the whole "algorithmic echo chambers" thing is overstated - my model suggests that most people's online discourse is limited by their own willingness to engage with opposing views, not by some sinister AI force.