I've been thinking for many years about e-readers and the consequences of tracking user reading behaviors. One likely outcome is that it will result in a wave of "pop literature". As I understand it, pop music is already heavily customized to appeal to the masses based on tracking loads of data, going as far as analyzing past hits to identify the combinations of tunes, vocals, instruments and audio effects that work best to create new hits. Record labels and producers apparently have it down to a (data) science.

Some may point out that mass-produced "pop literature" is not really new: for instance, the entire Hardy Boys/Nancy Drew series was just that. However, since this is not a widespread practice so far, (almost) nobody has complained about it.

But once this data starts getting used more and more, and feedback loops get more efficient, eventually pop literature may suffer from the same criticisms as pop music. In a way it's good: you're giving the customers what they want. But it also has the downside that creators will shape themselves to their audiences rather than express themselves freely, limiting the variety and individuality of creative output. As long as enough creators choose the latter at least once in a while, I think it's a worthwhile tradeoff.

However, this is just the low-hanging fruit and does not explore the full potential of e-readers. E-readers need not just be dumb terminals that phone home with reader behavior -- they're fully capable general-purpose computers, and we should use them as such! We should be thinking of e-readers that "adapt" to the reader and the material. And not just an "average reader" profiled aggregately over an entire audience, but individual readers. Let the authors freely express themselves, and let technology bridge the gap between the creative work and the reader.

I haven't thought it through completely, but we certainly have the technology -- things like NLP and machine learning and user profiling and heuristics can go a long way in performing basic adaptive tasks. User likes short chapters but the book has long ones? Re-segment the text to match the user's preferences! (Maybe using a variant of text tiling.) User looking for new reads? Recommend books based on similarity of reading behaviors! They could even change the language based on user profiles. User too young? Tone down the profanity! User too old? Replace "really whips the llamma's ass" with "totally funkadelic"!

Technology is, of course, not advanced enough, so we would still rely on human input to a great extent. Editorial staff (or other readers) would need to provide tons of annotations and metadata that can help e-readers adapt better.

But eventually it may happen that for the same material two users may end up having very different reading experiences.