Thursday, June 27, 2024

Is ChatGPT the key to stopping deepfakes? Study asks LLMs to spot AI-generated images

Large language models perform below state-of-the-art detectors, but their ability to plainly explain their analysis holds promise

UNIVERSITY AT BUFFALO

ChatGPT's analysis of deepfakes — IMAGE:
AN EXAMPLE OF CHATGPT'S ANALYSIS OF DEEPFAKE IMAGES. THE LARGE LANGUAGE MODEL WAS LESS ACCURATE THAN STATE-OF-THE-ART DEEPFAKE DETECTORS, BUT IMPRESSED RESEARCHERS WITH ITS ABILITY TO EXPLAIN ITS ANALYSIS IN PLAIN LANGUAGE.
view more
CREDIT: UNIVERSITY AT BUFFALO

BUFFALO, N.Y. — When most people think of artificial intelligence, they’re probably thinking of — and worrying about — ChatGPT and deepfakes. AI-generated text and images dominate our social media feeds and the other websites we visit, sometimes without us knowing it, and are often used to spread unreliable and misleading information.

But what if text-generating models like ChatGPT could actually spot deepfake images?

A University at Buffalo-led research team has applied large language models (LLMs), including OpenAI’s ChatGPT and Google’s Gemini, toward spotting deepfakes of human faces. Their study, presented last week at the IEEE/CVF Conference on Computer Vision & Pattern Recognition, found that LLMs' performance lagged behind that of state-of-the-art deepfake detection algorithms, but their natural language processing may actually make them the more practical detection tool in the future.

“What sets LLMs apart from existing detection methods is the ability to explain their findings in a way that’s comprehensible to humans, like identifying an incorrect shadow or a mismatched pair of earrings,” says the study’s lead author, Siwei Lyu, PhD, SUNY Empire Innovation Professor in the Department of Computer Science and Engineering, within the UB School of Engineering and Applied Sciences. “LLMs were not designed or trained for deepfake detection, but their semantic knowledge makes them well suited for it, so we expect to see more efforts toward this application.”

Collaborators on the study include the University at Albany and the Chinese University of Hong Kong, Shenzhen. The work was supported by the National Science Foundation.

How language models understand images

Trained on much of the available text on the internet — amounting to some 300 billion words — ChatGPT finds statistical patterns and relationships between words in order to generate responses.

The latest versions of ChatGPT and other LLMs can also analyze images. These multimodal LLMs use large databases of captioned photos to find the relationships between words and images.

“Humans do this as well. Whether it be a stop sign or a viral meme, we constantly assign a semantic description to images,” says the study’s first author, Shan Jai, assistant lab director in the UB Media Forensic Lab. “In this way, images become their own language.”

The Media Forensics Lab team decided to test if GPT-4 with vision (GPT-4V) and Gemini 1.0 could tell the difference between real faces and faces generated by AI. They gave it thousands of images of both real and deepfake faces and asked it to identify any potential signs of manipulation, or synthetic artifacts.

ChatGPT advantages

ChatGPT was accurate 79.5% of the time on detecting synthetic artifacts in images generated by latent diffusion, and 77.2% of the time on StyleGAN-generated images.

“This is comparable to earlier deepfake detection methods, so with proper prompt guidance, ChatGPT can do a fairly decent job at detecting AI-generated images,” says Lyu, who is also co-director of UB’s Center for Information Integrity.

More crucially, ChatGPT could explain its decision making in plain language. When provided an AI-generated photo of a man with glasses, the model correctly pointed out that “the hair on the left side of the image slightly blurs” and “the transition between the person and the background is a bit abrupt and lacks depth.”

“Existing deepfake detection models will tell us the probability of an image being real or fake, but they will very rarely tell us why they came to this conclusion. And even if we look into the model’s underlying mechanisms, there will be features that we simply can’t understand,” Lyu says. “Meanwhile, everything ChatGPT outputs is understandable to humans.”

That’s because ChatGPT bases its analysis on semantic knowledge alone. Whereas traditional deepfake detection algorithms distinguish real from fake by training on large datasets of images labeled real or fake, LLMs’ natural language abilities give them something of a common sense understanding of reality — at least when they’re not hallucinating — including the typical symmetry of human faces and the look of real photographs.

“Once the vision component of ChatGPT understands an image as a human face, the language component can make the inference that a face will typically have two eyes, and so on,” Lyu says. “The language component provides a deeper connection between visual and verbal concepts.”

ChatGPT’s semantic knowledge and natural language processing make it a more user-friendly deepfake tool for both users and developers, the study concluded.

“Typically, we take insights about detecting deepfakes and convert them into programming language. Now, all this knowledge is present within a single model and we need only use natural language to bring out that knowledge,” Lyu says.

ChatGPT drawbacks

ChatGPT’s performance was well below the latest deepfake detection algorithms, which have accuracy rates in the mid- to high-90s.

This was partly because LLMs can’t catch signal-level statistical differences that are invisible to the human eye but often used by detection algorithms to spot AI-generated images.

“ChatGPT focused only on semantic-level abnormalities,” Lyu says. “In this way, the semantic intuitiveness of the ChatGPT’s results may actually be a double-edged sword for deepfake detection.”

And other LLMs may not be as effective at explaining their analysis. Despite performing comparatively to ChatGPT at guessing the presence of synthetic artifacts, Gemini’s supporting evidence was often nonsensical, like pointing out nonexistent moles.

Another drawback is that LLMs often refused to analyze images. When asked directly whether a photo was generated by AI, ChatGPT typically replied with, “Sorry, I can’t assist with that request.”

“The model is programmed not to answer when it doesn’t reach a certain confidence level,” Lyu says. “We know that ChatGPT has information relevant to deepfake detection, but, again, a human operator is needed to excite that part of its knowledge base. Prompt engineering is effective, but not very efficient, so the next step is going one level down and actually fine tuning LLMs for this task specifically.”

JOURNAL

Computer

DOI

10.48550/arXiv.2403.14077

METHOD OF RESEARCH

Experimental study

SUBJECT OF RESEARCH

Not applicable

ARTICLE TITLE

Can ChatGPT Detect DeepFakes? A Study of Using Multimodal Large Language Models for Media Forensics

NIH funds critical center in Detroit to lead efforts to investigate and mitigate health impacts of community-voiced chemical and non-chemical stressors

WAYNE STATE UNIVERSITY - OFFICE OF THE VICE PRESIDENT FOR RESEARCH

DETROIT — Wayne State University received a four-year, $5.2 million P30 environmental health sciences core center (EHSCC) grant from the National Institute of Environmental Health Sciences (NIEHS) of the National Institutes of Health (NIH) in support of the “Center for Urban Responses to Environmental Stressors (CURES).”

This grant will allow the interdisciplinary CURES team of researchers, educators and community partners to continue its ongoing quest to understand the basis for urban environmental health disparities and the human health impact of environmental exposure to complex chemical and non-chemical stressors in Detroit's urban landscape. CURES is one of 26 P30 EHSCCs funded by NIEHS. The CURES team is a collaborative hub focused on community-engaged research and environmental health equity in Detroit and throughout the region.

“We are excited to receive this prestigious and highly competitive EHSCC award from NIEHS,” said Melissa Runge-Morris, M.D., director of CURES and of the Institute of Environmental Health Sciences at Wayne State University. “First launched in 2014, this new funding for CURES is a tribute to the great achievements that our researchers and community collaborators have made to date, and the important impact they have already made on Michigan and will continue to make moving forward.”

The integrative research made possible by this grant will study the hypothesis that diseases that are disproportionately prevalent in urban communities are a consequence of interactions between an individual’s susceptibility, socioeconomic framework and environmental exposures — particularly during vulnerable life windows that tip cellular networks from health toward disease. For the CURES team, Detroit’s history as an industrial focal point combined with its predominantly minority population of African Americans, as well as those originating from the Middle East and Eastern Europe, render it ideal to investigate all aspects of environmental health at the intersection of where pollutants, political, economic, cultural, health care and behavioral factors interact.

CURES making important impact

CURES researchers believe that to make a difference in urban environmental health outcomes, a team-science approach is required. Already, the dynamic collaborations incited by the interdisciplinary CURES team have achieved immeasurable advances in environmental health research that will pave the way to environmental health resilience in at-risk communities.

“Judy Westrick, Ph.D., director of Wayne State’s Lumigen Instrument Center in the Department of Chemistry and co-leader of CURES’ Exposure Signatures Facility Core, coordinated and assisted rapid response efforts to the East Palestine, Ohio, train derailment,” said Runge-Morris. “Her important work helped to lead community sampling, environmental monitoring, and chemical analysis of toxic chemicals released into the local environment. Because of her efforts, she and our research team have positioned CURES at the leading edge nationally as an essential resource for rapid-fire, community-engaged and technically sound responses to combat environmental disasters in the future.”

The CURES team has also become an important collaborative partner to community stakeholders as they advocate for a healthier Detroit. One such project led to a community grant from the American Rescue Plan Act to assist with community violence intervention. To address urban disparities in internet access, a major barrier to health care and workforce equity, CURES community and donor partnerships aided thousands of community members by providing 1,700 computer devices and internet access to older Detroit residents to assist with telehealth technology. CURES also developed a neighborhood deprivation index that is used to investigate how disadvantaged socioeconomics and neighborhood adversity predict Type 1 diabetes-related health outcomes in young Black adolescents.

CURES provided science principles, scientific foundation, guidance, pilot funding and analytical resources that has helped develop and germinate several other significant funded programs at Wayne State University, such as the Center for Leadership in Environmental Awareness and Research (CLEAR), an $11.3 million P42 Superfund Research Program center grant funded by NIEHS in 2022.

Additionally, CURES aligns with and supports the $18.15 million NIMHD P50 Center, ACHIEVE GREATER, led by Wayne State University’s Associate Vice President for Research Phillip Levy, M.D., which takes aim at cardiovascular health disparities in Detroit’s vulnerable urban population.

CURES core leader, Douglas Ruden, Ph.D., of Wayne State University, and CURES epidemiologists at Henry Ford Health Jennifer Straughen, Ph.D., and Christine Cole Johnson, Ph.D., play leadership roles in two comprehensive and large-scale Environmental Influences on Child Health Outcomes (ECHO) Program grants focused on environmental health and child development.

Examples of other high-impact projects stemming from the collaborative work of CURES researchers include NIEHS R01-funded studies to understand the role of gene-by-heavy-metal interactions in development of nonalcoholic fatty liver disease (PIs Wanqing Liu, Ph.D. and Andrea Cassidy-Bushrow, Ph.D.) and the impact of benzene, a major component of air pollution, on insulin resistance leading to Type 2 diabetes (PI Marianna Sadagurski, Ph.D.). CURES also provided pilot funds to a study that is investigating the impact of paternal exposures to per- and poly-fluorinated alkyl substances on the metabolic health of their offspring (PIs Michael Petriello, Ph.D. and Richard Pilsner, Ph.D.).

“The Center for Urban Responses to Environmental Stressors has made a major impact on the Detroit community through its important projects that are working to address critical health concerns of residents of urban cities like Detroit,” said Ezemenari Obasi, Ph.D., vice president for research at Wayne State University. “I am pleased that NIEHS is funding this project to allow them to continue studies that focus on the betterment of urban communities.”

The award number for this National Institutes of Environmental Health Sciences grant is P30ES036084.

# # #

About Wayne State University

Wayne State University is one of the nation’s pre-eminent public research universities in an urban setting. Through its multidisciplinary approach to research and education, and its ongoing collaboration with government, industry and other institutions, the university seeks to enhance economic growth and improve the quality of life in the city of Detroit, state of Michigan and throughout the world. For more information about research at Wayne State University, visit research.wayne.edu.

Wayne State University’s research efforts are dedicated to a prosperity agenda that betters the lives of our students, supports our faculty in pushing the boundaries of knowledge and innovation further, and strengthens the bonds that interconnect Wayne State and our community. To learn more about Wayne State University’s prosperity agenda, visit president.wayne.edu/prosperity-agenda.

Taxing shared micromobility: How cities are responding to emerging modes, and what's next

PORTLAND STATE UNIVERSITY

Shared micromobility (including shared electric scooters and bikes provided by private companies) is one of the newest transportation options that has come to cities in the last several decades. A new report explores the different ways cities charge shared micromobility companies to operate, and how these funds are used.

In the newly released report, John MacArthur of Portland State University, Kevin Fang of Sonoma State University and Calvin Thigpen of Lime examine data from 120 cities in 16 countries around the world. They also conducted a survey of cities’ shared micromobility program managers, with responses representing 33 jurisdictions in North America.

Download the report: "Taxing shared micromobility: assessing the global landscape of fees and taxes and their implications for cities, riders, and operators (PDF)"

"This study builds our understanding of a topic that is near and dear to the hearts of cities, riders, and micromobility operators: how to run a system that is affordable for riders while also remaining financially sustainable for micromobility operators. In the last 6 months alone, the industry has seen substantial upheaval through mergers, bankruptcies, and closures. So as cities revisit their program regulations, we hope they take into consideration that the industry has matured substantially since fees were initially established - with safer vehicles, better operations, and closer city collaboration - as well as the role shared micromobility can play in achieving sustainability and equity goals," Thigpen said

Digging into how each city made the decision of what to charge, the researchers find that taxes and fees vary dramatically from city to city and may not always reflect the city's stated policy goals.

"Though cities are using fees and taxes to mitigate the cost of program administration, which is very understandable given local budget constraints, these additional costs to riders can be at odds with a city’s broader goals for supporting sustainable and equitable transportation," MacArthur said.

The findings also reveal a trend of cities charging less for shared bicycles than for shared scooters. The notable exception to this pattern is Denver (which has both e-bikes and scooters), where the city does not differentiate between vehicle types and charges no program fees.

"One concern sometimes raised about shared micromobility are user fares. To the degree fees increase the cost of business and get passed along to riders, cities have a say in fares with their fee levels," Fang said.

Four Key Findings

1) Fees vary dramatically between cities. Some cities do not impose program fees at all, in line with municipal transportation goals. In cities that do assess program fees, the common types are per-trip, per-vehicle, flat annual, and flat one-time fees. There are large differences in the fee amounts that cities charge – for example, the highest per-vehicle fee is over four hundred times higher than the lowest.

"Fee levels were consistently inconsistent. In some cases, cities had zero permit fees. On the other end they could be a dollar or two for every trip," Fang said.

2) Shared micromobility is taxed twice—via sales tax and program fees—and these revenues can be substantial. On average, cities charging an annual fee received over a third of a million USD each year. If sales taxes/value added taxes (VAT) are included with fees, the average shared micromobility trip generates a fee + tax revenue of $0.70 USD per mile or $0.89 USD per trip. This means that globally, shared micromobility programs bring in an average rate of 16.4% of revenue from user fares in taxes and fees.

3) Shared micromobility taxes and fees are higher than most other modes of transportation, especially driving and ridehail. The research team found that fees and taxes on shared micromobility are significantly higher compared to other travel modes, being 23 times higher per mile than personal cars and over 5 times higher than ridehail trips.

"On average, fees and taxes on shared micromobility were quite a bit higher than charges on personal driving or ride-hail trips. This seems counter to many cities' goals of promoting alternative modes of travel," Fang said.

4) When deciding on fees, cities are especially concerned with covering administrative costs as well as influencing operator behaviors. The primary use of fee revenue is to cover program administration costs, rated as the top consideration by 77% of respondents (see chart below). Ensuring financial feasibility for scooter companies or lowering rider costs were less prioritized, even though both would benefit the shared micromobility system.

While cities’ concerns over budget are understandable, this consideration can be at odds with cities’ broader goals for supporting alternative transportation.

"After the boom of shared micromobility in 2017, cities looked at fees as a way to react to this new mode. We see that cities are still setting fees to cover program administration costs, but also as a way to influence operator behavior of how they operate their systems in the public rights-of-way," MacArthur said.

Download the report (PDF) for details about the research, including the survey that was distributed to shared micromobility program staff and the methods used to calculate taxes and fees for other modes of transportation. The report's appendix includes a complete summary of program fees in each jurisdiction as well as shared micromobility program fee revenues in 2022.

Who Can Use This Research?

Fees and taxes are relevant to all three of the major "stakeholders" in the shared micromobility field: cities, private micromobility companies, and travelers.

There have been numerous studies on cities’ shared micromobility policies around parking, ridership, safety, equitable distribution of vehicles, and sustainability. By contrast, there has been little research on the taxes and fees levied on shared micromobility systems and how they work to advance or deter municipal goals for shared micromobility.

The researchers provide case studies of cities taking different—and evolving—approaches to illustrate how different cities weigh tradeoffs. The information provided in this report can help inform cities who are working with shared micromobility companies to align program fee structures with their goals around climate, equity, congestion and more.

In a section offering rationales for lowering (or not charging) fees, the report notes that the shared micromobility landscape has changed since e-scooters first swept the world in 2017 and 2018. The shared micromobility industry no longer deploys at-will in city streets, but rather works through formal procurement processes to serve cities.

"Considering the newness of shared micromobility, it is not that surprising that approaches to fees have varied so widely initially. Today, though, cities are mostly on the same page with what they want with micromobility operations, so greater alignment on fees probably makes sense," Fang said.

Shared scooters and bikes are no longer just pilots; most cities now have multi-year permits with established operators. Both cities and companies are aware of the risks of poorly managed systems and have developed technologies and programs to address equity and operational challenges. This research offers a comprehensive look at how cities around the world are approaching the question of what to charge, and offers strategies to ensure that a city's fee structure supports transportation policy goals.

Portland State University's Transportation Research and Education Center (TREC) is home to the U.S. DOT funded National Institute for Transportation and Communities (NITC), the Initiative for Bicycle and Pedestrian Innovation (IBPI), PORTAL, BikePed Portal and other transportation grants and programs. We produce impactful research and tools for transportation decision makers, expand the diversity and capacity of the workforce, and engage students and professionals through education and participation in research. To get updates about what's going on at TREC, sign up for our monthly newsletter or follow us at the links below.

June research news from the Ecological Society of America

ECOLOGICAL SOCIETY OF AMERICA

Transplanted pine seedlings in Interior Alaska — IMAGE:
A NEW STUDY IN *ECOLOGICAL APPLICATIONS* TESTS WHETHER TRANSPLANTED PINE SEEDLINGS (LIKE THOSE SHOWN HERE) CAN SURVIVE IN PINE-FREE INTERIOR ALASKA.
view more
CREDIT: JILL JOHNSTONE

The Ecological Society of America (ESA) presents a roundup of four research articles recently published across its six esteemed journals. Widely recognized for fostering innovation and advancing ecological knowledge, ESA’s journals consistently feature illuminating and impactful studies. This compilation of papers explores the potential for pines to establish in pine-free interior Alaska, internet sleuthing to assess birds’ extinction risk and more, showcasing the Society’s commitment to promoting cutting-edge research that furthers our understanding of the natural world.

From Ecological Applications:

Lodgepole pine staking a claim to Alaska’s boreal region
Author contact: Xanthe J. Walker (xanthe.walker@nau.edu)

Shifts in species ranges driven by ongoing climate change are perhaps nowhere more evident than in the rapidly warming northern latitudes of the world. In this study, researchers used a combination of field experiments and model simulations to assess the susceptibility of Alaska’s boreal forests, where pine trees are currently absent, to colonization by lodgepole pine from forests in the neighboring Yukon Territory. They found that lodgepole could readily establish across their field sites as environmental conditions become increasingly suitable with little resistance from natural checks like browsing by moose and hare and competition from native trees. Based on their results, the authors warn that lodgepole pine could become widely distributed throughout the boreal regions of Alaska as natural constraints on their distribution wane under future climate conditions, resulting in a fundamental shift in ecosystem structure with considerable ecological, socioeconomic and cultural ramifications.

Read the article: Factors limiting the potential range expansion of lodgepole pine in Interior Alaska

A recipe for more deer in Washington state
Author contact: Taylor R. Ganz (taylorganz@gmail.com)

White-tailed deer in northeastern Washington State are vulnerable to apex predators and vehicle collisions on roads, but they also benefit from areas opened by timber extraction and agriculture according to the authors of this study. Analysis of radio-collar data from 280 deer revealed that predators—particularly cougars—had a marked effect on deer numbers. However, substantial increases in edible vegetation in recently logged areas, farmland and ranches boosted deer abundance and played an even larger role in driving deer population growth when compared to predation. Consequently, the authors propose that strategies promoting greater availability of forage would be the most effective approach for increasing white-tailed deer in the state. Moreover, a substantial portion of the tracked deer were killed by cars and trucks on the region’s roads, leading the authors to recommend actions to reduce deer-vehicle collisions—for both humans’ and deer’s sakes.

Read the article: White-tailed deer population dynamics in a multipredator landscape shaped by humans

From Ecosphere:

Honey bee pathogens are a bumble bee buzzkill
Author contact: Heather M. Hines (hmh19@psu.edu)

A new study suggests that honey bees may be serving as a winter reservoir of viral pathogens that go on to infect common eastern bumble bees in spring. Although parasite prevalence is relatively high in both bees in summer, loads decline to negligible levels in overwintering bumble bees. In contrast, honey bees appear to be less adept at ridding themselves of these nuisances and they maintain higher viral loads during the wintertime. Come spring, the larger parasite load carried by honey bees through the winter can then be passed to bumble bees. The findings have important ecological implications but may also aid in the development of more effective management strategies aimed at improving the survival of these indispensable pollinators.

Read the article: Comparison of seasonal viral prevalence supports honey bees as potential spring pathogen reservoirs for bumble bees

From Frontiers in Ecology and the Environment:

Looking online for long-lost birds
Author contact: Cameron L. Rutt (cameronrutt@gmail.com)

Catalogs of data collected by hobbyists and amateur scientists (both skilled and novice) on the world’s birds are now so extensive that they can serve as a reliable index of extinction risk, propose the authors of this new study. Analysis of 42 million photographic, audio and video records uploaded to popular online platforms like eBird and iNaturalist revealed that 144 (1.2%) of the nearly 12,000 known avian species across the globe meet the criteria of “lost,” which the authors define as the absence of solid documentation for more than a decade. Moreover, they suggest that the length of time since the last record of a species, combined with the collective effort spent searching for it, reflect a bird’s odds of being imperiled. Based on this measure, the researchers estimated that nearly 62% of the “lost” species are likely gone for good. The study highlights the value of these so-called citizen science or community science platforms, for biodiversity writ large.

Read the article: Global gaps in citizen-science data reveal the world’s “lost” birds

###

The Ecological Society of America, founded in 1915, is the world’s largest community of professional ecologists and a trusted source of ecological knowledge, committed to advancing the understanding of life on Earth. The 8,000 member Society publishes six journals and a membership bulletin and broadly shares ecological information through policy, media outreach and education initiatives. The Society’s Annual Meeting attracts 4,000 attendees and features the most recent advances in ecological science. Visit the ESA website at https://www.esa.org

Follow ESA on social media:
Twitter/X – @esa_org
Instagram – @ecologicalsociety
Facebook – @esa.org

JINGOISM

US states shape foreign policy amid national China unease, research shows

Peer-Reviewed Publication

UNIVERSITY OF NOTRE DAME

Kyle Jaros — IMAGE:
KYLE JAROS, ASSOCIATE PROFESSOR OF GLOBAL AFFAIRS AT THE UNIVERSITY OF NOTRE DAME, FIELDS QUESTIONS FROM STATE- AND LOCAL-LEVEL OFFICIALS DURING THE U.S.-CHINA SUBNATIONAL SYMPOSIUM AT THE UNIVERSITY OF MICHIGAN’S LIEBERTHAL-ROGEL CENTER FOR CHINESE STUDIES IN APRIL 2024. JAROS SHARED HIS EXPERTISE AS PART OF HIS ROLE AS A FELLOW OF THE NATIONAL COMMITTEE ON U.S.-CHINA RELATIONS’ PUBLIC INTELLECTUALS PROGRAM.
view more
CREDIT: UNIVERSITY OF NOTRE DAME

State-level officials such as governors, state legislators and attorneys general are shaping U.S.-China relations as the two countries navigate a strained geopolitical relationship, according to new research by political scientist Kyle Jaros.

“The state level has independent importance in the U.S.-China relationship — it’s not just a reflection of what’s happening at the national level,” said Jaros, associate professor of global affairs in the Keough School of Global Affairs at the University of Notre Dame. “The actions taken by state and local officials — and their Chinese counterparts — not only affect their own communities, but also play a key role in shaping the overall U.S.-China relationship.”

While the U.S. Constitution clearly states that foreign policy is the responsibility of the federal government, it also leaves space for cities and states to have international relationships and even to enter into certain kinds of agreements, Jaros said.

Known as subnational diplomacy or paradiplomacy, state-level relations with China are the focus of a new study by Jaros and Sara Newland of Smith College, published in Publius: The Journal of Federalism, a publication of Oxford University Press.

Jaros and Newland, who were recently awarded a grant from the Luce Foundation to fund this work, found that states vary widely in how they engage with China and range from confrontation to cooperation or a combination of the two.

“Broadly speaking, our data analysis shows some states that are clearly pro-engagement,” said Jaros, who currently serves as a visiting senior fellow for U.S.-China subnational relations for the Truman Center in Washington, D.C. “They feel that it is still appropriate to have at least some forms of business cooperation with China and may pursue climate cooperation, tourism cooperation or educational partnerships.”

To assess states’ engagement with China, Jaros and Newland tracked changing patterns of state-level U.S.-China relations using an original dataset on cooperative and confrontational policies across all 50 U.S. states.

Of the 50 states, California serves as a leading example of a state that has continued to adopt policies that foster people-to-people contact with China, Jaros said. For example, Gov. Gavin Newsom made a high-profile trip to China in October 2023 to advance climate partnerships and economic changes, laying the groundwork for Chinese leader Xi Jinping’s visit to San Francisco the following month. Florida and Texas lie at the other end of the spectrum, Jaros said, adopting confrontational policies that curb contact between state institutions (such as government or universities) and China. Florida, for example, now restricts Chinese citizens’ and businesses’ abilities to purchase land or real estate in the state, and Gov. Ron DeSantis has publicly articulated his concern that the Communist party is infiltrating Florida’s institutions.

In Indiana, the picture is more complicated, according to Jaros and Newland’s in-depth case study of the state, which was included in the research.

“Until 2022, Indiana engaged fairly regularly with official Chinese counterparts, even as concerns and criticisms from some quarters increased,” Jaros said. “While some forms of low-key economic and educational cooperation are continuing today, many areas that once seemed appropriate or safe are now seen by Indiana officials as off-limits or dangerous.”

For years, Indiana has pursued state-level cooperation with China, as the state is home to several major corporations, including leading pharmaceutical and engineering companies, that see the Chinese market as crucial to their overall strategy. In addition, numerous small businesses see China as a crucial part of their supply chain, relying on it as a sizable export market.

And yet, last year, Indiana state legislators voted to divest the state’s pension fund from China, and the state has banned new sister city relationships with China. In 2021, state Attorney General Todd Rokita launched an investigation into Valparaiso University’s Confucius Institute, alleging that it functioned as a propaganda arm for the Chinese Communist Party.

Jaros has met with the State Department’s Subnational Diplomacy Unit about his work, and he and Newland are studying how the federal government can help U.S. states and cities coordinate knowledge-sharing and also how it can provide useful information to lower levels of government.

In his capacity as a Truman fellow, Jaros is also expanding the research to include city-level diplomacy. This summer, he and Newland, who is also a Truman fellow, will meet with city officials, chamber of commerce members and other local groups in Los Angeles; Hartford, Connecticut; Des Moines, Iowa; and Jacksonville, Florida.

“We will solicit views from the local level both for their own sake and also to bring some of it back to share with policymakers in Washington, to help them have a better awareness of how what they are doing affects local communities,” Jaros said.

A fellow of the National Committee on U.S.-China Relations’ Public Intellectuals Program, Jaros regularly briefs state and local officials on the U.S.-China relationship. In April, he delivered a presentation at the committee’s subnational symposium, held in coordination with the University of Michigan.

Jaros said the next stage of the research includes examining the consequences of what is happening at the state level, including its policy implications.

“We see policy impacts at several different levels: city, state and federal,” he said. “This work has implications for how cities and states think about what kinds of interactions with China are appropriate right now and what kinds of caution are needed.”

Contact: Tracy DeStazio, associate director of media relations, 574-631-9958 or tdestazi@nd.edu