Information Cascade through Algorithms
Link: http://www.chicagotribune.com/business/bluesky/ct-google-fake-election-results-20161114-story.html
Information cascades are not limited only to the decisions of people, but also to those of algorithms. An article stating incorrectly that Donald Trump won the popular vote during the recent election briefly surfaced at the top of Google’s News results. The article, written by a pro-Trump website, was quickly taken down from the listing. However, it is alarming that a site as influential as Google could display such misinformation, and the results are possibly far-reaching, especially with countless numbers of people using websites such as Google as their source for reliable information. Google made a statement that the result was due to a fault in their news ranking algorithm, and that it is undergoing further improvement. The algorithms are written by humans, and are also susceptible to errors just like a human is when he/she chooses to ignore information in light of a cascade. These algorithms are ultimately vulnerable in some ways to being gamed by agents trying to influence the opinions of large chunks of the population.
Facebook, which acted as another large source of information for people during the election, also fell victim to this issue, with its algorithm for news story selection circulating fake stories. Zuckerberg denied the accusations of Facebook attempting at swaying the election through the promotion of such fake stories. In a post, he argued that it is not easy to discern whether something is true or fabricated, especially since many legitimate sources can sometimes mess up a few details of a story.
In today’s world, we have countless sources of information right on the phone screen in the palm of our hand. Although the information cascade model we looked at in class involved observing the decisions of other people, when we make decisions we also have access to the decisions made by (generally) well-informed organizations and sites. Since these sites tend to aggregate a lot more data when making their decision when compared to your friend, it is as if they have had access to more signals, and as we know, this makes us tend to trust them more over our own observation. This makes websites such as Google and Facebook incredibly influential in propagating information cascades, far more so than most individuals. However, since these sites process so much information, it is not possible for them to have humans make the decisions, and so they must use algorithms to pick the results they show. Algorithms can have mathematical vulnerabilities however, which can be exploited by those who want to utilize the trust people have in these sites to begin a false information cascade. These vulnerabilities are certainly not limited to the amount of people/sites peddling an idea, so they can provide a shortcut for an agent to start a cascade without having to change the opinions of many individuals first.