First-year MIT nuclear science and engineering (NSE) doctoral student Kaylee Cunningham is not the first person to notice that nuclear energy has a public relations problem. But her commitment to dispel myths about the alternative power source has earned her the moniker “Ms. Nuclear Energy” on TikTok and a devoted fan base on the social media platform.

Cunningham’s activism kicked into place shortly after a week-long trip to Iceland to study geothermal energy. During a discussion about how the country was going to achieve its net zero energy goals, a representative from the University of Reykjavik balked at Cunnigham’s suggestion of including a nuclear option in the alternative energy mix. “The response I got was that we’re a peace-loving nation, we don’t do that,” Cunningham remembers. “I was appalled by the reaction, I mean we’re talking energy not weapons here, right?” she asks. Incredulous, Cunningham made a TikTok that targeted misinformation. Overnight she garnered 10,000 followers and “Ms. Nuclear Energy” was off to the races. Ms. Nuclear Energy is now Cunningham’s TikTok handle.

A theater and science nerd

TikTok is a fitting platform for a theater nerd like Cunningham. Born in Melrose, Massachusetts, Cunningham’s childhood was punctuated by moves to places where her roofer father’s work took the family. She moved to North Carolina shortly after fifth grade and fell in love with theater. “I was doing theater classes, the spring musical, it was my entire world,” Cunningham remembers. When she moved again, this time to Florida halfway through her first year of high school, she found the spring musical had already been cast. But she could help behind the scenes. Through that work, Cunningham gained her first real exposure to hands-on tech. She was hooked.

Soon Cunningham was part of a team that represented her high school at the student Astronaut Challenge, an aerospace competition run by Florida State University. Statewide winners got to fly a space shuttle simulator at the Kennedy Space Center and participate in additional engineering challenges. Cunningham’s team was involved in creating a proposal to help NASA’s Asteroid Redirect Mission, designed to help the agency gather a large boulder from a near-earth asteroid. The task was Cunningham’s induction into an understanding of radiation and “anything nuclear.” Her high school engineering teacher, Nirmala Arunachalam, encouraged Cunningham’s interest in the subject.

The Astronaut Challenge might just have been the end of Cunningham’s path in nuclear engineering had it not been for her mother. In high school, Cunningham had also enrolled in computer science classes and her love of the subject earned her a scholarship at Norwich University in Vermont where she had pursued a camp in cybersecurity. Cunningham had already laid down the college deposit for Norwich.

But Cunningham’s mother persuaded her daughter to pay another visit to the University of Florida, where she had expressed interest in pursuing nuclear engineering. To her pleasant surprise, the department chair, Professor James Baciak, pulled out all the stops, bringing mother and daughter on a tour of the on-campus nuclear reactor and promising Cunningham a paid research position. Cunningham was sold and Backiak has been a mentor throughout her research career.

Merging nuclear engineering and computer science

Undergraduate research internships, including one at Oak Ridge National Laboratory, where she could combine her two loves, nuclear engineering and computer science, convinced Cunningham she wanted to pursue a similar path in graduate school.

Cunningham’s undergraduate application to MIT had been rejected but that didn’t deter her from applying to NSE for graduate school. Having spent her early years in an elementary school barely 20 minutes from campus, she had grown up hearing that “the smartest people in the world go to MIT.” Cunningham figured that if she got into MIT, it would be “like going back home to Massachusetts” and that she could fit right in.

Under the advisement of Professor Michael Short, Cunningham is looking to pursue her passions in both computer science and nuclear engineering in her doctoral studies.

The activism continues

Simultaneously, Cunningham is determined to keep her activism going.

Her ability to digest “complex topics into something understandable to people who have no connection to academia” has helped Cunningham on TikTok. “It’s been something I’ve been doing all my life with my parents and siblings and extended family,” she says.

Punctuating her video snippets with humor — a Simpsons reference is par for the course — helps Cunningham break through to her audience who love her goofy and tongue-in-cheek approach to the subject matter without compromising accuracy. “Sometimes I do stupid dances and make a total fool of myself, but I’ve really found my niche by being willing to engage and entertain people and educate them at the same time.”

Such education needs to be an important part of an industry that’s received its share of misunderstandings, Cunningham says. “Technical people trying to communicate in a way that the general people don’t understand is such a concerning thing,” she adds. Case in point: the response in the wake of the Three Mile Island accident, which prevented massive contamination leaks. It was a perfect example of how well our safety regulations actually work, Cunningham says, “but you’d never guess from the PR fallout from it all.”

As Ms. Nuclear Energy, Cunningham receives her share of skepticism. One viewer questioned the safety of nuclear reactors if “tons of pollution” was spewing out from them. Cunningham produced a TikTok that addressed this misconception. Pointing to the “pollution” in a photo, Cunningham clarifies that it’s just water vapor. The TikTok has garnered over a million views. “It really goes to show how starving for accurate information the public really is,” Cunningham says, “ in this age of having all the information we could ever want at our fingertips, it’s hard to sift through and decide what’s real and accurate and what isn’t.”

Another reason for her advocacy: doing her part to encourage young people toward a nuclear science or engineering career. “If we’re going to start putting up tons of small modular reactors around the country, we need people to build them, people to run them, and we need regulatory bodies to inspect and keep them safe,” Cunningham points out. “ And we don’t have enough people entering the workforce in comparison to those that are retiring from the workforce,” she adds. “I’m able to engage those younger audiences and put nuclear engineering on their radar,” Cunningham says. The advocacy has been paying off: Cunningham regularly receives — and responds to — inquiries from high school junior girls looking for advice on pursuing nuclear engineering.

All the activism is in service toward a clear end goal. “At the end of the day, the fight is to save the planet,” Cunningham says, “I honestly believe that nuclear power is the best chance we’ve got to fight climate change and keep our planet alive.”

The Singapore MIT-Alliance for Research and Technology (SMART), MIT’s research enterprise in Singapore, has launched a new interdisciplinary research group aimed at tackling key social and institutional challenges around the rise of artificial intelligence and other new technologies. The group, known as Mens, Manus and Machina: How AI Empowers People, Institutions and the City in Singapore (M3S), aims to advance knowledge in these fields and foster collaborative research that generates positive impact for society in Singapore and the world.

Seeking to redefine the boundaries of AI, automation, and robotics through interdisciplinary research, knowledge sharing, and impactful collaborations, SMART M3S endeavors to design inclusive, resilient, and innovative solutions that empower individuals, institutions, and cities. By exploring the intricate relationship between human capabilities, emerging technologies, and societal structures, it is envisioned that SMART M3S will drive scientific, societal, and commercial impact in Singapore and beyond.

In line with Singapore’s Smart Nation initiative and its National AI Strategy, the project will embark on an ambitious five-year endeavor supported by a multimillion-dollar grant from the National Research Foundation of Singapore under its Campus for Research Excellence And Technological Enterprise program. 

Bringing together a diverse team of 17 professors from MIT and institutions in Singapore, SMART M3S will draw expertise from local researchers from Singapore Management University (SMU), Singapore University of Technology and Design, the National University of Singapore, and the National Robotics Program of Singapore. M3S will be guided by lead principal investigator Jinhua Zhao of MIT, co-lead principal investigator Daniela Rus of MIT, and co-lead principal investigator Archan Misra of SMU.

Ranked No. 1 in the 2023 Smart City Index, Singapore has facilitated the integration of AI, automation, and robotics by strategic use of data analytics, internet-of-things technologies, and smart infrastructure. Amid the rise of AI and machine learning, SMART M3S will contribute to Singapore’s AI ecosystem by focusing on the human-machine relationship, enhancing existing AI initiatives in the city-state.

Inspired by MIT’s motto of “mens et manus,” Latin for “mind and hand,” the name M3S reflects the research group’s ideals to promote AI and machine use for practical application — technologies that are extensions of humans and augment their lives. M3S integrates research on robotics and AI with human capital development, economic growth, and public acceptability — an intersectional approach to the ongoing transformation of how we work and live.

This interdisciplinary approach encompasses tackling key issues such as physical and digital interfaces between humans and machines, machine learning fundamentals, and understanding the implications of AI for human and social capital development. Other areas of focus include work on structuring human-machine teams within organizations and the developing dynamics between humans and machines in resource allocation and human labor (as well as machine power) management.

Research conducted could significantly advance aspects of soft robotics, brain interfaces, learning algorithms, task allocation, team formation, model compression, sustainable technology, technology acceptability in the workplace, social acceptability of robotics and AI, and more. The impact of AI on human welfare and productivity and how AI technology can advance both areas are central considerations for the work at SMART M3S, as society navigates the transition toward an AI- and machine-enhanced future.

“As a species, humans have spent eons learning how to work effectively with each other, but at the scale of human history, we are still neophytes to computation and automation,” says Zhao, an MIT professor of urban studies and planning who is also founding director of the MIT Mobility Initiative. “We focus on two questions at M3S: How will we design AI and robotics technologies and train humans to build the skills and habits necessary for success in a robotics-heavy work environment? How will we adapt our social and business institutions to create the incentives and protections necessary to drive innovation and social welfare?”

“The M3S collaboration between researchers at MIT and in Singapore, through SMART, will break new ground in our understanding of AI’s impact on the future of work,” adds Rus, the Andrew (1956) and Erna Viterbi Professor of Electrical Engineering and Computer Science at MIT and director of the MIT Computer Science and Artificial Intelligence Laboratory. “By harnessing our collective expertise and innovative spirit, we aim to advance the state of the art in AI and turn this technological advancement into an engine for human potential and societal progress.”

“M3S is distinguished by its ambition to address the key challenges of human-AI synergy holistically, from both a scientific and societal perspective,” notes Misra, vice provost for research and the Lee Kong Chian Professor of Computer Science at SMU who is also co-director of the A*STAR-SMU Joint Lab in Social and Human-Centered Computing. “It will focus not just on the technical breakthroughs that will allow human workers and AI-enabled machines and software to work interactively, but also on the training and governance mechanisms that ensure that individuals and organizations adapt to and thrive in this new future of work. I’m especially excited to collaborate with MIT researchers on this important national priority for Singapore, which aligns perfectly with SMU’s strategic multidisciplinary research priority area of digital transformation.”

Through interdisciplinary research, knowledge sharing, and impactful collaborations, SMART M3S will explore the intricate interplay between human capabilities, emerging technologies, and societal structures, paving the way for designing inclusive, resilient, and innovative solutions that empower individuals, institutions, and cities in Singapore. By engaging with Singaporean collaborators, SMART M3S hopes to enhance Singapore’s ability to create forward-looking AI policies, invigorate Singapore’s economic standing within AI, and support local workforce training and mentorship on AI topics. 

“With our latest interdisciplinary research group, SMART M3S, we further our commitment to bringing scientific, social, and commercial impact to Singapore and beyond,” says Eugene A. Fitzgerald, CEO and director of SMART. “The focus on a human-centric approach to AI advancement should contribute towards Singapore being at the forefront of the future of work.”

Since its inception in Singapore in 2007, SMART has developed innovations that have transformed and are transforming a multitude of fields such as autonomous driving, agriculture, microelectronics, cell therapy, mechanics and microfluidics platforms for biology and medical diagnostics, and antimicrobial resistance.

SMART was established by MIT in coordination with the National Research Foundation of Singapore in 2007 to undertake cutting-edge research in areas of interest to both Singapore and MIT. SMART currently comprises an Innovation Center and four interdisciplinary research groups: Antimicrobial Resistance, Critical Analytics for Manufacturing Personalized-Medicine, Disruptive and Sustainable Technologies for Agricultural Precision, and M3S.

Steven Gonzalez is a PhD candidate in the MIT Doctoral Program in History, Anthropology, Science, Technology, and Society (HASTS), where he researches the environmental impacts of cloud computing and data centers in the United States, Iceland, and Puerto Rico. He is also an author. Writing under the name E.G. Condé, he recently published his first book, “Sordidez.” It’s described as an “Indigenous futurist science fiction novella set in Puerto Rico and the Yucatán.” Set in the near future, it follows the survivors of civil war and climate disaster led by protagonist Vero Diaz, as they reclaim their Indigenous heritage and heal their lands.

In this Q&A, Gonzalez describes the book’s themes, its inspirations, and its connection to research, people, and classes at MIT.

Q: Where did the inspiration for this story come from?

A: I actually began my time at MIT in September of 2017 when Hurricane María struck. It was a really difficult time for me at the Institute, starting a PhD program. And it’s MIT, so there’s a lot of pressure. I was still kind of navigating the new institutional space and trying to understand my place in it. But I had a lot of people at the Institute who were extremely supportive during that time. I had family members in Puerto Rico who were stranded as a result of the hurricane, who I didn’t hear from for a very long time — who I feared dead. It was a very, very chaotic, confusing, and emotionally turbulent time for me, and also incredibly difficult to be trying to be present in a PhD program for the first semester. Karen Gardner, our administrator, was really incredibly supportive in that. Also the folks at the MIT Association of Puerto Ricans, who hosted fundraisers and linked students with counseling resources. But that trauma of the hurricane and the images that I saw of the aftermath of the hurricane, specifically in the town where my grandmother’s house was where I spent time living as a child during the summers, and to me, it was the greenest place that I have ever known. It looked like somebody had torched the entire landscape. It was traumatizing to see that image. But that kind of seeded the idea of, is there a way to burn without fire? There’s climate change, but there’s also climate terror. And so that was sort of one of the premises of the book explores, geoengineering, but also the flip side of geoengineering and terraforming is, of course, climate terror. And in a way, we could frame what’s been happening with the fossil fuel industry as a form of climate terror, as well. So for me, this all began right when I started at MIT, these dual tracks of thought.

Q: What do you see as the core themes of your novella?

A: One major theme is rebuilding. As I said, this story was very influenced by the trauma of Hurricane María and the incredibly inspiring accounts from family members, from people in Puerto Rico that I know, of regular people stepping up when the government — both federal and local — essentially abandoned them. There were so many failures of governance. But people stepped up and did what they could to help each other, to help neighbors. Neighbors cleared trees from roads. They banded together to do this. They pooled resources, to run generators so that everyone in the same street could have food that day. They would share medical supplies like insulin and things that were scarce. This was incredibly inspiring for me. And a huge theme of the book is rebuilding in the aftermath of a fictive hurricane, which I call Teddy, named after President Theodore Roosevelt, where Puerto Rico’s journey began as a U.S. commonwealth or a colony.

Healing is also a huge theme. Healing in the sense of this story was also somewhat critical of Puerto Rican culture. And it’s refracted through my own experience as a queer person navigating the space of Puerto Rico as a very kind of religious and traditional place and a very complex place at that. The main character, Vero, is a trans man. This is a person who’s transitioned and has felt a lot of alienation and as a result of his gender transition, a lot of people don’t accept him and don’t accept his identity or who he is even though he’s incredibly helpful in this rebuilding effort to the point where he’s, in some ways, a leader, if not the leader. And it becomes, in a way, about healing from the trauma of rejection too. And of course, Vero, but other characters who have gone through various traumas that I think are very much shared across Latin America, the Latin American experiences of assimilation, for instance. Latin America is a very complex place. We have Spanish as our language, that is our kind of lingua franca. But there are many Indigenous languages that people speak that have been not valued or people who speak them or use them are actively punished. And there’s this deep trauma of losing language. And in the case of Puerto Rico, the Indigenous language of the Taínos has been destroyed by colonialism. The story is about rebuilding that language and healing and “becoming.” In some ways, it’s about re-indigenization. And then the last part, as I said, healing, reconstruction, but also transformation and metamorphosis. And becoming Taíno. Again, what does that mean? What does it mean to be an Indigenous Caribbean in the future? And so that’s one of the central themes of the story.

Q: How does the novella intersect with the work you’re doing as a PhD candidate in HASTS?

A: My research on cloud computing is very much about climate change. It’s pitched within the context of climate change and understanding how our digital ecosystem contributes to not only global warming, but things like desertification. As a social scientist, that’s what I study. My studies of infrastructure are also directly referenced in the book in a lot of ways. For instance, the now collapsed Arecibo Ionosphere Observatory, where some of my pandemic fieldwork occurred, is a setting in the book. And also, I am an anthropologist. I am Puerto Rican. I draw both from my personal experience and my anthropological lens to make a story that I think is very multicultural and multilingual. It’s set in Puerto Rico, but the other half is set in the Yucatán Peninsula in what we’ll call the former Maya world. And there’s a lot of intersections between the two settings. And that goes back to the deeper Indigenous history. Some people are calling this Indigenous futurism because it references the Taínos, who are the Indigenous people of Puerto Rico, but also the Mayas, and many different Maya groups that are throughout the Yucatán Peninsula, but also present-day Guatemala and Honduras. And the story is about exchange between these two worlds. As someone trained as an anthropologist, it’s a really difficult task to kind of pull that off. And I think that my training has really, really helped me achieve that.

Q: Are there any examples of ways being among the MIT community while writing this book influenced and, in some ways, made this project possible?

A: I relied on many of my colleagues for support. There’s some sign language in the book. In Puerto Rico, there’s a big tradition of sign language. There’s a version of American sign language called LSPR that’s only found in Puerto Rico. And that’s something I’ve been aware of ever since I was a kid. But I’m not fluent in sign language or deaf communities and their culture. I got a lot of help from Timothy Loh, who’s in the HASTS program, who was extremely helpful to steer me towards sensitivity readers in the deaf community in his networks. My advisor, Stefan Helmreich, is very much a science fiction person in a lot of ways. His research is on the ocean waves, the history and anthropology of biology. He’s done ethnography in deep-sea submersibles. He’s always kind of thinking in a science fictional lens. And he allowed me, for one of my qualifying exam lists, to mesh science fiction with social theory. And that was also a way that I felt very supported by the Institute. In my coursework, I also took a few science fiction courses in other departments. I worked with Shariann Lewitt, who actually read the first version of the story. I workshopped it in her 21W.759 (Writing Science Fiction) class, and got some really amazing feedback that led to what is now a publication and a dream fulfilled in so many ways. She took me under her wing and really believed in this book.

The MIT Stephen A. Schwarzman College of Computing has awarded seed grants to seven projects that are exploring how artificial intelligence and human-computer interaction can be leveraged to enhance modern work spaces to achieve better management and higher productivity.

Funded by Andrew W. Houston ’05 and Dropbox Inc., the projects are intended to be interdisciplinary and bring together researchers from computing, social sciences, and management.

The seed grants can enable the project teams to conduct research that leads to bigger endeavors in this rapidly evolving area, as well as build community around questions related to AI-augmented management.

The seven selected projects and research leads include:

“LLMex: Implementing Vannevar Bush’s Vision of the Memex Using Large Language Models,” led by Patti Maes of the Media Lab and David Karger of the Department of Electrical Engineering and Computer Science (EECS) and the Computer Science and Artificial Intelligence Laboratory (CSAIL). Inspired by Vannevar Bush’s Memex, this project proposes to design, implement, and test the concept of memory prosthetics using large language models (LLMs). The AI-based system will intelligently help an individual keep track of vast amounts of information, accelerate productivity, and reduce errors by automatically recording their work actions and meetings, supporting retrieval based on metadata and vague descriptions, and suggesting relevant, personalized information proactively based on the user’s current focus and context.

“Using AI Agents to Simulate Social Scenarios,” led by John Horton of the MIT Sloan School of Management and Jacob Andreas of EECS and CSAIL. This project imagines the ability to easily simulate policies, organizational arrangements, and communication tools with AI agents before implementation. Tapping into the capabilities of modern LLMs to serve as a computational model of humans makes this vision of social simulation more realistic, and potentially more predictive.

“Human Expertise in the Age of AI: Can We Have Our Cake and Eat it Too?” led by Manish Raghavan of MIT Sloan and EECS, and Devavrat Shah of EECS and the Laboratory for Information and Decision Systems. Progress in machine learning, AI, and in algorithmic decision aids has raised the prospect that algorithms may complement human decision-making in a wide variety of settings. Rather than replacing human professionals, this project sees a future where AI and algorithmic decision aids play a role that is complementary to human expertise.

“Implementing Generative AI in U.S. Hospitals,” led by Julie Shah of the Department of Aeronautics and Astronautics and CSAIL, Retsef Levi of MIT Sloan and the Operations Research Center, Kate Kellog of MIT Sloan, and Ben Armstrong of the Industrial Performance Center. In recent years, studies have linked a rise in burnout from doctors and nurses in the United States with increased administrative burdens associated with electronic health records and other technologies. This project aims to develop a holistic framework to study how generative AI technologies can both increase productivity for organizations and improve job quality for workers in health care settings.

“Generative AI Augmented Software Tools to Democratize Programming,” led by Harold Abelson of EECS and CSAIL, Cynthia Breazeal of the Media Lab, and Eric Klopfer of the Comparative Media Studies/Writing. Progress in generative AI over the past year is fomenting an upheaval in assumptions about future careers in software and deprecating the role of coding. This project will stimulate a similar transformation in computing education for those who have no prior technical training by creating a software tool that could eliminate much of the need for learners to deal with code when creating applications.

“Acquiring Expertise and Societal Productivity in a World of Artificial Intelligence,” led by David Atkin and Martin Beraja of the Department of Economics, and Danielle Li of MIT Sloan. Generative AI is thought to augment the capabilities of workers performing cognitive tasks. This project seeks to better understand how the arrival of AI technologies may impact skill acquisition and productivity, and to explore complementary policy interventions that will allow society to maximize the gains from such technologies.

“AI Augmented Onboarding and Support,” led by Tim Kraska of EECS and CSAIL, and Christoph Paus of the Department of Physics. While LLMs have made enormous leaps forward in recent years and are poised to fundamentally change the way students and professionals learn about new tools and systems, there is often a steep learning curve which people have to climb in order to make full use of the resource. To help mitigate the issue, this project proposes the development of new LLM-powered onboarding and support systems that will positively impact the way support teams operate and improve the user experience.

Prior to receiving a PhD in computer science from MIT in 2017, Marzyeh Ghassemi had already begun to wonder whether the use of AI techniques might enhance the biases that already existed in health care. She was one of the early researchers to take up this issue, and she’s been exploring it ever since. In a new paper, Ghassemi, now an assistant professor in MIT’s Department of Electrical Science and Engineering (EECS), and three collaborators based at the Computer Science and Artificial Intelligence Laboratory, have probed the roots of the disparities that can arise in machine learning, often causing models that perform well overall to falter when it comes to subgroups for which relatively few data have been collected and utilized in the training process. The paper — written by two MIT PhD students, Yuzhe Yang and Haoran Zhang, EECS computer scientist Dina Katabi (the Thuan and Nicole Pham Professor), and Ghassemi — was presented last month at the 40th International Conference on Machine Learning in Honolulu, Hawaii.

In their analysis, the researchers focused on “subpopulation shifts” — differences in the way machine learning models perform for one subgroup as compared to another. “We want the models to be fair and work equally well for all groups, but instead we consistently observe the presence of shifts among different groups that can lead to inferior medical diagnosis and treatment,” says Yang, who along with Zhang are the two lead authors on the paper. The main point of their inquiry is to determine the kinds of subpopulation shifts that can occur and to uncover the mechanisms behind them so that, ultimately, more equitable models can be developed.

The new paper “significantly advances our understanding” of the subpopulation shift phenomenon, claims Stanford University computer scientist Sanmi Koyejo. “This research contributes valuable insights for future advancements in machine learning models’ performance on underrepresented subgroups.”

Camels and cattle

The MIT group has identified four principal types of shifts — spurious correlations, attribute imbalance, class imbalance, and attribute generalization — which, according to Yang, “have never been put together into a coherent and unified framework. We’ve come up with a single equation that shows you where biases can come from.”

Biases can, in fact, stem from what the researchers call the class, or from the attribute, or both. To pick a simple example, suppose the task assigned to the machine learning model is to sort images of objects — animals in this case — into two classes: cows and camels. Attributes are descriptors that don’t specifically relate to the class itself. It might turn out, for instance, that all the images used in the analysis show cows standing on grass and camels on sand — grass and sand serving as the attributes here. Given the data available to it, the machine could reach an erroneous conclusion — namely that cows can only be found on grass, not on sand, with the opposite being true for camels. Such a finding would be incorrect, however, giving rise to a spurious correlation, which, Yang explains, is a “special case” among subpopulation shifts — “one in which you have a bias in both the class and the attribute.”

In a medical setting, one could rely on machine learning models to determine whether a person has pneumonia or not based on an examination of X-ray images. There would be two classes in this situation, one consisting of people who have the lung ailment, another for those who are infection-free. A relatively straightforward case would involve just two attributes: the people getting X-rayed are either female or male. If, in this particular dataset, there were 100 males diagnosed with pneumonia for every one female diagnosed with pneumonia, that could lead to an attribute imbalance, and the model would likely do a better job of correctly detecting pneumonia for a man than for a woman. Similarly, having 1,000 times more healthy (pneumonia-free) subjects than sick ones would lead to a class imbalance, with the model biased toward healthy cases. Attribute generalization is the last shift highlighted in the new study. If your sample contained 100 male patients with pneumonia and zero female subjects with the same illness, you still would like the model to be able to generalize and make predictions about female subjects even though there are no samples in the training data for females with pneumonia.

The team then took 20 advanced algorithms, designed to carry out classification tasks, and tested them on a dozen datasets to see how they performed across different population groups. They reached some unexpected conclusions: By improving the “classifier,” which is the last layer of the neural network, they were able to reduce the occurrence of spurious correlations and class imbalance, but the other shifts were unaffected. Improvements to the “encoder,” one of the uppermost layers in the neural network, could reduce the problem of attribute imbalance. “However, no matter what we did to the encoder or classifier, we did not see any improvements in terms of attribute generalization,” Yang says, “and we don’t yet know how to address that.”

Precisely accurate

There is also the question of assessing how well your model actually works in terms of evenhandedness among different population groups. The metric normally used, called worst-group accuracy or WGA, is based on the assumption that if you can improve the accuracy — of, say, medical diagnosis — for the group that has the worst model performance, you would have improved the model as a whole. “The WGA is considered the gold standard in subpopulation evaluation,” the authors contend, but they made a surprising discovery: boosting worst-group accuracy results in a decrease in what they call “worst-case precision.” In medical decision-making of all sorts, one needs both accuracy — which speaks to the validity of the findings — and precision, which relates to the reliability of the methodology. “Precision and accuracy are both very important metrics in classification tasks, and that is especially true in medical diagnostics,” Yang explains. “You should never trade precision for accuracy. You always need to balance the two.”

The MIT scientists are putting their theories into practice. In a study they’re conducting with a medical center, they’re looking at public datasets for tens of thousands of patients and hundreds of thousands of chest X-rays, trying to see whether it’s possible for machine learning models to work in an unbiased manner for all populations. That’s still far from the case, even though more awareness has been drawn to this problem, Yang says. “We are finding many disparities across different ages, gender, ethnicity, and intersectional groups.”

He and his colleagues agree on the eventual goal, which is to achieve fairness in health care among all populations. But before we can reach that point, they maintain, we still need a better understanding of the sources of unfairness and how they permeate our current system. Reforming the system as a whole will not be easy, they acknowledge. In fact, the title of the paper they introduced at the Honolulu conference, “Change is Hard,” gives some indications as to the challenges that they and like-minded researchers face.

Since the invention of the telegraph, humans have been able to communicate across great distances in real-time. Today, we can choose among myriad technologies — radio, telephone, video conference platforms — to connect with colleagues and loved ones in different time zones, countries, and continents. These technologies create a telepresence — a sense of nearness between living beings separated only by space.

“The purpose of telepresence is to connect people who are alive,” says Hiroshi Ishii, the Jerome B. Wiesner Professor of Media Arts and Sciences at the MIT Media Lab, where he directs the Tangible Media research group. “But what about communicating with people who are no longer with us? That is the aim of TeleAbsence, our speculative design project. We attempt to bridge the vast emotional distance caused by bereavement. To create the illusion that we are communicating and interacting with a loved one who has departed. And to discover whether this illusory communication can help soothe our grief.”

Launched in the late 1990s, the Tangible Media research group works to give the virtual world a physical form. The group has invented tangible interface technologies, developed urban planning and simulation tools, and designed dozens of user interface devices that facilitate a merger of real and virtual environments. 

The TeleAbsence project, supported, in part, by the Center for Art, Science and Technology (CAST), is one of the group’s most ambitious efforts. In addition to blending the real and virtual worlds, it also probes — and imitates — the way humans process feelings of belonging, love, and loss. Originally inspired by bereavement, the project has evolved, and now addresses other forms of loss and emotional distance.

Dialing up past lives

“My practice has always explored the way objects and environments create identity and preserve memory,” says Danny Pillis, a graduate student in media arts and sciences and affiliate of the Media Lab. “And I’ve always been fascinated by the way experiences imprint themselves into the human mind.”

For the TeleAbsence project, Pillis collaborated with faculty and students at the Berklee College of Music and Xiao Xiao at the De Vinci Innovation Center in Paris to build the AmbientPhoneBooth. A personal immersive media environment, the AmbientPhoneBooth is an actual phone booth where users can connect with places and homes from their past. The user enters the booth, sits down, and dials a number on a rotary phone. A slow crescendo of sounds emerges, a heartbeat that then morphs into the cadence of a train traveling along a track, and then into a soothing lullaby. Designed by Ziaire Trinidad Sherman, the audio soundscape is controlled by the rotary dial. 

A beguiling and seamless blend of technologies across time, the AmbientPhoneBooth juxtaposes legacy technologies like rotary phones and locomotives with state-of-the-art virtual reality headsets and computation. “I am particularly interested in artifacts and media that predate and also anticipate the birth of modern computing,” says Pillis. “I want to see how these tangible artifacts can be advanced by contemporary technology. In a sense, it’s about connecting the Industrial Revolution with the digital revolution.”

Inspired, in part, by the Wind Phone in Japan’s Iwate Prefecture — a deactivated phone booth that thousands of people visit each year to “speak” to lost loved ones — the AmbientPhoneBooth is a work in progress. In the future, the user will be able to input family photos, home movies, and highly detailed scans of their childhood homes, and then experience these images as a hyper-real virtual reality recreation. At present, the virtual reality headset recreates the past experiences of a family home in Pittsburgh. “The goal is to create a template for future home media,” Pillis explains. “To provide people with the tools to create interactive virtual memories of a home or place they once inhabited. We all share a common human story, of space and place and identity.”  

 Words no longer yours

For Kyung Yun Choi, a media arts and sciences graduate student and Media Lab affiliate from Seoul, South Korea, the TeleAbsence project meant probing the relationship between her external and inner self. “Part of my project comes from speaking in English here in the U.S. instead of Korean,” says Choi. “In this new language, the words I spoke didn’t feel like mine. It was as if I’d become a different person, a person I was observing from a distance.”

To recreate that sense of distance — an alienation from self she describes as simply “weird” — Choi created an interactive installation in which an user speaks into a microphone. Those spoken words are relayed to an automated typewriter that translates the speech into Morse code — excluding the words “I” and “you.” The installation simulates the perception of distance and the experience of losing authority over one’s own thoughts. “In the past, people wrote on typewriters, and used Morse code to deliver messages across distance,” she explains. “Now we use speech recognition to send text messages. I wanted to capture that history of technology, to bring back the memories associated with them, and make them shareable.” 

Brushstrokes through time

While Choi’s TeleAbsence project aims to help people connect with their inner selves, Cathy Fang’s project was born out of a desire to connect more deeply with her grandparents, who live halfway around the world. 

“During my childhood, in Shanghai, I connected with my grandparents through Chinese calligraphy,” says Fang MS ’23, a PhD student in media arts and sciences and Media Lab affiliate. “What resonates with me, for this TeleAbsence project, was the experience of having a connection to a person and having that connection somehow lost or altered because of time and space.” 

While Fang does communicate with her grandparents by phone and video — she also visits with them when she returns to China — she misses the intimacy she shared with them through the rich sensory experience of tracing brushstrokes on paper. So she created a digitally-driven, three-dimensional plotter that recreates her grandfather’s brush strokes; the machine drafts elegant Chinese characters in black ink on paper. But it’s not the final product that interests her. “This isn’t a tool to recreate characters,” she explains. “It’s an attempt to draw attention to the person behind the traces: their movements and the pressure of their hand on the brush bearing down on paper that are otherwise ephemeral. Traces that can evoke your shared memories and connections.”

Ishii believes these projects can help ease the sense of loss and distance that all humans experience.

“Our goal is to soothe the pain of ‘Saudade,’” he says, “the hard-to-translate Portuguese word for melancholy and longing. And we will do that through tangible objects, abstract ambient media, and by capturing the shadows that people cast as they move through life. Ultimately, I want to see the hand of Cathy Fang’s grandfather in the place of the plotter, so it seems that he, and not the machine, is drawing the characters. That is when I will be satisfied.”  

When you picture MIT doctoral students taking small PhD courses together, you probably don’t imagine them going on class field trips. But it does happen, sometimes, and one of those trips changed Andy Sun’s career.

Today, Sun is a faculty member at the MIT Sloan School of Management and a leading global expert on integrating renewable energy into the electric grid. Back in 2007, Sun was an operations research PhD candidate with a diversified academic background: He had studied electrical engineering, quantum computing, and analog computing but was still searching for a doctoral research subject involving energy. 

One day, as part of a graduate energy class taught by visiting professor Ignacio J. Pérez Arriaga, the students visited the headquarters of ISO-New England, the organization that operates New England’s entire power grid and wholesale electricity market. Suddenly, it hit Sun. His understanding of engineering, used to design and optimize computing systems, could be applied to the grid as a whole, with all its connections, circuitry, and need for efficiency. 

“The power grids in the U.S. continent are composed of two major interconnections, the Western Interconnection, the Eastern Interconnection, and one minor interconnection, the Texas grid,” Sun says. “Within each interconnection, the power grid is one big machine, essentially. It’s connected by tens of thousands of miles of transmission lines, thousands of generators, and consumers, and if anything is not synchronized, the system may collapse. It’s one of the most complicated engineering systems.”

And just like that, Sun had a subject he was motivated to pursue. “That’s how I got into this field,” he says. “Taking a field trip.”

Sun has barely looked back. He has published dozens of papers about optimizing the flow of intermittent renewable energy through the electricity grid, a major practical issue for grid operators, while also thinking broadly about the future form of the grid and the process of making almost all energy renewable. Sun, who in 2022 rejoined MIT as the Iberdrola-Avangrid Associate Professor in Electric Power Systems, and is also an associate professor of operations research, emphasizes the urgency of rapidly switching to renewables.

“The decarbonization of our energy system is fundamental,” Sun says. “It will change a lot of things because it has to. We don’t have much time to get there. Two decades, three decades is the window in which we have to get a lot of things done. If you think about how much money will need to be invested, it’s not actually that much. We should embrace this future that we have to get to.”

Successful operations

Unexpected as it may have been, Sun’s journey toward being an electricity grid expert was informed by all the stages of his higher education. Sun grew up in China, and received his BA in electronic engineering from Tsinghua University in Beijing, in 2003. He then moved to MIT, joining the Media Lab as a graduate student. Sun intended to study quantum computing but instead began working on analog computer circuit design for Professor Neil Gershenfeld, another person whose worldview influenced Sun.  

“He had this vision about how optimization is very important in things,” Sun says. “I had never heard of optimization before.” 

To learn more about it, Sun started taking MIT courses in operations research. “I really enjoyed it, especially the nonlinear optimization course taught by Robert Freund in the Operations Research Center,” he recalls. 

Sun enjoyed it so much that after a while, he joined MIT’s PhD program in operations research, thanks to the guidance of Freund. Later, he started working with MIT Sloan Professor Dimitri Bertsimas, a leading figure in the field. Still, Sun hadn’t quite nailed down what he wanted to focus on within operations research. Thinking of Sun’s engineering skills, Bertsimas suggested that Sun look for a research topic related to energy. 

“He wasn’t an expert in energy at that time, but he knew that there are important problems there and encouraged me to go ahead and learn,” Sun says. 

So it was that Sun found himself in ISO-New England headquarters one day in 2007, finally knowing what he wanted to study, and quickly finding opportunities to start learning from the organization’s experts on electricity markets. By 2011, Sun had finished his MIT PhD dissertation. Based in part on ISO-New England data, the thesis presented new modeling to more efficiently integrate renewable energy into the grid; built some new modeling tools grid operators could use; and developed a way to add fair short-term energy auctions to an efficient grid system.

The core problem Sun deals with is that, unlike some other sources of electricity, renewables tend to be intermittent, generating power in an uneven pattern over time. That’s not an insurmountable problem for grid operators, but it does require some new approaches. Many of the papers Sun has written focus on precisely how to increasingly draw upon intermittent energy sources while ensuring that the grid’s current level of functionality remains intact. This is also the focus of his 2021 book, co-authored with Antonio J. Conejo, “Robust Optimiziation in Electric Energy Systems.”

“A major theme of my research is how to achieve the integration of renewables and still operate the system reliably,” Sun says. “You have to keep the balance of supply and demand. This requires many time scales of operation from multidecade planning, to monthly or annual maintenance, to daily operations, down through second-by-second. I work on problems in all these timescales.”

“I sit in the interface between power engineering and operations research,” Sun says. “I’m not a power engineer, but I sit in this boundary, and I keep the problems in optimization as my motivation.”

Culture shift

Sun’s presence on the MIT campus represents a homecoming of sorts. After receiving his doctorate from MIT, Sun spent a year as a postdoc at IBM’s Thomas J. Watson Research Center, then joined the faculty at Georgia Tech, where he remained for a decade. He returned to the Institute in January of 2022.

“I’m just very excited about the opportunity of being back at MIT,” Sun says. “The MIT Energy Initiative is a such a vibrant place, where many people come together to work on energy. I sit in Sloan, but one very strong point of MIT is there are not many barriers, institutionally. I really look forward to working with colleagues from engineering, Sloan, everywhere, moving forward. We’re moving in the right direction, with a lot of people coming together to break the traditional academic boundaries.” 

Still, Sun warns that some people may be underestimating the severity of the challenge ahead and the need to implement changes right now. The assets in power grids have long life time, lasting multiple decades. That means investment decisions made now could affect how much clean power is being used a generation from now. 

“We’re talking about a short timeline, for changing something as huge as how a society fundamentally powers itself with energy,” Sun says. “A lot of that must come from the technology we have today. Renewables are becoming much better and cheaper, so their use has to go up.”

And that means more people need to work on issues of how to deploy and integrate renewables into everyday life, in the electric grid, transportation, and more. Sun hopes people will increasingly recognize energy as a huge growth area for research and applied work. For instance, when MIT President Sally Kornbluth gave her inaugural address on May 1 this year, she emphasized tackling the climate crisis as her highest priority, something Sun noticed and applauded. 

“I think the most important thing is the culture,” Sun says. “Bring climate up to the front, and create the platform to encourage people to come together and work on this issue.”

Thanks to the support of libraries participating in Direct to Open (D2O), the MIT Press will publish its full list of 2023 scholarly monographs and edited collections open access.

Launched in 2021, D2O from the MIT Press is a sustainable framework that harnesses the collective power of libraries to support open and equitable access to vital, leading scholarship. D2O moves scholarly books from a solely market-based, purchase model, where individuals and libraries buy single e-books, to a collaborative open-access model, supported by libraries. Instead of purchasing a title once for a single collection, libraries now have the opportunity to fund MIT Press books one time for the world through participant fees.

“With the successful conclusion of our second year of Direct to Open, we are thrilled to make the press’s complete list of 2023 monographs openly available,” says Amy Brand, director and publisher of the MIT Press. “This achievement comes at a pivotal time for open science, research, and publishing and would not be possible without the partnership and collaboration of D2O member libraries and consortia. Together, we are proving open access scholarship is not only achievable, but sustainable and scalable.”

In its second year, D2O received support from 322 libraries around the globe, an increase of 33 percent from the first year. Expanding D2O’s international footprint, the press also entered into all-in agreements with Big Ten Academic Alliance and the Konsortium der sächsischen Hochschulbibliotheken, as well as central licensing and invoicing agreements with Council of Australian University Librarians, Center for Research Libraries, Greater Western Library Alliance, MOBIUS, Northeast Research Libraries, Jisc, Partnership for Academic Library Collaboration and Innovation, Statewide California Electronic Library Consortium, and Lyrasis.

“When we launched Direct to Open two years ago, we passionately believed that taking action to foster a more equitable, sustainable, and open scholarly communication ecosystem was vital and urgent,” says Amy Harris, senior manager of library relations and sales at the MIT Press. “Success was not guaranteed and has required dedicated, hard work to achieve this year; but we have been truly humbled by the support of all of the participating libraries and our consortia partners.”

As the MIT Press enters its third year of D2O, momentum among libraries and readers continues to build. D2O titles were accessed over 320,000 times in their first year of availability, with titles like “Active Inference: The Free Energy Principle in Mind, Brain, and Behavior” by Thomas Parr, Giovanni Pezzulo and Karl J. Friston; “The Entangled Brain: How Perception, Cognition, and Emotion Are Woven Together” by Luiz Pessoa; and “Cognitive Robotics,” edited by Angelo Cangelosi and Minoru Asada, finding large international audiences. The collection also features MIT-affiliated authors and books like “Dare to Invent the Future: Knowledge in the Service of and through Problem-Solving” by Clapperton Chakanetsa Mavhunga, professor of science, technology, and society at MIT and “Playing Oppression: The Legacy of Conquest and Empire in Colonialist Board Games” by Mary Flanagan and Mikael Jakobsson, who teaches and conducts research at the MIT Game Lab.

Readers can browse the complete D2O collection here.

Fifth-generation, or 5G, mobile network technology is all the hype these days. Compared to 4G, this newest way of connecting wireless devices to cellular networks is designed to provide higher data rates, ultralow latency, improved reliability, expanded configurability, increased network capacity and availability, and connectivity among a larger number of users.

The U.S. Department of Defense (DoD) would like to leverage these commercial advances in their communications systems, but 5G, like its predecessors, lacks sufficiently robust security features. For military applications, wireless connectivity leaves communications vulnerable to unwanted detection (identifying the presence of signals), unwarranted geolocation (determining the origin of signals), and purposeful jamming (hindering the transmission and reception of signals). Before the DoD can fully harness 5G technology, networking vulnerabilities must be identified, quantified, and mitigated.

“For commercial communications, you may worry about interference a bit, but you don’t worry about anybody intentionally seeking to find you and disrupt your communications, as is the case in the military,” explains Nicholas Smith, a researcher in the Tactical Networks Group, part of the Communication Systems R&D area at MIT Lincoln Laboratory. “The military also has to contend with more challenging mobility scenarios beyond people walking or driving around, such as airplanes traveling at Mach speeds.”

Smith is part of a Lincoln Laboratory team assessing the vulnerabilities of 5G and developing potential solutions to make this latest-generation technology resilient enough for military use.

Mountains of data        

In April 2022, with funding provided by the DoD FutureG and 5G Office, the Lincoln Laboratory 5G vulnerability assessment team headed to Hill Air Force Base (AFB) near Salt Lake City, Utah, to conduct an over-the-air test campaign at the newly opened 5G network test bed designed and installed by Nokia Corporation. The team is among the first to leverage this test bed at Hill AFB, which is one of five DoD FutureG and 5G Office test beds at U.S. military installations serving as locations for evaluating the capabilities and functionality of 5G networks. Though 5G vulnerabilities had previously been modeled, this testing campaign represented one of the first red-teaming campaigns against 5G in the field.

Over two weeks, the team deployed GPS-equipped antenna arrays connected to software-defined radios to collect network signals, which were then analyzed by a standalone computer server. Each day, the team drove three trucks, each containing one of these sensor systems, to different sites on base and asked the Hill AFB liaisons to tune certain network parameters — for example, to turn certain base stations on or off, increase or decrease the power of the base stations, or adjust the beam-steering directions. With each adjustment, the team collected data to determine how difficult it was to detect, geolocate, and jam 5G signals. The mountainous terrain enabled the team to obtain results from different elevations.

Before heading out to the field, the team performed modeling and simulation to prepare for their experimental setup, considering factors such as how far away from a 5G base station signals can be detected, where to place the sensors for the lowest geolocation error, and what the best sensor geometry is. They also verified the algorithms used for detection and geolocation.

On site at Hill AFB, the team consistently detected 5G signals through several types of detection algorithms, from general energy detectors (which measure the energy, or power, of a received signal) to more-specific matched-filter detectors (which compare the energy of an unknown received signal to the energy of a known signal). They detected signals up to the horizon (to around 20 kilometers out and verified further distances through simulation) — a very far range, particularly for a specific type of signal called the signal synchronization block (SSB). The SSB is detectable by design; mobile devices need to detect the SSB in order to synchronize to a wireless network’s time and frequency and ultimately access the network. However, this detectability means the SSB poses a considerable vulnerability.

“Detection facilitates jamming,” Smith says. “Once adversaries detect a signal, they can jam it. Because the SSB is periodic in time and frequency, it is quite easy to detect and then jam.”

To geolocate the signals, the team performed angle-of-arrival estimation using the MUSIC (for MUltiple SIgnal Classification) algorithm, which estimates the direction of arrival of signals received by an antenna array. As Smith explained, if you have two sensors spaced out on opposite sides of the map and know the angle that the signal is coming from for both sensors, you can draw straight lines that will intersect; where they intersect is the geolocation point.

“One of our objectives was to see how inexpensive or easy it would be to detect, geolocate, and jam 5G signals,” Smith explains. “Our results show that you don’t need to be highly sophisticated; commercially available off-the-shelf, low-cost hardware setups and open-source algorithms are effective.”

This 5G vulnerability assessment is an extension of previous 4G vulnerability assessments conducted by the laboratory.

Generational advances

New generations of wireless communications technology typically appear once per decade. Focusing on voice, the first generation, 1G, paved the way for the first mobile telephones in the 1980s. The second generation, 2G, enabled more secure voice transmission with less static and introduced short message services (SMS), or text messaging. With the debut of 3G in the early 2000s came the core network speeds needed to launch the first smartphones, bringing internet to our phones to support mobile applications such as maps and video calling. And 4G, providing even higher data-transfer rates, enabled high-definition video streaming, enhanced voice call quality (through long-term evolution, or LTE, technology), and internet-of-things devices such as smartwatches and digital home assistants.

The rollout of 5G, which began in earnest in 2019 and continues to evolve, comes with orders-of-magnitude improvements in several areas, including speed, latency, connectivity, and flexibility. For example, 4G theoretically tops out at 1 gigabit per second for data speed, while 5G tops out at 20 gigabits per second — a rate 20 times faster. In addition to operating at low-band frequencies (below 6 GHz), 5G can operate at less-crowded millimeter-wave frequencies (above 24 GHz). The abundant spectrum available at these higher frequencies enables extreme capacity, ultrahigh throughput, and ultralow latency. However, because high-frequency signals experience scattering as they travel through the atmosphere, their range is limited. To address this limitation, researchers are introducing concepts to complement the currently large cellphone towers (macrocells), which are located miles apart, with smaller towers (microcells, picocells, or femtocells) spaced closer together, particularly in high-density urban areas. With these small cells, the high frequencies don’t have to travel as far and can provide high data rates to lots of users.

Massive multiple-input, multiple-output (MIMO) antenna arrays provide another means to serve concurrent users. Providing a large number of antennas at 5G base stations means wireless signals can be tightly focused in targeted directions toward a desired receiving device such as a cellphone, laptop, or autonomous car, instead of spreading in all directions. Called beamforming, this focusing technique helps users get more precise, reliable wireless connections with faster data transfer and prevents the data from going to unintended recipients.

“5G presents an opportunity for communications to be much more based on beamforming and massive MIMO,” Smith says. “With these technologies, 5G has the potential to be less detectable and geolocatable and more anti-jam than all of the previous generations. But we need to be informed on how to configure the network to do that, because 5G is not inherently secure.”

Improved resilience

Over the past year, the team has been applying the insights from their field-testing campaign to enhance the resiliency of standard 5G components and processes.

“Our goal is to make the resiliency enhancements as simple and cost-effective as possible for the DoD to implement, leveraging existing 5G technology and not having to modify 5G hardware, at least on the cellphone side,” Smith says. 

Going forward, Smith is excited to design more complex algorithms, especially ones that use machine learning to detect and geolocate 5G signals. He also expressed the team’s interest in potentially using 5G for drone swarms, which, according to Smith, are “one of the hardest problems as far as communications go” because of factors like movement complexity and power limitations.

If the 10-year technology cycle keeps up, 6G will likely launch around 2030. New capabilities may include applying artificial intelligence to manage network resources; extending frequencies to even higher (terahertz) ranges; and integrating communications across land, air, sea, and space into a cohesive ecosystem.

“Our current program is actually called 5G-to-nG [next generation],” says Smith. “We’re already looking ahead to 6G and the vulnerabilities it may bring for the DoD.”

For a small percentage of cancer patients, doctors are unable to determine where their cancer originated. This makes it much more difficult to choose a treatment for those patients, because many cancer drugs are typically developed for specific cancer types.

A new approach developed by researchers at MIT and Dana-Farber Cancer Institute may make it easier to identify the sites of origin for those enigmatic cancers. Using machine learning, the researchers created a computational model that can analyze the sequence of about 400 genes and use that information to predict where a given tumor originated in the body.

Using this model, the researchers showed that they could accurately classify at least 40 percent of tumors of unknown origin with high confidence, in a dataset of about 900 patients. This approach enabled a 2.2-fold increase in the number of patients who could have been eligible for a genomically guided, targeted treatment, based on where their cancer originated.

“That was the most important finding in our paper, that this model could be potentially used to aid treatment decisions, guiding doctors toward personalized treatments for patients with cancers of unknown primary origin,” says Intae Moon, an MIT graduate student in electrical engineering and computer science who is the lead author of the new study.

Alexander Gusev, an associate professor of medicine at Harvard Medical School and Dana-Farber Cancer Institute, is the senior author of the paper, which appears today in Nature Medicine.

Mysterious origins

In 3 to 5 percent of cancer patients, particularly in cases where tumors have metastasized throughout the body, oncologists don’t have an easy way to determine where the cancer originated. These tumors are classified as cancers of unknown primary (CUP).

This lack of knowledge often prevents doctors from being able to give patients “precision” drugs, which are typically approved for specific cancer types where they are known to work. These targeted treatments tend to be more effective and have fewer side effects than treatments that are used for a broad spectrum of cancers, which are commonly prescribed to CUP patients.

“A sizeable number of individuals develop these cancers of unknown primary every year, and because most therapies are approved in a site-specific way, where you have to know the primary site to deploy them, they have very limited treatment options,” Gusev says.

Moon, an affiliate of the Computer Science and Artificial Intelligence Laboratory who is co-advised by Gusev, decided to analyze genetic data that is routinely collected at Dana-Farber to see if it could be used to predict cancer type. The data consist of genetic sequences for about 400 genes that are often mutated in cancer. The researchers trained a machine-learning model on data from nearly 30,000 patients who had been diagnosed with one of 22 known cancer types. That set of data included patients from Memorial Sloan Kettering Cancer Center and Vanderbilt-Ingram Cancer Center, as well as Dana-Farber.

The researchers then tested the resulting model on about 7,000 tumors that it hadn’t seen before, but whose site of origin was known. The model, which the researchers named OncoNPC, was able to predict their origins with about 80 percent accuracy. For tumors with high-confidence predictions, which constituted about 65 percent of the total, its accuracy rose to roughly 95 percent.

After those encouraging results, the researchers used the model to analyze a set of about 900 tumors from patients with CUP, which were all from Dana-Farber. They found that for 40 percent of these tumors, the model was able to make high-confidence predictions.

The researchers then compared the model’s predictions with an analysis of the germline, or inherited, mutations in a subset of tumors with available data, which can reveal whether the patients have a genetic predisposition to develop a particular type of cancer. The researchers found that the model’s predictions were much more likely to match the type of cancer most strongly predicted by the germline mutations than any other type of cancer.

Guiding drug decisions

To further validate the model’s predictions, the researchers compared data on the CUP patients’ survival time with the typical prognosis for the type of cancer that the model predicted. They found that CUP patients who were predicted to have cancer with a poor prognosis, such as pancreatic cancer, showed correspondingly shorter survival times. Meanwhile, CUP patients who were predicted to have cancers that typically have better prognoses, such as neuroendocrine tumors, had longer survival times.

Another indication that the model’s predictions could be useful came from looking at the types of treatments that CUP patients analyzed in the study had received. About 10 percent of these patients had received a targeted treatment, based on their oncologists’ best guess about where their cancer had originated. Among those patients, those who received a treatment consistent with the type of cancer that the model predicted for them fared better than patients who received a treatment typically given for a different type of cancer than what the model predicted for them.

Using this model, the researchers also identified an additional 15 percent of patients (2.2-fold increase) who could have received an existing targeted treatment, if their cancer type had been known. Instead, those patients ended up receiving more general chemotherapy drugs.

“That potentially makes these findings more clinically actionable because we’re not requiring a new drug to be approved. What we’re saying is that this population can now be eligible for precision treatments that already exist,” Gusev says.

The researchers now hope to expand their model to include other types of data, such as pathology images and radiology images, to provide a more comprehensive prediction using multiple data modalities. This would also provide the model with a comprehensive perspective of tumors, enabling it to predict not just the type of tumor and patient outcome, but potentially even the optimal treatment.

The research was funded by the National Institutes of Health, the Louis B. Mayer Foundation, the Doris Duke Charitable Foundation, the Phi Beta Psi Sorority, and the Emerson Collective.