assistance with an AIML idea

Hi,
Im using python for my bots that use AIML and I was wondering about something. I usually keep logs of what my bot says and its responses from other people. Now what Im thinking is having another python script that goes through these responses and generate or update its current AIML markups based on reversing the roles of responses and the statements by and from the responder (does this make sense?)
For example if someone asked if today is sunday and the bot's response is that it doesnt know and the responder tells it to ignore it. The script will update the AIML so that if someone tells the bot that they dont know what day it is today, the bot will tell them to ignore it.
 
Has this been done before? Im just thinking in a very general manner here, im sure there are alot of ways in which this can be improved.
Any comments or thoughts or obstacles with this aside from the obvious overfitting/overspecialized AIML; but I intend to counter this with more general chats.
Cheers