🏛️🧨🪛 Will AI break democracy or fix it? Yes.

An excerpt from Bruce Schneier and Nathan E. Sanders' new book

We’re seeking a Social Media Fellow and a Research Fellow for AI & The Future of Social Conversation

It’s time we talked about the elephant in the room.

So much of what we care about at New_ Public, including how we talk to each other online, where we get news and information, and how the internet shapes our lives and communities, is rapidly changing. Generative AI might rearrange everything.

It’s tough to talk about AI with confidence and certitude — it’s evolving quickly, it means something different to everyone, and it’s extraordinarily divisive along unusual axes. Depending on who you ask, it’s a grift, a bubble, a toolset, or a world-changing epochal shift. But without doubt, AI has arrived, and it is already making an impact.

Will we retread the same patterns of social media, and will AI be built to maximize growth, engagement, and profit at all costs? Will everything about the social internet be transformed, unrecognizably, with entirely different incentives and outcomes?

We’re excited to begin to take on this topic, and we have a great introduction in this newsletter.

Below we have an exclusive excerpt of the new book by Bruce Schneier and Nathan E. Sanders, Rewiring Democracy: How AI Will Transform Our Politics, Government, and Citizenship. Bruce, a longtime friend of New_ Public, is an internationally renowned security technologist and bestselling author. He’s a Lecturer at the Harvard Kennedy School and a board member of EFF. Nathan is a data scientist focused on making policymaking more participatory, with research on many subjects, including AI.

In Rewiring Democracy, Bruce and Nathan do some predicting and explaining, but they emphasize the agency we all have to shape AI’s development, and our power to use it to make democracy stronger and more collaborative. This chapter, on social media, journalism, and even politics, is fascinating and central to our interests. For more, please consider pre-ordering the book, out this month from MIT Press. (Bolding below added by me.)

–Josh Kramer, Head of Editorial, New_ Public

Informing The Public

Bruce Schneier and Nathan E. Sanders
An excerpt from their new book, Rewiring Democracy

Media, especially local reporting, is integral to a functioning democracy. Yet social media and industry consolidation have all but destroyed local journalism. For as many new and exciting experiments as there are in nontraditional media enterprises to replace these lost institutions, a reliably reproducible model has not yet been established.

With the help of AI, reporting will grow even more nontraditional. AI will displace human political commentary and automate local news delivery because some media owners are deciding it is the only economic way to deliver local news. AI will supercharge the fragmentation of the media sparked by the growth of the internet. Social media, blogs, streaming video, and podcasting already provide platforms — and audiences — for every interest and viewpoint, no matter how marginal it might seem. AI can drive that fragmentation to the individual level, not only personalizing content feeds but also creating new and different content to fit singular tastes.

Modern generative AI has far superior capabilities in scope and sophistication compared to earlier technologies like predictive news feeds. Instead of picking one out of a few versions of a headline, it can create a custom headline just for you, or even an entire story, in text, audio, or video, with your preferred content and tone. Instead of simply learning from the links that you click or ignore, AI can solicit and integrate more specific feedback. It can ask you whether you believed the AI report, whether it was biased against your own interests, or whether it was just plain boring, and modify the content accordingly.

Google took a big step towards these capabilities with the 2024 release of NotebookLM Audio Overviews, an AI tool that can transform any document — even yawn-inducing texts like the US Federal Register — into a conversational podcast. Around Boston, this ability was immediately put to work generating a local news podcast. This tool demonstrates one means by which AI can usefully respond to detailed feedback from users; NotebookLM will tweak its synthetic conversations if you ask it to focus on particular pages of the document or specific subjects, or to appeal to certain audiences. This provides a wholly new capability for news consumers.

There’s real potential here for AI to replace human reporters and commentators, and thereby fill a gap in local coverage that the market is increasingly failing to provide for. Even those consumers who yearn for the human perspective and voice may find AI-personalized news unavoidable, as it generates content on niche topics no human commentators are writing about and as it becomes increasingly indistinguishable from human-generated content.

Yet even when AI personalization fills a real void, it might still have negative effects. Social media has demonstrated how catering to individual predilections gives rise to “echo chambers” or “filter bubbles”: feedback loops of hyperpartisan, extremely ideological, frequently false content that entraps consumers and coarsens public discourse.

AI can personalize news in vastly different ways than social media. The level of personalization in predictive social media feeds is often tailored at the trivial level of “you like cats, so here’s a cat.” AI-generated content will sometimes have the same aggravating, obsequious quality, yet it can also be truly individualized and responsive to the input and feedback you offer about the information you seek.

Currently, those superior capabilities are being largely developed and deployed by the same Big Tech companies that developed the current generation of social algorithms. Their incentives haven’t really changed; this matters more than the capabilities. Early AI-generated news content has emulated the clickbait that has been promoted on social media for years: inane articles about “one small trick,” nonsense medical cures, celebrity listicles. It doesn’t have to work like this. AI could be used to follow your directions, instead of to hack your attention. For example, in 2023, the AI-powered news aggregator Artifact introduced a feature that allowed users to flag misleading headlines and trigger AI to rewrite them more accurately.

However, tech companies aren’t building these tools as part of a mission to improve our lives. Primarily and inevitably, their mission is monetization. Platform companies are happy to extract profit from all political factions or to cater to whichever party favors their business. People seeking objective news may trust an AI model more than any of the dwindling number of mainstream corporate alternatives. But they should be aware that tech companies distributing content from those AIs will be willing to serve, and profit from, misinformation along with verifiable facts.

Politicians will be eager adopters of these personalizing technologies. AI tools allow candidates to publish content in every format, covering any and all stories in the day’s news from their own perspective and curating the content in the way most beneficial to their campaign. A 2024 candidate for the US state of Georgia’s House of Representatives was among the first to use AI to generate blog posts, images, and even podcasts expressing his policy positions.

Distressingly, major losses have already been felt by journalists and traditional media companies. Journalism is one of the first fields where AI systems have displaced human labor, especially in reporting on sports, fashion, and business. Even though human journalists are still superior to AI and possess capabilities that AI does not, that doesn’t stop media companies from choosing the cheaper AI alternative. We don’t expect major national newspapers to assign opinion columns to an AI, but your local paper very well might — and then not tell you. News organizations can also use AI responsibly, with disclosure, and to useful effect. In 2024, the Washington Post began using AI to augment reader comments by prompting and summarizing their reactions to stories. Journalists can also use AI to their own advantage — more on that in the next chapter.

Bruce Schneier (left) and Nathan E. Sanders

We can even imagine a future where human influencers and opinion leaders are less relevant, and are supplanted by a personalized AI news feed. More likely, the most successful pundits will become even more influential, while the average ones will be replaced by AIs. Such a development would represent a continuation of the effects of the internet on opinion journalism. A surviving few national and international outlets have elevated platforms for their star columnists, while local beats are increasingly neglected.

How will political behavior change as the media becomes more automated and personalized? You can expect it to become more transactional. The politician’s goal will be to work their way into the individualized content feeds of a million constituents rather than to make it onto the monolithic nightly news. The winning strategy will entail addressing the specific interests of everyone, everywhere. An elegant speech on a general policy topic has a low probability of keenly interesting anyone, so it may not land in many feeds. But a narrowly targeted promise could interest a specific constituency. If you can make enough promises or pander to enough constituencies — a tax cut here, a new park there, shaving a regulation or three — you can maximize your probability of landing in almost everyone’s feed. A steady stream of such content can help a candidate win the attention contest that is politics.

The strategy of deluging media consumers with content from every angle to maximize attention is not new, but AI will change its implementation and effects. The modern phrase “flooding the zone” emerged when political controversies from the early 2000s began to be shaped by a proliferation of bloggers, each with small audiences. Mass media narratives emerged from their fragmented and chaotic writings. Government institutions like the US military employed this digital flood strategy to shape public perception around issues like the Iraq War, funding and coordinating networks of bloggers and writers to propagate their agenda. Political campaigns and elected leaders increasingly do this and, often, the flood consists of half-truths, untruths, misinformation, and demagoguery. As we discussed earlier, AI empowers campaigns with superhuman coordination. An AI-enhanced campaign can engage in individual outreach to cultivate every journalist, every social media influencer, and every voter as a potential contributor to the deluge.

The capabilities of AI to personalize the news are both real and compelling, so we should direct our attention to the incentives and means of distribution for that content. We should expect politicians and parties to increasingly leverage AI to flood the zone with content that advances their narrative, and expect many voters to increasingly consume that content. Democratic, open societies may not outlaw this kind of speech, but they can choose to regulate the channels through which it is distributed, and to induce them to prioritize values beyond short-term revenue and click-through rates. Next, we will turn to how we can use the same AI tools to inject more substantive, fact-based reporting and deliberation into the political bloodstream.

Excerpted from Rewiring Democracy: How AI Will Transform Our Politics, Government, and Citizenship by Bruce Schneier and Nathan E. Sanders. Reprinted with permission from The MIT Press. Copyright 2025.

Thanks Bruce and Nathan! If you are a research, writer, and thought partner intrigued by this topic and looking for work, please check out this new role directly working with our Co-Director Eli Pariser.

Exploring the origins of democracy IRL in Greece right now,

–Josh