Aaron Wall posted over at Threadwatch and also commented on his personal blog how fantomaster had some good responses, aftrer reading these responses I agree they are worth reading. fantomaster makes good points.
My take is that as long as content is relevant to the readers it should not matter how it is derived. Programing software to identify content, say, by keywords, is no different than instructing a junior employee to source articles with a brief, say, ‘find some articles about dodgy car dealers in Birmingham ‘. In the end the readers are the judge of what works and what doesn’t surely, they will just move on if content is irrelevant.
After all nobody ever had a long term relationship with Eliza did they?