@Unfiltered/AI: Bridging Bytes and Ballots | Edition 3
By Lacy Crawford, Alan Rosenblatt, PhD, and Craig Johnson, Unfiltered.Media
Welcome back to the Bridging Bytes and Ballots newsletter—helping you navigate the intersection of AI and politics. This week Alan Rosenblatt, partner at Unfiltered.Media, a digital and social media strategist, organizer, professor and thought leader with over 30 years’ experience in the field, is joining the conversation. Let’s get to it.
Meta is Polarizing, What’s New?
Meta let researchers study whether it stokes polarization. The results were polarizing.
Columbia Journalism Review
Lacy: Craig, what are your thoughts on social media and polarization? I feel like it’s a chicken and egg problem. In my view, polarization was getting worse even before social media.
Craig: Polarization is the great debate in politics, and it’s also what makes this field one of the hardest to study and draw strong conclusions from. The best that political science can generally prove is correlations—and this is kind of a perfect example of it. Which came first—social media polarizing us? Or economic inequality, racial reactionaries, and a 40-year mission by the right to push further to the right? I always think it’s interesting that the democratization of primaries in the 1970’s directly overlaps with the rise in polarization in America. Let’s follow “correlation” to its logical conclusion—does that mean democracy causes polarization?
Lacy: Alan, why do we think Meta would allow researchers to even conduct this study? In the past they’ve been cagey on letting folks see the data. Can you walk us through how this study went down?
Alan: First of all, Meta let this happen because it was a Meta-sponsored study, and they thought they would benefit from the apparent transparency while being able to get away with drawing the conclusions. They failed at that. But here’s a quick explanation of how this study went down: Meta provided unfettered data access to a group of academic researchers from the Center for Social Media and Politics (CSMaP) at NYU. Four articles covering their findings were published in Science and Nature. They looked at the effect the Facebook newsfeed and its infamous algorithm have on political polarization and knowledge. The studies found that frequency of seeing political and policy news stories in their newsfeed increases news knowledge, but does not have much effect on polarization over the period covered in the study.
This opens questions regarding the effect of long-term, more frequent exposure to one side’s messages compared to the other. If one side dominated the frequency of delivering messages to people over time, would that cause a shift in their “news knowledge” enough that leads to a shift in their political position? Does the frequency and sources of reshares add credibility to the information shared? Would the ideological composition of the people someone follows affect the bias and influence of reshares they see in their newsfeed (self-ghettoization)?
If that frequency is artificially enhanced with bots resharing content, whether they are AI bots or not, that can create a distortion of online public opinion and an obvious effort to manipulate it. If the news being reshared by bot networks is AI-generated news designed to be compelling to specific audiences, that is a whole new level of manipulation. There is a lot more to unpack down this rabbit hole.
Craig taps into an important point that runs through this research, namely that social media is revealing underlying polarization already in the polity. The research on Facebook’s newsfeed suggests that the focal point of outreach there should be on raising levels of knowledge for the long-run, because you are not going to shift people ideologically in the short-run.
Why This Senator Thinks We Don't Need More AI Laws
Congress unlikely to pass sweeping new AI laws, key GOP senator says
Politico
Lacy: Craig, in previous newsletters we’ve heard your thoughts on members of Congress and technology, particularly on their lack of understanding when it comes to AI and social media at large. How can we help close that gap? Alan, same question.
Craig: The Senator at the center of this Politico article knows so little about technology and its impact that he believes current laws, which can’t keep up with Web 2.0 issues as it is, can somehow help regulate something as complex and as multifaceted as AI. A big part of why the Republican Party quite frankly sucks is that they are incapable of having interesting policy ideas outside of their tired mantra of “government = bad.” This attitude of “we don’t need more AI laws” stems directly from that. This approach is going to be as effective as it is for gun control and mass shootings, “Enforce the Gun Laws We Already Have.”
Alan: Congress has always moved too slowly on digital policy and with too little understanding of how it all works. Compounding this with the need to protect First Amendment rights and the deep partisan division in Congress, anticipating no forthcoming AI laws from Congress is neither a surprise nor necessarily a bad thing. We do need regulations and laws for this stuff, but guess what? They have to be effective ones.
Artificial Intelligence is Reshaping Business
From OpenAI To Stripe, Artificial Intelligence Is Remaking The Cloud
Forbes
Lacy: I’m going to learn how to code so I can make something with AI. I’m sorta joking but I know there’s a large number of people like me, with no technical background in generative AI, trying to jump in the fray. I dislike this word, but how can this “disrupt” the political industry?
Craig: There has been an unbelievable surge in small, and even hobbyist, AI creations. Between people modifying the recently released Llama 2.0 model in various useful home chat bots, and NPC conversation integration in Skyrim, the sky's the limit with these new tools. People had no idea what happened when Nvidia released the RTX 3090 card—but it basically enables enterprise-level speeds on your home computer. It’s a $1,400 MSRP card that’s doing AI computations nearly as fast as the $4,500 A6000 card, but it’s widely available at even lower prices. Install any operating system and you’re good to go. This means anyone can set up their own custom tool to make whatever goal, nefarious or not, x100 times easier to achieve. Want to promote climate denialism propaganda customized for individuals? Now anyone in their basement can do it.
Alan: Each industry has its own language, value systems, and context. There is an inevitable need to integrate AI into each of them in a way that is matched to those differentiators. Much of this will depend on the good work of people who understand both how to build customized AI/machine learning technology AND what the needs and cultural expectations are for their respective industries. In politics, we see this with respect to the messages AI generates and how well, or how poorly, it matches to the norms of the audience for which it is generated.