May 27, 2023: Misinformation
Good afternoon. Today we will look at the nature of misinformation.
This discussion is prompted by a recent article at The New Atlantis on deepfakes. A deepfake is a piece of synthetic media designed to resemble something else. With recent advances in generative artificial intelligence, it is feared that there will be an onslaught of convincing deepfakes that will destroy the general public’s ability to distinguish fact from fiction.
According to the article, these fears have been large speculative, and so far they have not materialized in a significant way. One problem is that creation of a deepfake still requires sophistication and resources; more traditional methods of deception are still more cost effective, even for malign state actors. A tendency in the development of AI is that, over time, results that were once difficult to produce become easier, both in terms of technical skill and compute required, so this could change.
A broader problem is that misinformation discussions tend to ignore demand. It helps to think of the information space as a market, with similar properties to any other market, including both supply and demand curves. Common points in the exposition of the supply side might be found in Richard Hasen’s book, Cheap Speech: How Disinformation Poisons Our Politics―and How to Cure It. Hasen argues that the lemon problem prevails in the market for information.
If you will pardon a brief diversion, in the used car market, a “lemon” refers to a low-quality used car. The used car market might have both high- and low-quality cars, but it is difficult to distinguish between the two. Therefore, sellers tend to put low-quality cars on the market, buyers have to assume that cars for sale are of low quality, and the market devolves into a lemon market. This state of affairs constitutes a market failure. Hasen argues that a similar phenomenon prevails for information. It may be difficult to distinguish between good and bad information, and so “sellers” have an incentive to put out low-quality information, “buyers” have reason to believe that the information they see is bad, and the market degenerates into misinformation.
But it is simply not sound to assume that news consumers desire accurate information. For an explanation of the demand for misinformation, I would start with Guy-Uriel Charles’ post and Ilya Somin’s article on Reason. Both articles refer to the large volume of research that establishes the correlation between partisan affiliation and the propensity for misinformation, and the widespread nature of motivated reasoning. Yet these observations often don’t filter into the discussion on misinformation.
Revelations coming out of the Dominion Papers earlier this year paint demonstrate how the 2020 stolen election notion was driven by FOX’s viewers, not the network. Tucker Carlson, who was forced out over the scandal, wrote of the conspiracy theories, “Our viewers are good people and they believe it” and clearly did not believe them himself. I recently highlighted an article from the Risk Monger (David Zurak) on a similar dynamic within Greenpeace, where their campaign against Golden Rice in the Philippines was driven by an activist base, and they had to go ahead with the campaign for fear that, if they did not accommodate the activist base’s ideology, the base would bolt for another organization that would.
Deepfake worries are part of a broader moral panic about the impact of novel technologies on the information ecosystem. And I do use the phrase “moral panic” deliberately. Several recent studies (e.g this and this) have cast doubt on the notion that social media is a driver of political polarization. See this Toward Data Science post on the problems around Facebook’s notorious 2014 emotional contagion study (among other problems, the small magnitude of the effect, which can be established with sufficiently large amounts of data, is neglected in most popular discussions of the topic).
The nature of misinformation is something else that needs to be interrogated much more closely. Popular discussion tends to consider misinformation as a polarizing binary; either something is unambiguously misinformation or it is sound information. But clearly there is a spectrum of truth, with the vast majority of ideas occupying a state that does not admit an unambiguous classification. Many ideas are a matter of interpretation (a couple of which are highlighted in the Quick Hits section below) which are not easily classified. And of course there is the widespread and well-founded critique that “misinformation” and “non-mainstream” are often used interchangeably, when in reality there is at best a weak correlation between the two.
Now, the capacity for AI to facilitate the spread of misinformation has been rolled into a generalized fear about the technology. This further increases the need to understand what misinformation is really about, and right now that understanding is very poor.
Quick Hits
Roger Pielke Jr. discusses the recent “beyond growth” event hosted by the European Parliament. This is only a small part of the inroads the degrowth movement has made into high levels of European (and around the world) policymaking, generally in an opaque and undemocratic fashion. If one looks at ideas that are not explicitly called degrowth, but expound the principles of the degrowth movement, then one finds them to be widespread.
According to official statistics, India recently surpassed China as the world’s most populous country. However, one might conclude from research by Yi Fuxian that this milestone occurred a few years ago, as he has convincingly documented ways in which the Chinese government and World Population Prospects have inflated population figures. Not so fast, Yi argues in a new editorial, because India’s figures might also be inflated. According to Yi, the international community through the UN has abetted population control policies in both countries, which have included forced abortions and forced sterilization, policies which continue to this day. Thus there is reason to fudge the numbers. One must wonder if WPP’s figures are reliable for other countries.