, or new knowledge is gained, this often is a more correct solution than the old one - even though it has been correct in the past.
Let me illustrate this with an example: If you plan a drive with Google Maps, it will suggest the route - and expected duration - based on lots of data. But while you are driving, if the traffic pattern changes (e.g. due to an accident) it will automatically change your routing, because the new information beats the old, even though it may be based on much fewer data points.
This automatic adaptation to a changing environment is the intelligence that I would like to see in SMAX, and a first step is giving priority to new information in place of historic data.