Daily Tao – The Hype Machine, Sinan Aral – 5

One thing Damon Centola and his team learned while studying networked crowds was that crowd wisdom can be improved by social influence when the most influential individuals are also the most accurate. Networks that put more weight on the opinions of those with the greatest accuracy, reliability, or access to the truth can perform even better than independent crowds. (The call to “listen to the scientists” on climate change and pandemic responses comes to mind.) But how can we engineer the Hype Machine to put more emphasis or weight on these peers? The Hype Machine is already replete with feedback mechanisms—they’re just designed to feed back the wrong signals. Take likes, for example. The “like” button is the engine of the attention economy. It is designed to capture our attention, to elicit our approval or disapproval of the content we see, and to incentivize us to produce more content by giving us a dopamine rush. The more we like social media content, the more engaged we are and the more opportunity there is to serve us ads. Likes serve another purpose, however, because as we like more content, we signal our preferences to the Hype Machine, which enables the ads served in those impressions to be targeted at the right people. Now imagine a world in which we went back to the invention of the “like” button and replaced it with a “truth” button (for content we think is true), or a “reliability” button (for content we think is from a reliable source), or a “wholesomeness” button (for content that is good for us), or an “educational” button (for content that taught us something). The thought exercise forces us to rethink the feedback we see on social media and to consider how code changes could reengineer the Hype Machine toward positivity. In fact, we already use norms to participate in this reengineering effort. For example, we have, as a society, largely accepted that on Twitter “retweets do not necessarily mean endorsements,” because we have adopted the ubiquitous “RT ≠ Endorsement” tag to reengineer the meaning of a retweet. Research shows that feedback is essential to our ability to process social information in collectively useful ways. So how we formally and informally design that feedback will help shape how the Hype Machine shapes us. What if every time we posted content to social media we were given the option to relate how “confident” we were in the material, or if we were asked whether we thought other people’s posts were true? How long would it take before all Americans knew the correct capitals of all fifty states? How long would it take before everyone in the United States knew their Miranda rights? Feedback is not just about weighting the information we receive in socially beneficial ways. It also allows us to adapt the network itself. Would we change who we are following on Twitter if their profiles displayed a “veracity” score that recorded the percentage of their posts that were fact-checked to be true or false? If decisions on who to follow were affected by how truthful people were, and if truth tellers amassed larger followings, would everyone be inspired to be more truthful? Would that limit the number of reshares of false information and the followers of false-news-peddling accounts?

In a network, the most influential individuals can shape what the crowd thinks and have the most impact in increasing overall crowd wisdom. This dynamic is a double-edged sword, allowing for great harm or good.

One suggestion in this excerpt that stands out is simply how we can reframe the audience mindset towards posts by simply changing the “like” button to something else. Such feedback, will not only reframe the audience mindset and get them to think critically, but also might be a sign of feedback to the poster on their thoughts.

Being the cynical individual that I am though, such mechanisms probably wouldn’t work and people who are “wrong” might end up doubling down on their views when faced with contrasting evidence. Also, discussions online largely ignore context, and statistics devoid of context can be found to support almost any viewpoint you want.

Share