How would you design an unbiased news recommendation algorithm? You can't - the definition of 'trending' has to include decisions and values. Is the most-share

How would you design an unbiased news recommendation algorithm? You can't - the definition of 'trending' has to include decisions and values. Is the most-shared story always relevant to a wide audience? An organised brigade (think a grassroots political campaign) could use coordinated sharing to bootstrap their way onto the trending list. That would devalue the trending list. More broadly, you have to teach code to autonomously infer relevance from ambiguous signals. And once you do that, you're still not unbiased; you've simply delegated bias to your audience. Humans are not perfect, but I would still privilege human editorial judgement above algorithms for the time being. But I agree the question of what issues and perspectives people are exposed to, and who controls what they see, is incredibly important. Just too important to abstract away to a machine.