Social Media Systems and Democracy

• Economically speaking, shallow votes are significantly cheaper to make.
• It’s nearly inevitable that any company that is winning at capitalism cares less about whether you like their product and more about whether you use it.
• It’s a classic systems pitfall: you get what you optimize for, not what you pretend that thing represents. When you create a system to opitmize for most user actions, it doesn’t matter whether you’re imagining user actions represent meaningful engagement. It’s hard to quantify and optimize for meaningful engagement, but a love for data-driven approaches doesn’t justify using a data-driven approach on the wrong data. It is better to use fallible human judgment and intuition on trying to solve the right problem than an algorithmic approach that is guaranteed to solve the wrong one. If you optimize for clicks, don’t be surprised if all you get is clicks.
• For example, YouTube’s shift in the meaning of subscriptions. At some point, some manager or board member decided that increasing subscription numbers was an important goal for YouTube. A YouTuber with 10 million subscribers would look good for the economic health of the company. Successful megacreators make it appear as if investing in creating youtube content is a reasonable choice. And so it was no surprise when YouTube quickly reached the goal of having a youtuber with 10 million subscribers after making some optimizations. All they had to do was change what it meant to subscribe to someone, as well as automatically subscribe all new users to a list of heavily-subscribed youtubers in an opt-out system. The system prefers ten million people shallowly engaging with one shallow content creator than to have those same ten million divided up among a thousand niche creators that they feel a meaningful connection to. YouTube will reap all the short term benefits and face all the long term hazards of cultivating a monoculture.
• if you are browsing in a part of the web that promotes things using internet votes, you are all but guaranteed to only find things that elicit a quick easy user action and then leave the user unsatisfied and looking for more. In practice, inspiring and satisfying pieces of content are dead ends for user actions. Thoughtful pieces of content that take twenty minutes to read get one vote in the time it takes for pretty pictures and amsuing memes to get dozens.
• For companies that run on advertising, this is the economical choice in th short term, though as internet ads become worth less it might not be enough.
• Google and YouTube’s latest attempt at a subscription fee, for example, would have made sense a few years ago, but seems very strange indeed given their recent change in direction.
• The communities that survive best in their isolated bubbles are those that develop scripts for how to respond to conflicting opinions; these automated defense reactions mean you can out-respond anyone who considers their words in real time.
• When voting is asynchronous and public, the fastest reaction gets disproportionate influence on future voter behavior. The fastest reaction is also not usually a very well-thought-out one. (also see stack exchange’s fastest gun in the west problem).
• Gamergate’s mistake was trying to effect real world change, when their existence was only supported by internet systems. Scripted responess and instant outrage are amplified by internet structures, but real world actions are not.