|
This was a much meatier writeup than I expected from the premise, which usually gets a pretty surface, opinionated treatment. Maybe I misunderstand the capabilities of offered solutions, but I can't imagine that ML would be a tremendously effective tool to gauge content veracity, at least on its own. Maybe it would be a good supplement to a headstrong human effort, but that headstrong human effort seems for major gatekeepers of content (Google, Facebook, etc) like it would be an underpowered cost-center to placate a concerned public at best, and a non-starter at worst. The final suggestion resonated with me a lot: > Finally, we think the research community should continue to build up our understanding of how this content is created; a deep enough understanding could allow us to adapt the strategies of bad actors as a tool against them. It’s time to fight fire with fire. This, to me, while maybe not the most tidy solution to the issue, acknowledges the messy reality of the situation and gives one way to effectively approach it: flood the zone. I think we did see tactical mimicry with things like Correct the Record during the 2016 election cycle. However, I believe that lacked the sort of ethical/factual ambivalence that got the 4chan/reddit pipeline running as well as it did (and continues to do). Essentially, the disseminators of propaganda operate asymmetrically, as they are governed by more dire motivations and by fewer rules. Those who would respond to nullify the effect of that propaganda can't half-ass it, they would need to match the shrewdness and power of what continues to crop up. It's very Brave New World. |