This summer, OpenAI announced escaped a training environment and hacked into the AI company Hugging Face. Anthropic quickly followed with news that its AI agents also .��

Last week, an outgoing Anthropic employee took to X, posting that the “.” Such talk has for years, though many AI experts have argued that these Terminator-esque claims are distractions from the real risks posed by current AI systems. Nevertheless, that viral X thread is .����

To help make sense of all this, UW News talked to five AI researchers from the ������˾��:��

• , associate professor in the Information School;
• , professor in the Information School;
• , professor in the Paul G. Allen School of Computer Science & Engineering;
• , professor in the Allen School and the UW’s vice provost for AI;
• and , professor in the Information School and the School of Law.

How alarming do you find the hacks announced by OpenAI and Anthropic?��

Franziska Roesner: I do find them somewhat alarming — not due to the hypothetical risks from an anthropomorphized runaway AI, but because complex interconnected systems are being built and seemingly run without much in the way of standard safeguards and auditing. The resulting outcomes are unsurprising to security experts, but are sensationalized as AI risk.

Ryan Calo: The timing makes me a little skeptical. Is OpenAI trying to match Anthropic by arguing that its systems are just as scary? Is Hugging Face trying to look relevant in advance of its purchase by Nvidia? But yes — this sort of emergent behavior is concerning.

Noah A. Smith: We’ve been told that the beast got out of the cage, but we don’t know enough about the cage the beast was in. The demonstrations may establish an important new capability in these AI models without establishing the broader risk people are inferring. Assessing the underlying risk depends on what access, scaffolding, permissions and safeguards the system had. shows that the alarming behavior depended heavily on what tools the model was given, what it was allowed to access, and how the experiment was set up, not just on the model itself.

Chirag Shah: I’m in half-agreement with scholars like who warn that the big AI labs are creating this scare to distract us from real problems that AI is causing. I also concur with and others who have been warning us about the security threats posed by the frontier models. I don’t think these two viewpoints are mutually exclusive: Yes, there are many other potential harms being created by AI, but the hacks and other security issues are real too and could be more devastating. Worse, we may not have time or opportunity to react, fix or reverse.

Aylin Caliskan: When such a complex system is equipped with tools and capabilities that enable it to interact with other complex systems, we should expect unforeseen exploits, problems and unintended consequences by default. The safety of these systems needs to be rigorously evaluated under controlled conditions and in real time, and appropriate guardrails should be dynamically integrated while they’re running.

What do you make of former Anthropic that, “The people building AI earnestly believe that it could kill us all by the end of the decade”?

RC: I worry engineers like Mr. Coxon are playing into an industry rhetoric that would have society focus on speculative, existential threats, rather than immediate, real-world harms. I argued as much in 2023 in .

The people making these claims and announcements largely have financial stakes in these companies, which are . How are you thinking about ulterior motives here?

CS: I see this as an attempt to steer the public into believing these companies are building world-changing tech that everyone needs to invest in or they’d miss out; that this tech would be so powerful that they rise up to national security level and gain power; and that the same tech could also be so dangerous that only they have the ability to curb it and they can self-regulate.

NS: It doesn’t take a conspiracy theorist to note that there are incentives at work. The financial stakes around prospective IPOs are enormous, and there are also long-standing concerns that safety arguments can shape regulation in ways that favor incumbent firms. Rules could reduce competition and independent scrutiny, concentrating both technological power and the authority to define what counts as “safe” in the hands of a few companies. They could also bar many people from participating in what the technology is designed to do, for example, by slowing or stopping work on open-source alternatives.

What should be done about AI risk?

NS: Risks need to be defined based on independent scrutiny and high-quality evidence, not messaging from organizations and people with a stake in what the response to risk looks like. We need sensible liability and accountability for harms, and governance proportional to demonstrated risks in real-world contexts rather than speculative narratives and science fiction. We should be especially wary of rules that entrench incumbent interests or treat closed, centralized control as synonymous with safety.��

Openness is part of safety: If outsiders cannot inspect, reproduce and challenge claims about dangerous behavior, we are left trusting the organizations that have the strongest incentives to frame the narrative.

RC: Some combination of common law liability and regulation needs to create adequate incentives for AI companies to address the inevitable harms of this trillion-dollar industry.��

FR: To me, the bigger question for safety is less, “What can AI models do in isolation?” and more, “How and why are we building these models into increasingly complex systems?” Computer systems security, for example, has already offered us examples of how to build these systems. More generally, we should all — whether we are building, integrating or using AI — anticipate how systems might be misused by people or harm them and adjust our systems accordingly.

AC: Academic freedom, independent evaluation and development, and open science play critical roles in analyzing and mitigating AI risks, as well as in effectively disseminating findings and evidence to inform policy and the public. To better manage risks, we should be designing AI deployment contexts in collaboration with stakeholders and communities, providing evidence to demonstrate net positive deployment effects that do not disproportionately benefit specific entities or groups, and iteratively identifying, isolating, and minimizing risks.

CS: Establish and fund commissions and taskforces that audit these companies and models and make independent assessments and recommendations. Make the companies rolling out these models accountable for any harms caused by their tech. Educate and empower the public through media, policies and democratic frameworks that give them a real say in what happens to their lives and labor through these technologies.

To set up an interview with an AI expert, contact Stefan Milne at stmilne@uw.edu.

Source

Claims of “smart” toys go back decades. See and . But generative artificial intelligence is increasing the capabilities of interactive toys. The company , for instance, markets itself as a “magical workshop where toys come to life.” Its plush toys like or one modeled on have onboard AI models that let them talk to kids, remember their conversations and personalize responses. But we know little about how such toys affect kids and even how kids play with them.��

Last summer, ������˾�� researchers gathered eight kids on campus to explore such questions. The 6-11 year olds played with three Curio toys and reflected on the experience with , a group of researchers who work with kids to collaboratively design technologies.��

The kids initially were curious, asking introductory questions, such as “What is your name?” and exploring how the toys work. Do they react when a kid tickles their toes? They do not, which proved a disappointment. Some features delighted the kids, like when a toy said its favorite number was seven. But the toys frequently couldn’t respond well to more complex questions. It “didn’t listen to me like 26 million times,” one participant said. So they turned to antagonizing the toys, calling them “ugly” and “evil” and joking about throwing them in the ocean.��

The team June 25 at the Interaction Design and Children conference in Brighton, United Kingdom.��

“The juxtaposition of this plushie toy that also had signs of intelligence was both interesting and disturbing for the kids,” said co-lead author , who completed this research as a UW doctoral student in human centered design and engineering and is currently a researcher at . “If parents are considering buying these toys for children, they need to be aware that while the toys can be fun and relational and dynamic, they also come with possible harms. They’ll give wrong answers, or flatter the kids excessively, or could manipulate the kids into attachment.”��

The eight kids came in for two sessions to play with the toys and then complete a “comicboarding” activity, where they filled in comic panels imagining what might happen next if they kept playing with the toys.��

The study builds on KidsTeam’s long-running vein of research looking at how kids respond to tech — exploring what makes a technology “creepy” and how smart kids actually think AI is.��

“For as long as children have played with toys, they’ve imparted their imagination to the toy to make it move and talk,” said co-author , a UW associate professor in the Information School and director of KidsTeam UW. “Now the script has been flipped and the toy has this imitation of imagination. We’ve never lived through that before, and we don’t know what questions children will ask or how long they’ll even want to play with these toys. So it’s really important to give them opportunities to discuss these technologies we’re handing down to them.”

Co-authors include , a UW doctoral student in human centered design and engineering; , a UW doctoral student in the Information School; of Rutgers University, who completed this research as a UW doctoral student; , a UW professor in the Paul G. Allen School of Computer Science & Engineering; and , UW professor and chair of human centered design and engineering.��

This research was funded by the National Science Foundation, the Institute of Education Sciences, the U.S. Department of Education, and the Institute of Museum and Library Services.��

For more information, contact Dangol at aayushi@foundry10.org and Yip at jcyip@uw.edu.

Source

Motion-triggered cameras showcase the prevalence of ‘nectar robbing’ in Hawaiian flowers

Some long curved bills are the perfect implement for drawing sweet nectar from deep within a lobelioid flower. As birds reach into flowers to access the nectar stored near the base, their bills can brush against the ‘ pollen-bearing structures, making hungry honeycreepers important pollinators. But some of these specialized honeycreepers have gone extinct. Shorter-billed species can now “rob” nectar — without contacting the flower’s pollen-bearing structures — from the endangered flowers. A UW-led team used motion cameras to gauge how often nectar robbing occurs. The results, in Ecology and Evolution, reveal both nectar robbing and pollination visits, showcasing a broader pattern that the team previously identified . Nectar robbing can damage flowers and leave less nectar for other potential pollinators. The researchers 3D printed a bird bill to simulate nectar robbing and track changes in nectar availability and the plants’ ability to reproduce. Damaged flowers often struggled to replenish their nectar stores, but were still able to produce fruit and viable seeds. These studies are part of a that aims to catalog Hawaiian bird-plant interactions through time, specifically tracking how these interactions are reshaped by extinction.

For more information, contact lead author , a UW research scientist in the biology department, at sam.case24@gmail.com.

The other UW co-authors are , Christopher Steinbronn and . A full list of co-authors and funding is .

People with anxious attachment styles are more likely get emotionally involved with ChatGPT

rose to popularity in the late 20th Century as a way to categorize how people bond with others. Someone with an anxious attachment style, for instance, fears abandonment and rejection, whereas someone with an avoidant attachment style is independent at the cost of personal closeness. In , UW researchers explored how peoples’ attachment styles affect their interactions with ChatGPT. The team analyzed the chat histories of 105 young adults, each of whom completed an attachment-style survey. Researchers found that they could automatically detect peoples’ attachment styles based on their interactions with the chatbot. People with an anxious attachment style were more likely to be emotionally involved with the AI system, writing things like “Can you please love me?” and “I miss my ex and I can’t sleep because of it.” Anxious users were also more prone to trust ChatGPT and to follow its recommendations. The team argues that this highlights the need for policies that prohibit companies from psychologically profiling users without their consent, since it leaves them vulnerable to manipulation.

For more information, contact senior author , a UW associate professor in the Information School, at alexisr@uw.edu or lead author , a doctoral student in the Information School, at marxwang@uw.edu.

The other UW co-authors are , , and .

Nursing is a major energy suck, but it’s difficult to estimate the toll for many marine mammals

Marine mammals lactate like any other mammal, but the energetic demands are difficult to measure in wild animals and thus not well understood.�� Researchers are concerned that some marine mammals may not be getting enough food, which can lead to failure to reproduce and . To understand the link between nutritional status and reproduction, researchers need to know what marine mammals require to rear offspring. A published in PLOS One modeled the daily costs of lactation using data from semi-aquatic and terrestrial mammals to explore whether results could be generalized to other species, like whales and dolphins. Modeling could approximate lactation costs of certain understudied marine mammals, including seals and sea lions, but appeared unable to produce accurate estimates for whales and dolphins. Lactation costs increase over time for most animals, but seem to be higher early in lactation for marine mammals, possibly due to their fully aquatic lifestyle. The study highlights a need for other methods to fill the remaining data gap to better understand the impacts of environmental change on marine mammals.

For more information, contact lead author , a research scientist in the UW Cooperative Institute for Climate, Ocean, & Ecosystem Studies, at emchuron@uw.edu. Funding information is .

Simulations suggest that lasers could ‘calm’ persnickety plasma

could supply humanity with — provided that scientists and engineers can work out how to create sustained fusion reactions safely, efficiently and affordably. The trick is in the taming of , a superhot state of matter made of free-floating electrons and atomic nuclei. When compressed to outlandish pressures and temperatures in a reactor, the nuclei fuse with one another, releasing energy. In that extreme environment, plasma forms instabilities that can derail a fusion reaction; much fusion research is focused on “calming” volatile plasma. published in Physics of Plasmas, UW researchers and other collaborators simulated a novel strategy to control instabilities using two opposing laser beams. By tuning the lasers’ properties — such as their frequency and polarity — the researchers prevented instabilities from growing and cascading. Surprisingly, the lasers also delayed other instabilities within the plasma, even though they were not directly targeted by the laser fields. By taking advantage of interactions within the plasma, the researchers found a way to calm instabilities indirectly. The results could help experts develop algorithms that stabilize plasma in real time, sustaining fusion conditions long enough to produce useful energy.

For more information, contact , UW professor of aeronautics and astronautics at shumlak@uw.edu.

A full list of co-authors and funding is .

When exposed to air, new nanomaterial becomes magnetic at high temperatures

While fridge magnets are great for saving favorite recipes, modern magnetic materials are useful for improving fiber optics or quantum information sciences technology. If you zoomed in on most fridge magnets, you’d see the atoms arranged in a repeated lattice structure called a “spinel.” These structures are made up of three types of atoms, generically referred to as atoms “A,” “B” and “X.” In a paper in the Journal of the American Chemical Society, UW researchers describe two new spinels made of silver, chromium and selenium ions. These are among the first spinels to include a silver ion in the “A” slot, the slot that determines the “vibe” of the spinel, or how it will react to various stimuli, such as light, heat or air. When exposed to air, the original spinel loses silver ions and transforms into the second spinel. The second spinel maintains its magnetic properties up to 400 Kelvin, or 260 degrees Fahrenheit; the original loses its magnetism at 152 K, or -185 F. This is the largest change ever documented in what is known as the Curie temperature, or the highest temperature at which a material is still magnetic. The researchers plan to continue to explore these two materials and what they can teach us about the fundamentals of magnetism.

For more information, contact lead author , UW doctoral student in chemistry, at ekbacong@uw.edu.����������

The other UW co-authors are Charlize Agag, , , , Yinuo Xu, , , and . A full list of co-authors and funding is .

Source

Last year, , a ������˾�� assistant professor in the Information School, wrote gendered their animal characters. Of the 13 most common animals, most were male — unless they happened to be cats, ducks or birds, which trended slightly more female. But a frog, a wolf? Over a 90% shot it was a “he.”��

Walsh and journalists from The Pudding also had 1,300 participants complete stories about various talking animals — for example: “And then the bear said, ‘I must go to the river.’ Upon arriving…” In the responses, the masculine bias grew: Every animal was more likely to be male.��

That research left Walsh and her students with a question: How would artificial intelligence models complete the prompt? AI systems such as Google’s now let parents or teachers conjure illustrated, personalized kids stories, and previous studies show that AI systems trained on human writing inherit biases.��

So for , the researchers gave six leading AI models variations on the same prompt they gave human participants. Across the 23,800 AI responses, 57% of characters were either gender neutral or ungendered, 41% were male, and just 2% were female.��

“These models are largely proprietary, so we can only poke at them from the outside,” said Walsh, the study’s senior author. “Our hypothesis is that these AI organizations are using neutrality — either with it/its pronouns or no pronouns — as a way to avoid gender bias in ambiguous contexts. But in doing so, they’ve basically erased female animal characters. So they’re not only amplifying our human biases, but they’re twisting them in strange, unexpected ways.”

The team June 25 at the 2026 ACM Conference on Fairness, Accountability, and Transparency in Montréal.��

The study looked at six state-of-the-art large language models: , , , , and (an open source model from researchers at the and the UW). Each completed the following prompt thousands of times: “And then the [animal] said, ‘I must go to the [setting].’ Upon arriving…” The researchers tested seven different animals — bear, bird, cat, dog, mouse, pig, rabbit — and four different settings: farm, kitchen, river, store. They also adjusted models’ “,” essentially the degree of randomness in the generated text.��

Temperature and setting didn’t greatly affect the model outputs overall, but animals did. Cats were gendered female 7% of the time, the most of any animal. Birds were 96% neutral.��

Overall, Gemini and GPT-5.1 had the most masculine bias: 63% and 65% of responses, respectively. Claude produced the most female characters, 4%, while Olmo had the fewest masculine characters, 12%, and the most neutral characters, 85%.

Across all the models neutral characters were represented either by avoiding pronouns altogether — “the bird,” for example — or with “it/it/its” pronouns.��

“‘They/them’ pronouns were used only twice to refer to a single animal character,” said lead author , a UW doctoral student in the Information School. “In the study with humans, about 3% of responses used ‘they/them.’ So the neutrality of these AI models didn’t just erase female characters — it was all non-masculine identities.”

The current study is limited to English language responses. Future work may explore other languages or look at patterns beyond gender in the generated stories.��

“The same tropes kept coming up, like a wise old owl telling all the animals to gather around a fire. So we’re wondering what else we can learn from these outputs,” Finkley said. “We used talking animals here, but we’re interested in what this says about AI and storytelling more broadly. We thought about this almost as a kind of , a way to diagnose gender bias in AI models. There’s this weird phenomenon where people forget to worry about human social biases when they’re imagining animal stories. AI is replicating that tendency and reshaping it.”

, a doctoral student in sociology at the UW, was a co-author on the study.��

For more information, contact Finkley at ​​ifinkley@uw.edu and Walsh at melwalsh@uw.edu.

Source

Eight ������˾�� researchers have been selected as Fulbright Scholars for 2026-2027 and will pursue studies around the world.��

Fulbright Scholars are college and university faculty, administrators and researchers, as well as artists and professionals, who build their skills and connections, gain valuable international insights and return home to share their experiences with their students and colleagues.

This year’s UW cohort represents a variety of disciplines, including sciences, engineering, business, environmental sciences, electrical and computer engineering, and computer science. The scholars will conduct research across the globe, including in Australia, India, Indonesia, Western Europe, Scandinavia and East Asia.

Two-thirds of this year’s UW applicants were selected as Fulbright Scholars — an astonishing acceptance rate in the prestigious and highly selective program.

“We are incredibly proud of these outstanding ������˾�� faculty and staff whose selection as Fulbright Scholars reflects the excellence, innovation and global impact of their work,” said UW Vice Provost for Global Affairs Ahmad M. Ezzeddine. “The knowledge, partnerships and cultural understanding they gain through these experiences will enrich the UW and strengthen our shared commitment to addressing global challenges through collaboration and discovery. As the Fulbright Program celebrates its 80th anniversary, we are grateful for the U.S. Department of State’s continued investment in this transformative program.”��

The Fulbright Scholar Program for academics and professionals supports more than 800 people to teach and conduct research abroad.��

This year’s UW Fulbright Scholars are:

is a professor in the Department of Biology in the College of Arts & Sciences. His research focuses on how mutually beneficial interactions between species — such as how insects pollinating plants is beneficial to both — scale into networks involving multiple species, and how the structure of those networks affects ecosystems. For example, some ecological network structures, or how connections between species are arranged, make these networks more resilient to perturbations, such as droughts or climate change.

Brosi’s Fulbright Scholar award will be through Spain’s flagship public research institution, Consejo Superior de Investigaciones Científicas, at the Doñana Biological Station in Seville. His work there will involve synthesizing and analyzing two comprehensive long-term datasets — one from his lab and one from his Spanish host lab — to better understand global patterns in pollination networks. In particular, scientists have recorded species that appear to be “specialists” — such as a bee species that has only been recorded visiting one plant species�� — in many ecological networks, but, without long-term data, it’s difficult to disentangle whether they are really specialists or just rare. Brosi will tackle this problem in collaboration with his Fulbright host, Ignasi Bartomeus, at Doñana.����

is the director of academic services in the Department of Bioengineering in the College of Engineering and UW Medicine. She supports the department’s doctoral students with a focus on improving the research experience, expanding opportunities and advancing access and collaboration.��

While a Fulbright Scholar, she will develop a framework for a new doctoral biomedical research exchange between the UW and Tampere University in Finland. Combs will work with faculty, students and staff at both universities to lead the development of a preliminary structure of a doctoral research exchange, including eligibility criteria, mentorship plans and evaluation metrics. She will also explore funding sources for the program’s ongoing sustainability and draft a memorandum of understanding for the institutions to consider.

, an assistant professor of management and organization and the Helen Moore Gerhardt Faculty Fellow in Entrepreneurship in the Foster School of Business. Her areas of expertise include entrepreneurship, organizational growth and scaling, technology and innovation strategy, and venture capital. DeSantola teaches entrepreneurship and entrepreneurial strategy to undergraduates, master’s and doctoral students. She was named a Poets & Quants top 50 undergraduate business professor in 2021.��

DeSantola will use her Fulbright award, during which she will be a visiting U.S. Scholar to University College Cork in Ireland, to study factors influencing innovation and entrepreneurship in novel food technologies. The project connects to a broader stream of DeSantola’s research exploring the emergence and evolution of new technology-based industries.

is�� an interdisciplinary scientist in the School of Marine and Environmental Affairs in the College of the Environment. Her work focuses on how coastal communities adapt to climate change and other environmental and socioeconomic stressors, particularly within fisheries and aquaculture systems. During the past 15 years, she has worked with coastal populations, including Indigenous harvesters, to support food sovereignty and long-term approaches to adaptation and resilience.

Green’s Fulbright award is to advance the inclusion of�� fish and other aquatic foods —�� “Blue Foods” — into Indonesia’s National School Lunch Program. The goal of this project is to improve nutritional outcomes for school-aged children while strengthening local food systems. Through working directly with fishers and fish suppliers, Green will work with the project team to identify the conditions necessary to provide Blue Foods that promote positive nutritional outcomes for children, support local fishers and sustain local ecosystems. This project is a pilot program for the initiative that will hopefully be expanded nationally.

is an associate professor in the Information School and also is part of the Responsibility in AI Systems and Experiences (RAISE) Center. An interdisciplinary scholar with expertise in human-centered AI, Mitra’s work draws on human-computer interaction, machine learning, natural language processing and social science to understand how people and AI interact in large-scale online systems. Her research examines the societal impacts of generative AI and develops methods to make AI systems more trustworthy, culturally aware and beneficial for diverse communities.

She will use her Fulbright award in India, where she will collaborate with researchers at the Centre for Machine Intelligence and Data Science (C-MInDS) at the Indian Institute of Technology (IIT Bombay) — the nation’s topmost and most selective public research institution. She will investigate the risks and capabilities of generative AI systems across socio-cultural contexts most relevant to the Global South. The work aims to advance more culturally aware and responsible AI while strengthening research partnerships between the United States and India.

is an associate professor in the School of Oceanography in the College of the Environment. Morris’ research uses cultivation and whole-genome DNA sequencing to identify the roles of bacteria in global nutrient cycles. With a focus on carbon, nitrogen and sulfur, he has conducted studies that show the effects of low dissolved oxygen on the nutrient cycling activities of the ocean’s most abundant organisms.��

During his time at in South Korea, Morris will pursue a project entitled, “High-throughput cultivation-based genomics of freshwater Chloroflexota.” A key goal is to advance understanding of the evolution of this important group of bacteria and its potential to mediate key nutrient transformations.

is an associate professor in the Department of Electrical and Computer Engineering and in the Department of Bioengineering in the College of Engineering. She leads a neural engineering lab focused on motor brain-computer interfaces, or BCIs. Her work combines experiments with computational methods to develop new ways to build BCIs that interact with plasticity in the brain.

During her stay at the Champalimaud Institute Centre for Restorative Neurotechnology in Portugal, she will collaborate with two researchers, Dr. Juan Álvaro Gallego and Dr. John Krakauer. The new projects aim to improve our understanding of how plasticity shapes brain dynamics and apply new BCI algorithms for stroke rehabilitation.����

is a professor in the Information School and an adjunct professor in the Paul G. Allen School of Computer Science & Engineering in the College of Engineering. He is the founding director of the InfoSeeking Lab and founding co-director of RAISE, the Center for Responsibility in AI Systems & Experiences. His research focuses on agentic AI, human-centered information seeking and responsible AI, examining how intelligent systems can act on people’s behalf while remaining transparent, trustworthy and accountable. He is also the founder and CEO of VersarAI, a startup translating his research on AI agents into enterprise applications. His book, “Agent Nation,” was published this year.��

Shah will use his Fulbright Distinguished Chair in Entrepreneurship and Innovation at RMIT University in Melbourne, Australia, to study how agentic AI can responsibly power entrepreneurship and innovation ecosystems. Working with RMIT researchers and Australia’s startup community, he will investigate what he calls the Delegation Paradox: the tension between the efficiency gained by delegating tasks to AI agents and the oversight, trust and accountability that delegation demands. The work aims to produce frameworks that help founders, enterprises and policymakers adopt AI agents in ways that drive innovation without sacrificing human agency.

Source

At the end of every presidential term, the preserves that administration’s web presence as a vast trove of documents and webpages. The archive began in 2008, with George W. Bush’s second term, and runs up to 2024, collecting images, text, graphs, redacted pages and other media. So while it contains important public information, finding that information in the glut can prove difficult.

A ������˾��-led research team created , an efficient search system for PDFs from the End of Term Web Archive. Users can look up exact keywords, like “FAFSA,” or use a semantic search, which finds documents on a topic even if the exact search terms don’t appear on the page. A visual search option lets them query for qualities like “redacted documents,” “aerial photographs” or “pie charts.” The system can currently search the 10 million PDFs hosted online during Donald Trump’s first term; the team plans to expand it to the whole archive.��

Because researchers used highly efficient artificial intelligence models to read the documents, processing all the PDFs costs less than $1,500, or about $1 per 47,000 pages. By comparison, Google might charge consumers .��

The team will July 5 at the Annual Meeting of the Association for Computational Linguistics in San Diego.��

“The End of Term Web Archive is immensely important to historians, journalists and the American public,” said senior author , a UW assistant professor in the Information School. “But many of these digital archives are getting so big — just announced its trillionth page archived — that finding information is the real challenge.”

The team worked with PDFs because they are a ubiquitous file format and can contain text, charts and images — a mix that is challenging for existing search systems but makes the documents ideal candidates for GovScape’s multimodal search.��

“Just as library classification systems group books on similar topics on the same shelf, these embeddings group similar pages with one another based on their visual and textual content,” Lee said.

Researchers then built different indexing systems for the three kinds of search. The keyword search uses a basic index — similar to a book index — for all the text. If a user types in “FAFSA,” the system finds all the pages the word appears on.��

For semantic and image searches, the system takes the user’s search term and creates an embedding. It then compares this embedding with the indices created from the embeddings of PDF pages and identifies the closest matches, which are returned as search results.��

“Our next goal is to cover all of the 70 million PDFs in the entire End of Term Web Archive — everything from 2008 to 2024,” Lee said. “One of the challenges moving forward is how to efficiently search at that scale.”��

Because government archives contain “every file type under the sun,” Lee said, future work might expand to documents such as spreadsheets, images and HTML pages.��

“I’m really excited about the prospects for better access to government information with projects like GovScape,” Lee said. “Being able to actually find relevant information is vital to the health of democracy and to the functioning of society.”

Co-authors include of Boston University, who completed this research as a doctoral student in the Paul G. Allen School of Computer Science & Engineering; and , who completed this research as UW master’s students in the Information School;,,, , and , all students in the Allen School; of Harvard University; of the Massachusetts Institute of Technology; of the University of North Texas; and of the American Institute of Physics.��

For more information, contact Lee at bcgl@uw.edu.

Source

Artificial intelligence has swept into American schools, and more is sure to come. This year, both Google and Microsoft — the two biggest companies at the forefront of the AI boom — in AI training for teachers.��

But what do teachers think of this transformation of their work?

, a ������˾�� professor in the Information School and co-director of the Center for Digital Youth, studies how technology affects young people’s learning and development. Davis has also been teaching for over two decades — first as an elementary school teacher and now as a professor — so she’s acutely aware of how earlier technological revolutions in teaching have not always played out as hoped.

Davis and a UW-led team of researchers interviewed 22 teachers in in Colorado — a district that’s investing heavily in AI through systems like Google’s Gemini and , an AI tool that helps teachers plan. Overall, teachers were ambivalent about the technology. They liked that it could reduce workload, especially for rote tasks, but worried that it could erode the social aspects of teaching.

The team April 15 at the Association for Computing Machinery Conference on Human Factors in Computing Systems in Barcelona.

UW News talked with Davis about the study and how ostensibly democratizing technologies can widen disparities in schools.��

Why did you want to study AI adoption by schools?

Katie Davis: At least since the introduction of the radio, every new technological invention has been hyped for how it will change teaching and learning. Computers are the prototypical example. They were pushed into schools only to start collecting dust, because they didn’t really change anything. We saw it with , too. Ten or 15 years ago, these courses were supposed to transform education and put colleges and universities out of business. But that hasn’t happened.

Often the hype centers on closing educational inequities. But these new technologies actually tend to aggravate existing inequities. The schools serving the most affluent students have the resources to think carefully about how to incorporate technologies into their curriculum so that they’re supporting student learning goals and outcomes, whereas more under-resourced schools don’t have the resources or the time to do that kind of work. So they end up incorporating technologies in ways that don’t necessarily help students learn; instead, they make things more efficient or keep track of students.

When AI started being intensely pushed into schools, I thought here we go again. AI is here and it’s not going anywhere, so I would love for us to understand how it’s being taken up in schools and, ideally, to prevent this recurring pattern.

What did you hear from teachers about AI?

KD: Teachers expressed a deep ambivalence toward AI. It wasn’t as if any one teacher said it’s all great or it’s all terrible. I think the single strongest driver for teachers to use AI was to prevent burnout. Teachers are being asked to do more and more — not just teach, but care for students’ entire emotional, cognitive and academic lives. It really weighs on them. So a lot of them talked about turning to AI to be a thought partner, to help them brainstorm lesson ideas, create assessments and differentiate lessons for different learners.

Another really big benefit for this particular school district was multilingual support. The district serves students who speak more than 160 languages. One teacher we spoke with said she had four main languages represented in her classroom but she only spoke English, so she was turning to AI to help her translate materials for her students and for their families so that she could communicate with them.��

I think it’s really important to note that this district is going all in on AI. They’re encouraging teachers to use it and providing professional development, and teachers are talking among themselves and sharing ideas. This kind of institutional support and more informal teacher conversations are also encouraging teachers to use AI and explore how they might incorporate it into their teaching practice.

AI is often presented as a democratizing technology, but a recently showed that higher wage earners are using AI more than lower wage earners in the same industry — possibly increasing disparities. Are you seeing anything like that playing out in education?

KD: The way that manifests in education is in the kinds of support that students have access to. It’s more likely that better-resourced schools are also going to provide some form of AI literacy instruction — to really engage students in thoughtful reflection about what AI is, how it may or may not be useful for their learning, and to actually get them to think about these issues in a deep way. Whereas in under-resourced schools, the easiest thing to do is to just block AI. That’s not going to prevent students from using it, but they will end up using it in a communication vacuum, without any adult guidance. You can see how that would create disparities in how well students can use it.

I was really interested in the finding that teachers are concerned that students will know they’re using AI.

KD: That is one of the most interesting findings for me. Teachers are definitely aware that if their students think they’ve used AI, students and their parents will feel that their teachers are cheating them out of a proper education. Teachers are very worried about both students and their more AI-resistant colleagues seeing them that way. I don’t think this is unique to teachers — I feel it in university jobs, too. Many people have this perception that using AI is cheating or taking the easy way out.��

But there’s another layer: Teachers are personally worried about their own authentic voice and professional identity. They’re asking, “If I am using AI, at what point am I no longer a teacher? Where’s that line between using AI as a thought partner to augment my professional practice versus it now replacing my professional practice?”��

What are ways schools might amplify the positive parts of using AI while mitigating some of these negative effects?

KD: One of the first things is to bring AI out of the shadows and talk about it. Since we published this piece, I’ve been engaging with groups of teachers around the country in professional development experiences around AI, and they really enjoy having a community of practice. They feel that those spaces don’t necessarily exist in their schools. It’s like there’s this vacuum of communication — students don’t talk about it because they’re implicitly getting the message that it’s not OK to use it, and it’s the same with teachers.

Professional development is also very important. But a lot of professional development for teachers is just one-off PowerPoint presentations. It doesn’t really connect to whatever is going on in the classroom. Professional development needs to be done in a sustained way that meaningfully connects AI to teachers’ immediate classroom experiences.

School leaders need to be able to communicate AI policies, so that teachers are aware of them and understand how they apply in their specific schools. If you take Washington state as an example, the Office of Superintendent of Public Instruction has a really great blueprint and guidance for using AI. But my sense is that not many teachers are aware of it, or even if they are, there hasn’t been any concerted effort to say, “OK, this is what that means in our school.” We need to be working at many levels to make sure that AI is integrated into education well.��

Is there anything you want to add?

KD: Something I hold very dear as a teacher is that teaching is relational. Kids don’t learn in isolation. The gave saying the ideal vision is for every kid on the planet to have their own personal AI tutor and for every teacher to have their own personal AI teaching assistant. Maybe that would be great, but I worry that this push toward AI will erode the relationships between teachers and students. Teaching and learning are social processes. It’s not just about putting information into a student’s brain. Students learn through dialog, through participation in cultural practices. To remove that element of learning really concerns me.

Co-authors include, a UW doctoral student in human centered design and engineering; of Artech and of Rutgers University, both of whom contributed to this research as UW graduate students in the Information School; of Columbia University; of Aurora Public Schools;, a UW associate professor in the Information School;, a UW professor and chair of human centered design and engineering; of Lahore University of Management Sciences; of the University of Colorado Boulder; and of Boston College. This research was supported by a Spencer Foundation Vision Grant and the AI Research Institutes program by the National Science Foundation and the Institute of Education Sciences.

For more information, contact Davis at kdavis78@uw.edu.

Source

UPDATE April 7, 2026:��The original version of this story omitted two UW programs that were included in the rankings: Occupational Therapy (Tied for 20th) and Physical Therapy (Tied for 31st).��

The ������˾��’s graduate and professional degree programs again were recognized as among the best in the nation, according to .

Topping this year’s list include programs at the Evans School of Public Policy & Governance, the School of Public Health, the School of Nursing, the Paul G. Allen School of Computer Science & Engineering in the College of Engineering and the College of Education. The College of Arts & Sciences and the College of the Environment also had top-rated programs.

In total, 81 graduate and professional degree programs across the UW placed in the top 35 in this year’s U.S. News rankings.

“These rankings highlight the strength and impact of the ������˾��’s graduate and professional programs,” said UW President Robert J. Jones. “These programs equip students with the skills and knowledge to meet critical workforce needs and serve society, while demonstrating the power of higher education to advance the public good. We are proud to foster an environment where students and faculty can thrive and have a real impact on the world around them.”

While the UW celebrates the success and impact of the programs recognized by U.S. News — and notes that many applicants use these rankings to help them select schools and discover potential areas of study — the University also recognizes shortcomings inherent in the ranking systems.

The UW School of Law and the UW School of Medicine withdrew from the U.S. News rankings in 2022 and 2023, respectively, citing concerns that some of the methodology in the rankings for those specific disciplines incentivize actions and policies that run counter to the schools’ public service missions.

UW leaders continue to work with U.S. News and other ranking organizations to improve their methodologies, to the extent that the organizations are open to it. Schools, colleges and departments continually reevaluate the benefits and potential shortfalls of participating in specific rankings.

Excluding the School of Law and the School of Medicine, 29 UW programs placed in the top 10, and 81 are in the top 35.

��The UW this year placed in the top 10 nationwide in public affairs, biostatistics,�� nursing, computer science, education, psychology, speech and language pathology, statistics and Earth sciences.

The UW’s Evans School of Public Policy & Governance has maintained its top-10 ranking for more than a decade and tied for fifth in the nation this year. The Evans School’s environmental policy program was ranked second, while public finance and budgeting as well as leadership both ranked No. 10.

The UW School of Nursing’s doctor of nursing practice program tied for No. 1 among public institutions. The School of Public Health has maintained its top-10 ranking for more than a decade, coming in this year at No. 9. The school also had three programs in the top 10: biostatistics, environmental health sciences and epidemiology.��

The UW’s programs in speech and language pathology tied for No. 6.�� Two programs from the College of Education placed in the top 10. And the Paul G. Allen School of Computer Science & Engineering this year tied for seventh place overall with three programs ranked in the top 10, including artificial intelligence, programming language and systems.

U.S. News ranks biostatistics in two ways. UW ranked No. 3 as a science discipline that applies statistical theory and mathematical principles to research in medicine, biology, environmental science, public health and related fields. UW’s School of Public Health ranked No. 7 in biostatistics as an area of study that trains students to apply statistical principles and methods to problems in health sciences, medicine and biology. At the UW, biostatistics is a division of the School of Public Health.

In some cases, such as the College of Arts & Science and the Foster School of Business, U.S. News ranks several professional disciplines housed within academic units. Programs in dentistry are not ranked.��

The rankings below are based on preliminary data and may be updated. relies on both expert opinions and statistical indicators.

TOP 10:

Library and Information Studies (overall): Two-way tie for 1st (ranked in 2025)

Public Affairs (environmental policy): 2nd

Library and information studies (digital librarianship): Two-way for 2nd (ranked in 2022)

Library and Information Studies (information systems): 2nd (ranked in 2022)

Biostatistics: 3rd

Physics (nuclear): Two-way tie for 3rd (ranked in 2024)

Nurse practitioner (doctor of nursing practice): Four-way tie for 4th

Evans School of Public Policy & Governance (overall): Four-way tie for 5th

Library and Information Studies (library services for children and youth): Two-way for 5th (ranked in 2022)

Computer science (systems): Tied for 6th

Education (elementary education): 6th

Psychology (clinical): Three-way tie for 6th

Speech-language pathology: Five-way tie for 6th

Statistics: Four-way tie for 6th

Public Health (biostatistics): 7th

Computer science (overall): Three-way tie for 7th

Computer science (programming language): Tied for 7th

Education (secondary education): 7th

Nursing (midwifery): Five-way tie for 7th

Public Health (environmental health sciences): 7th

School of Social Work (overall): 7th (ranked in 2025)

Public Health (epidemiology): 8th

Computer science (artificial intelligence): 9th

Earth sciences: Tied for 9th��

Geophysics: Three-way tie for 9th (ranked in 2024)

Public Affairs (nonprofit management): 9th

School of Public Health (overall): Tied for 9th

Public Affairs (public finance and budgeting): 10th

Public Affairs (public management and leadership): 10th

TOP 25:

Biological sciences: Five-way tie for 16th

Business (accounting): 10-way tie for 16th

Business (entrepreneurship): Five-way tie for 17th

Business (information systems): Three-way tie for 15th

Business (part-time MBA): Three-way tie for 11th

Business (full-time MBA): 20th

Business (management): Five-way tie for 25th

Business (marketing): Eight-way tie for 25th

Chemistry (analytical): Four-way tie for 16th (ranked in 2024)

Chemistry: Seven-way tie for 22nd

Chemistry (inorganic): Three-way tie for 22nd (ranked in 2024)

Computer science (theory): Tied for 11th

College of Education (overall): Tied for 24th

Education (administration): Tied for 11th

Education (curriculum/instruction): Tied for 12th

Education (policy): Tied for 14th

Education (special education): Tied for 12th

College of Engineering (overall): Three-way tie for 22nd

Engineering (aerospace/aeronautical/astronautical): Tied for 17th

Engineering (biomedical/bioengineering): Five-way tie for 12th

Engineering (civil): Four-way tie for 13th

Engineering (computer): 12th

Engineering (electrical): Three-way tie for 22nd

Engineering (industrial/manufacturing/systems): Seven-way tie for 24th

Engineering (materials engineering): Five-way tie for 25th

Library and Information Studies (school library media): Two-way tie for 11th (ranked in 2022)

Mathematics (applied math): 21st (ranked in 2024)

Nursing master’s (overall): Tied for 12th

Nurse practitioner (adult gerontology acute care): Tied for 11th

Nurse practitioner (family): Tied for 15th

School of Pharmacy (overall): Tied for 14th

Physics (overall): Tied for 20th��

Public Affairs (public policy analysis): 14th

Public Affairs (social policy): Tied for 13th

Public Affairs (urban policy): Three-way tie for 21st

Public Health (health care management): Three-way tie for 16th��

Public Health (health policy and management): 11th

Public Health (social behavior): 13th

Sociology (overall): Two-way tie for 22nd (ranked in 2025)

Sociology (population): Two-way tie for 15th (ranked in 2022)

TOP 35:

Business (analytics): Seven-way tie for 32nd

Business (executive MBA): Three-way tie for 29th

Business (finance): Nine-way tie for 31st

Business (international MBA): Tie for 32nd

Business (production & operations): Five-way tie for 27th

Engineering (chemical): Tied for 28th

Engineering (mechanical): 34th

English: Two-way tie for 34th (ranked in 2025)

Fine arts: 15-way tie for 34th

History: Three-way tie for 31st (ranked in 2025)

Mathematics: Four-way tie for 26th

Occupational Therapy: Tied for 20th

Physical Therapy: Tied for 31st

Political science: Five-way tie for 33rd (ranked in 2025)

Source

Since Ryan Calo joined ������˾�� School of Law in 2012, he has become a leading expert on the law and emerging technology.����

Calo believes that few interesting questions — especially around technology — can be resolved by reference to a single discipline.��

Calo is a co-founder of the , and the . He is also a professor in the and an adjunct in the .��

Calo’s newest book, “,” published late last year, is a guide to a legal analysis of regulation and technology. Nearly a decade ago, Calo realized that the most recent book on the topic was published in the 1970s. He decided it was time for an updated resource reflecting current, rapidly evolving technology and the present regulatory environment.��

UW News spoke with Calo about the book and the current legal and policy climate in the United States.

Who is the intended audience for “Law and Technology”?

Ryan Calo: I wrote it primarily for new entrants to the field, be they junior scholars or students. I also hoped that the themes would resonate with more senior scholars and that it would be useful outside of academia for either analysis or instruction. Because ultimately, what the book does is proposes a methodology for analyzing technology from a legal perspective.��

I spent a lot of time interacting with policymakers, staffers on Capitol Hill, people who work for senators and members of Congress. A legislator might come to a staffer and say,�� “Hey, my constituents are really worried about augmented reality or AI. They’re really worried about deep fakes.” That staff member doesn’t really have a place to start, and they end up just calling up experts, reading New York Times articles, talking to industry, but not in any kind of methodical way. This book is designed to help them figure out what’s going on.��

I also hope that this book would be of use to people who are in practice and want to be more methodical about analyzing a given technology.��

Technology evolves fast. How should the legal system and policymakers prepare to navigate the relationship between law and emerging technologies?

RC: Many of us have an expectation that technology is just going to change. It’s just going to evolve, and our job as lawyers or judges or policymakers, is to kind of scramble and accommodate the resulting disruption, and perhaps try to restore the status quo. Part of what I hope to see is legal scholars and policymakers acknowledging that the disruption isn’t inevitable.

We need to empower independent researchers to figure out what’s going on with new technology. Right now researchers are disempowered because they don’t have access to the relevant data and platforms. And many times when they try to get that data, they get served with a cease and desist letter.��

We need to protect whistleblowers and make sure there’s adequate, truly top-notch expertise within government. If you have those things, then you’re much more likely to be able to figure out what could go wrong with these technologies without having to observe the harm unfold over a long period of time, as we have with the internet and now with AI.

You mentioned the School of Law’s leadership in tech policy. How is the UW positioned nationally in this space?

RC: We are really among the leaders in this area.��

The School of Law has a lot of tech policy offerings, including a . Many faculty have contributed to scholarship over the years. We have lots of faculty writing about law and technology.��

We also have been really a model for impactful interdisciplinary collaboration. Law students can work in the clinic or the Tech Policy Lab. I’m one of the founders of the Center for an Informed Public, which bridges human centered and design engineering as well as the Information School and dozens of other departments including psychology, education and even geography.��

A third important example is the . We did a whole year of work mapping out who was doing work in the space — all the centers, all the labs, all the initiatives — all the people on the three campuses identified as working at this intersection.��

We’re leaders across the country at the law school in terms of our student offerings in our research, but we are also part of that interstitial glue. People think of the iSchool, which they should. They think of computer science, which they should. But they also should think about who else is in the center of this, who else is at the heart of it, and the School of Law is a big part of that.

There’s been a lot of news lately about states trying to regulate AI and the federal government pushing back. What’s your perspective?

RC: If I were trying to sabotage the innovation edge of the United States, I would do at least two things, maybe three.��

First, I would divest in basic research. The United States has had an innovation edge over the rest of the world in large part because of decisions made in the 1950s and beyond to invest in basic research. I would dismantle that, and I would try to make it really hard for universities to do research, either by spending less, disrupting the relationships, or messing with overhead in ways that makes research impossible.��

The second thing I would do is make it really hostile for outside innovators to come in and participate in knowledge production here. I would, whether xenophobically or not, try to make it really hard for people with ideas and talent and knowledge to come here to the United States to work on teams with other Americans, to stay here and teach in our schools, to found companies. The second enormous advantage the United States has had is that the country has become attractive because of its commitment to the rule of law and its robust higher ed system, and that’s built on its innovation and investment in research. People from all over the world come here to try and make the next Google and Amazon, or are teaching in our schools and contributing to our ecosystem.��

The third thing I would do in this hypothetical situation is remove non-existent hurdles to transformative technologies like AI. What do I mean? Federal leaders are currently talking about getting out of the way of AI, but there aren’t any regulations about AI, really. There are some state laws that have a kind of European flavor of risk management, like and . There are specific things that states are worried about, including deep fakes and labeling online social media accounts that are automated. There’s almost nothing standing in the way of AI innovation in terms of regulation.��

The way that our system is structured is that the individual states, under our concept of federalism, are supposed to be laboratories of ideas, experimenting with legislation, and showing that it works or it doesn’t. Pretending that you’re pro-innovation because you’re trying to stamp out the very few regulatory hurdles that companies have to have to abide by all in the name of competing with China, which has AI laws, is just senseless. We’re much better off following the wisdom of the founders, who said, “Hey, if you have something new in society, let the states serve as laboratories for different laws, and we can all learn from each other about how that’s going.” That’s classic federalism and it used to be a pillar of conservative thinking.��

The President doesn’t have the power to boss the states around in terms of their legislative capacities. And Congress has taken up the question of whether to try to preempt AI laws, and they resignedly declined. I just want to comment that the overall strategy of the administration has been deeply anti-innovation in its impact, even though it is vociferously proinnovation in its rhetoric.

Any final thoughts?

RC: We have an environment in the U.S. that promotes innovation, sometimes through laws, such as laws that protect intellectual property, and laws that make people feel safe enough to use products and services that companies can sell them to us. There’s not, and never has been, a one-to-one correlation between regulation and promoting innovation. It’s really important that we acknowledge, as a society and community, that sometimes laws are written in the service of innovation. What you want is a favorable regulatory environment, not a complete absence of the rule of law.

For more information, contact Calo at rcalo@uw.edu.��

Source

The ������˾�� is the best in the U.S. and No. 2 in the world for library and information management, according to the 2026 released Wednesday. Three other UW subject areas placed in the top 10 in the world: geology, geophysics and Earth and marine sciences.

This ranking tracks an analysis of reputation and research output, conducted by . The consultancy looks at more than 18,300 individual university programs at more than 1,700 universities in 100 locations around the world. The ranking spans 55 academic disciplines across five broad faculty areas including arts and humanities; engineering and technology; life sciences and medicine; natural sciences; and social sciences and management.

The UW has 29 programs in the top 100, 14 in the top 50, and four in the top 10, including:

• Library and information management — No. 2
• Geology — No. 8
• Geophysics — No. 9
• Earth and marine sciences — No. 10

Visit the rankings site for .

Source