Facebook Political Ads and Information Cascades
Facebook has been pressured to make political advertisement funding more transparent to prevent election manipulation. Last May, they created a new policy that promised users the publication of the buyer’s names (ie: “paid for by the GOP”). However, a new report shows that the buyers have found loopholes by purchasing advertisements under obscure names like “a freedom loving American Citizen exercising my natural law right, protected by the 1st Amendment and protected by the 2nd Amendment.” Filing under these names allows buyers to remain hidden, and undermine Facebook’s transparency policy.
We can see how this strategy might influence public opinion by means of information cascades. We will use a similar model as the Herding Experiment:
- Decision to make – A person can either share or not share the political ad
- People share one after another, in a sequence.
- Each person has some private information on the subject.
- A person doesn’t know why another user shared the ad, but only that they did.
Let’s use the example of a foreign power buying misleading ads on Facebook. A user might choose to share the ad if they mistake it to be true and unbiased. Then, a second user might choose to share the ad, but they are influenced by the first user. If they see that the first user shared the ad, they will be more likely to believe it is true. Then a third user sees that it was shared by two of their friends, and is even more likely to not fact check and believe it to be true…etc. The cascading effect occurs, and misinformation spreads rapidly.
So how does transparency stop this information cascade? Suppose Facebook policy was successful, and under the misleading ad, it was written “paid for by a malicious foreign power”. Even if the first user chooses to share the ad, the second user has additional information. Rather than mainly relying on the first user’s decision, the second user can take into account the buyer’s name and think otherwise. That is, no matter how many people share the post, one piece of information is always constant: it is known who is backing the ad. Because there is a piece of global information, a user doesn’t have to rely on inferring other people’s private information. This breaks our model, and weakens the cascading influence of the network.