Most genAI coverage looks at nat'l elections & deepfakes. This week we launched a new project, the AI Political Archive, which seeks to go deeper by tracking the full range of generative AI uses — not just deepfakes — across state, local, and national races in 2024.
In a big win for transparency, yesterday's #SCOTUS ruling on FL/TX social media laws signaled that well-drafted disclosure laws should withstand constitutional scrutiny. Read our @lawfare.bsky.social article explaining why these cases are important for researchers.
Facebook is in its 'fuck around find out' era. A colossal, decaying platform emboldened by Musk telling advertisers to go fuck themselves. Experts it used to consult with on hard issues say they haven't heard from the company in years
Online misinformation seems worse than ever, at least based on vibes. But it's now harder than ever for outside researchers to study misinformation because platforms are have reduced transparency, writes @vox.com's A.W. Ohlheiser.
While we haven't seen a deluge of AI-generated deepfakes in 2024, deceptively edited cheap fakes are successfully pushing misleading messages to Americans. See these two stories from @caseynewton.bsky.social and @davidingram.bsky.social for more.
New survey data from Pew highlights how, for large swaths of TikTok and other social media users, politics just isn't the priority that lots of campaigns and policymakers want or imagine it to be: www.cnn.com/2024/06/12/t...
Some tremendous insights in this @knightfdn.bsky.social report about how research has helped shape public policy and discourse on digital media and democracy. It also highlights several challenges -- and opportunities -- for the field going forward.
Earlier this year we outlined much of this research for Brookings, explaining that most people don't see misinfo, exposure is heavily concentrated among a small fringe, and it doesn't influence our political behaviors and beliefs.
Conventional wisdom says online misinfo is rampant, algorithms are at fault, & this makes us more polarized. In important new article in @natureportfolio.bsky.social debunks these myths, explaining what the research actually shows.
AI is already playing a role in U.S. politics – and there’s concern about how it may impact elections in 2024 and beyond.
Through interviews with experts in digital politics, we look at how & why AI is being used and what it could mean for the future: bit.ly/CMEElectionsAI
In this week's @natureportfolio.bsky.social, a collection of articles probe the problem of online misinformation and try to assess the risks posed to society. Highly recommend this editorial, which does an excellent job summarizing insights from each piece.
Some have criticized this meme as "lazy activism" or "slacktivism." Our research from a decade ago finds that activating these kinds of peripheral users is an important way to expand the audience for protest messages.
The debate over children's online safety legislation takes for granted that social media harms young people and the proposed reforms would make a difference. But a close look at the research finds this isn't necessarily the case.
🚨NEW REPORT🚨: Child Online Safety Legislation: A Primer - We break down intl, state, & federal legislation like KOSA - Will they improve youth mental health? VERY unclear - They WILL require widespread age-gating & harm marginalized ppl citap.pubpub.org/pub/cosl/
New: TikTok officials offered the U.S. government everything they thought it wanted: board control, code-source review, even a kill switch. The U.S. chose a blunter option.
"A complete absence of faith in [its] ability to regulate tech."
Social media research tends to focus on nat'l & int'l political discussions on large platforms (FB, X, YT). These findings contribute to our understanding of a widely used local platform by showing what the real-life neighborhoods of Nextdoor’s ‘digital town squares’ look like 9/
3️⃣ Using a sample of 116k posts & 164k comments, collected from 30 Nextdoor neighborhoods, we find that posts seeking and offering services are by far the most popular — but posts reporting potentially suspicious people or activities received the highest engagement. 8/
Why are police more likely to be present in neighborhoods w/more non-white residents? We don't know. They could be there to engage w/the community, facilitate surveillance by users, or some combo of the two. Analyzing police behavior on ND is a promising line of future work. 7/
2️⃣ Government agencies, and law enforcement agencies in particular, are more likely to be present on Nextdoor in communities that have a higher proportion of non-white residents, greater levels of income inequality, a higher average home value, and higher incomes.
6/
Using US Census data, other public info, and posts from a sample of Nextdoor neighborhoods, the research yields three main findings:
1️⃣ Nextdoor neighborhoods are more likely to be located in communities that are less dense, whiter, wealthier, older, and more educated.
5/
Our paper seeks to better understand these dynamics by generating a demographic portrait of communities in which Nextdoor neighborhoods exist, the presence of public agencies in those communities, & what topics are most often discussed. 4/
In addition, the presence of law enforcement agencies on Nextdoor, paired with posts categorized as “crime and safety,” has drawn criticism for contributing to community policing and surveillance, often with racialized impacts. 3/
Nextdoor is present in ~220k US neighborhoods, but we know little about the make-up of those neighborhoods or what users discuss. This limits our understanding of the digital media ecosystem as a whole, and its relationship with local communities and civic life in particular. 2/
Social media research & reporting often focuses on large platforms. Much less is known about hyperlocal sites like Nextdoor, used by 1/3 US households. In @journalqd.bsky.social, we help fill this gap w/new data exploring communities & conversations on the site.
New research from several fact checking organizations shows that most image-based disinformation is now AI-generated—but the way researchers collected their data suggests that the problem is even worse than they claim. www.404media.co/google-says-...
How can search engines feed people reliable information? How can they innovate and adapt to competitive pressures, such as AI, while remaining a stable part of our digital infrastructure? Our recent event covered these questions & more. Read our recap: csmapnyu.org/impact/news/...
Even seeing one link in search results to a low-quality article containing misinfo can make someone believe it. An AI summary of these results might exclude that info, taking the onus off the consumer to determine what's high-quality and what's not.
Latinos in the U.S. who use social media in Spanish are more likely to believe disinformation when they come across it.
Laura Zommer writes about our new research on news consumption habits of Latino communities for
@IJNet@icfjorg.bsky.social
As millions of Indians vote, there’s one platform most of them will turn to for information: WhatsApp. And no one controls the political narrative on WhatsApp like the BJP, the ruling party helmed by Prime Minister Modi.
Thank you to Zoe Darme of Google, @miasato.bsky.social, and @ftripodi.bsky.social for joining us yesterday to discuss how search engines influence the information landscape — and what role AI could play going forward. Here's the video in case you missed it.
We address exactly this point -- about the risks of overstating the impact of foreign influence campaigns on creating a sense that U.S. electoral system is more fragile than it actually is -- in the discussion section of that paper.
See as well our work about the reach of Russian foreign influence campaigns on Twitter in 2016, which was heavily concentrated, a tiny fraction of the people's exposure to news on Twitter, and not related to changes in opinions or vote preferences.
3/4
We need to prepare voters for possible Russian disinformation tactics in the 2024 election. But overstating the importance of videos that have "fail[ed] to get traction" risks exaggerating the efficacy and impact of these campaigns.
1/4
Big victory for transparency here in Musk's lawsuit over data scraping. The judge says allowing social platforms total control over their data risks creating "information monopolies that would disserve the public interest."
Join us on Wednesday at 12pm ET for this conversation about the future of search engines in the age of AI, featuring experts from academia, journalism, and industry.
Details 👇
If you're at #VivaTech later this month, check out @jatucker.bsky.social's session on 5/23: "Polarization or Conversation: Can We Still Fix Social Networks?"
According to our research, Russia's 2016 influence campaign on Twitter didn’t change attitudes or voting behavior. But its largest effects may have been convincing Americans the campaign was successful.
For Wired, @davidgilbert.bsky.social reports on China’s state-sponsored disinformation campaign, which have been incredibly ineffective despite running for seven years.