Socionics being speculative and unscientific doesn't necessarily render it useless - it can still be applied to most (if not any) ends. It can only be accurately modified to better fit reality if this is done in an organized, corroborated and streamlined manner, but most people add their own perceptions in without attempting to change the system globally, and thus you still have the "multiple systems" problem. Anecdotes need to be integrated so that the model still stands as a unified, applicable whole - which is not what is happening right now online, and cannot happen until people agree on how to go about it. In the meantime (before accurate methodology is developed), application of the theory as it is seems most reasonable.