The two arguments presented - to keep a legal path for a higher priority routing - during late 2015 Trilogue Negotiations that framed the directive the way it is:

Driverless cars Medical applications (remote operations rooms were mentioned)

You have to give to telecoms - they built the infrastructure and are desperate to design some services that could become additional source of revenues. For now, it seems, the door is shut - maybe with the new automation coming they could dig it up - akin to "you don't want your house to send fire warning to city grid too slow, do you?".

That's a very interesting point you are making. Would seem true if one establishes that any off-the-personal-baseline analysis by a human-wetware cost more, the further off the analysed position is.

Your idea would to be to sneak in notions that are outside of a personal, or group, bubble. Do it gently and slowly. It would work if the positions would be presented with known and familiar analogies. In some setup that could work, we think in analogies.

One could argue, that that's what we do all the time - convincing ourselves and others of value of existing or new positions while ever-slightly changing our own positions. Do we want to dress that process up with machine learning, though?