The AI has trouble knowing what is important in the conversation. It seems like such difficulty creates an incentive for a user to ask someone something again instead of looking it up in history (although this also may be a signal for AI that something in the history is important).
Other reason may be that it is used by services - they want as long history as possible to find out what kind of person you are in case you are being investigated.
Anyway these artificial obstacles most often have sinister motives.
Other reason may be that it is used by services - they want as long history as possible to find out what kind of person you are in case you are being investigated.
Anyway these artificial obstacles most often have sinister motives.