The intersection of art, science, and technology presents a unique, sometimes challenging, viewpoint for both scientists and artists. It is in this nexus that art historian Lindsay Caplan positions herself: “My work as an art historian focuses on the ways that artists across the 20th century engage with new technologies like computers, video, and television, not merely as new materials for making art as they already understand it, but as conceptual platforms for reorienting and reimagining the foundational assumptions of their practice.”

With this introduction, Caplan, an assistant professor at Brown University, opened the inaugural Resonances Lecture — a new series by STUDIO.nano to explore the generative edge where art, science, and technology meet. Delivered on April 28 to an interdisciplinary crowd at MIT.nano, Caplan’s lecture, titled “Analogical Engines — Collaborations across Art and Technology in the 1960s,” traced how artists across Europe and the Americas in the 1960s engaged with and responded to the emerging technological advances of computer science, cybernetics, and early AI. “By the time we reached the 1960s,” she said, “analogies between humans and machines, drawn from computer science and fields like information theory and cybernetics, abound among art historians and artists alike.”

Kaplan’s talk centered on two artistic networks, with a particular emphasis on American artist Liliane Lijn: New Tendencies exhibitions (1961-79) and the Signals gallery in London (1964-66). She deftly analyzed the artist’s material experimentation with contemporary advances in emergent technologies — quantum physics and mathematical formalism, particularly Heisenberg's uncertainty principle. She argued that both art historical formalism and mathematical formalism share struggles with representation, indeterminacy, and the tension between constructed and essential truths.

Following her talk, Caplan was joined by MIT faculty Mark Jarzombek, professor of the history and theory of architecture, and Gediminas Urbonas, associate professor of art, culture, and technology (ACT), for a panel discussion moderated by Ardalan SadeghiKivi SM ’22, lecturer of comparative media studies. The conversation expanded on Caplan’s themes with discussions of artists’ attraction to newly developed materials and technology, and the critical dimension of reimagining and repurposing technologies that were originally designed with an entirely different purpose.

Urbonas echoed the urgency of these conversations. “It is exceptionally exciting to witness artists working in dialectical tension with scientists — a tradition that traces back to the founding of the Center for Advanced Visual Studies at MIT and continues at ACT today,” reflected Urbonas. “The dual ontology of science and art enables us to grasp the world as a web of becoming, where new materials, social imaginaries, and aesthetic values are co-constituted through interdisciplinary inquiry. Such collaborations are urgent today, offering tools to reimagine agency, subjectivity, and the role of culture in shaping the future.”

The event concluded with a reception in MIT.nano’s East Lobby, where attendees could view MIT ACT student projects currently on exhibition in MIT.nano’s gallery spaces. The reception was, itself, an intersection of art and technology. “The first lecture of the Resonances Lecture Series lived up to the title,” reflects Jarzombek. “A brilliant talk by Lindsay Caplan proved that the historical and aesthetical dimensions in the sciences have just as much relevance to a critical posture as the technical.”

The Resonances lecture and panel series seeks to gather artists, designers, scientists, engineers, and historians who examine how scientific endeavors shape artistic production, and vice versa. Their insights expose the historical context on how art and science are made and distributed in society and offer hints at the possible futures of such productions.

“When we were considering who to invite to launch this lecture series, Lindsay Caplan immediately came to mind,” says Tobias Putrih, ACT lecturer and academic advisor for STUDIO.nano. “She is one of the most exciting thinkers and historians writing about the intersection between art, technology, and science today. We hope her insights and ideas will encourage further collaborative projects.”

The Resonances series is one of several new activities organized by STUDIO,nano, a program within MIT.nano, to connect the arts with cutting-edge research environments. “MIT.nano generates extraordinary scientific work,” says Samantha Farrell, manager of STUDIO.nano, “but it’s just as vital to create space for cultural reflection. STUDIO.nano invites artists to engage directly with new technologies — and with the questions they raise.”

In addition to the Resonances lectures, STUDIO.nano organizes exhibitions in the public spaces at MIT.nano, and an Encounters series, launched last fall, to bring artists to MIT.nano. To learn about current installations and ongoing collaborations, visit the STUDIO.nano web page.

For weeks, the whiteboard in the lab was crowded with scribbles, diagrams, and chemical formulas. A research team across the Olivetti Group and the MIT Concrete Sustainability Hub (CSHub) was working intensely on a key problem: How can we reduce the amount of cement in concrete to save on costs and emissions? 

The question was certainly not new; materials like fly ash, a byproduct of coal production, and slag, a byproduct of steelmaking, have long been used to replace some of the cement in concrete mixes. However, the demand for these products is outpacing supply as industry looks to reduce its climate impacts by expanding their use, making the search for alternatives urgent. The challenge that the team discovered wasn’t a lack of candidates; the problem was that there were too many to sort through.

On May 17, the team, led by postdoc Soroush Mahjoubi, published an open-access paper in Nature’s Communications Materials outlining their solution. “We realized that AI was the key to moving forward,” notes Mahjoubi. “There is so much data out there on potential materials — hundreds of thousands of pages of scientific literature. Sorting through them would have taken many lifetimes of work, by which time more materials would have been discovered!”

With large language models, like the chatbots many of us use daily, the team built a machine-learning framework that evaluates and sorts candidate materials based on their physical and chemical properties. 

“First, there is hydraulic reactivity. The reason that concrete is strong is that cement — the ‘glue’ that holds it together — hardens when exposed to water. So, if we replace this glue, we need to make sure the substitute reacts similarly,” explains Mahjoubi. “Second, there is pozzolanicity. This is when a material reacts with calcium hydroxide, a byproduct created when cement meets water, to make the concrete harder and stronger over time.  We need to balance the hydraulic and pozzolanic materials in the mix so the concrete performs at its best.”

Analyzing scientific literature and over 1 million rock samples, the team used the framework to sort candidate materials into 19 types, ranging from biomass to mining byproducts to demolished construction materials. Mahjoubi and his team found that suitable materials were available globally — and, more impressively, many could be incorporated into concrete mixes just by grinding them. This means it’s possible to extract emissions and cost savings without much additional processing. 

“Some of the most interesting materials that could replace a portion of cement are ceramics,” notes Mahjoubi. “Old tiles, bricks, pottery — all these materials may have high reactivity. That’s something we’ve observed in ancient Roman concrete, where ceramics were added to help waterproof structures. I’ve had many interesting conversations on this with Professor Admir Masic, who leads a lot of the ancient concrete studies here at MIT.”

The potential of everyday materials like ceramics and industrial materials like mine tailings is an example of how materials like concrete can help enable a circular economy. By identifying and repurposing materials that would otherwise end up in landfills, researchers and industry can help to give these materials a second life as part of our buildings and infrastructure.

Looking ahead, the research team is planning to upgrade the framework to be capable of assessing even more materials, while experimentally validating some of the best candidates. “AI tools have gotten this research far in a short time, and we are excited to see how the latest developments in large language models enable the next steps,” says Professor Elsa Olivetti, senior author on the work and member of the MIT Department of Materials Science and Engineering. She serves as an MIT Climate Project mission director, a CSHub principal investigator, and the leader of the Olivetti Group.

“Concrete is the backbone of the built environment,” says Randolph Kirchain, co-author and CSHub director. “By applying data science and AI tools to material design, we hope to support industry efforts to build more sustainably, without compromising on strength, safety, or durability.

In addition to Mahjoubi, Olivetti, and Kirchain, co-authors on the work include MIT postdoc Vineeth Venugopal, Ipek Bensu Manav SM ’21, PhD ’24; and CSHub Deputy Director Hessam AzariJafari.

This spring, 25 MIT students and a postdoc traveled to Washington, where they met with congressional offices to advocate for federal science funding and specific, science-based policies based on insights from their research on pressing issues — including artificial intelligence, health, climate and ocean science, energy, and industrial decarbonization. Organized annually by the Science Policy Initiative (SPI), this year’s trip came at a particularly critical moment, as science agencies are facing unprecedented funding cuts.

Over the course of two days, the group met with 66 congressional offices across 35 states and select committees, advocating for stable funding for science agencies such as the Department of Energy, the National Oceanic and Atmospheric Administration, the National Science Foundation, NASA, and the Department of Defense.

Congressional Visit Days (CVD), organized by SPI, offer students and researchers a hands-on introduction to federal policymaking. In addition to meetings on Capitol Hill, participants connected with MIT alumni in government and explored potential career paths in science policy.

This year’s trip was co-organized by Mallory Kastner, a PhD student in biological oceanography at MIT and Woods Hole Oceanographic Institution (WHOI), and Julian Ufert, a PhD student in chemical engineering at MIT. Ahead of the trip, participants attended training sessions hosted by SPI, the MIT Washington Office, and the MIT Policy Lab. These sessions covered effective ways to translate scientific findings into policy, strategies for a successful advocacy meeting, and hands-on demos of a congressional meeting.

Participants then contacted their representatives’ offices in advance and tailored their talking points to each office’s committees and priorities. This structure gave participants direct experience initiating policy conversations with those actively working on issues they cared about.

Audrey Parker, a PhD student in civil and environmental engineering studying methane abatement, emphasizes the value of connecting scientific research with priorities in Congress: “Through CVD, I had the opportunity to contribute to conversations on science-backed solutions and advocate for the role of research in shaping policies that address national priorities — including energy, sustainability, and climate change.”

To many of the participants, stepping into the shoes of a policy advisor was a welcome diversion from their academic duties and scientific routine. For Alex Fan, an undergraduate majoring in electrical engineering and computer science, the trip was enlightening: “It showed me that student voices really do matter in shaping science policy. Meeting with lawmakers, especially my own representative, Congresswoman Bonamici, made the experience personal and inspiring. It has made me seriously consider a future at the intersection of research and policy.”

“I was truly impressed by the curiosity and dedication of our participants, as well as the preparation they brought to each meeting,” says Ufert. “It was inspiring to watch them grow into confident advocates, leveraging their experience as students and their expertise as researchers to advise on policy needs.”

Kastner adds: “It was eye-opening to see the disconnect between scientists and policymakers. A lot of knowledge we generate as scientists rarely makes it onto the desk of congressional staff, and even more rarely onto the congressperson’s. CVD was an incredibly empowering experience for me as a scientist — not only am I more motivated to broaden my scientific outreach to legislators, but I now also have the skills to do so.”

Funding is the bedrock that allows scientists to carry out research and make discoveries. In the United States, federal funding for science has enabled major technological breakthroughs and advancements in manufacturing and other industrial sectors, and led to important environmental protection standards. While participants found the degree of support for science funding variable among offices from across the political spectrum, they were reassured by the fact that many offices on both sides of the aisle still recognized the significance of science. 

When you’re trying to communicate or understand ideas, words don’t always do the trick. Sometimes the more efficient approach is to do a simple sketch of that concept — for example, diagramming a circuit might help make sense of how the system works.

But what if artificial intelligence could help us explore these visualizations? While these systems are typically proficient at creating realistic paintings and cartoonish drawings, many models fail to capture the essence of sketching: its stroke-by-stroke, iterative process, which helps humans brainstorm and edit how they want to represent their ideas.

A new drawing system from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Stanford University can sketch more like we do. Their method, called “SketchAgent,” uses a multimodal language model — AI systems that train on text and images, like Anthropic’s Claude 3.5 Sonnet — to turn natural language prompts into sketches in a few seconds. For example, it can doodle a house either on its own or through collaboration, drawing with a human or incorporating text-based input to sketch each part separately.

The researchers showed that SketchAgent can create abstract drawings of diverse concepts, like a robot, butterfly, DNA helix, flowchart, and even the Sydney Opera House. One day, the tool could be expanded into an interactive art game that helps teachers and researchers diagram complex concepts or give users a quick drawing lesson.

CSAIL postdoc Yael Vinker, who is the lead author of a paper introducing SketchAgent, notes that the system introduces a more natural way for humans to communicate with AI.

“Not everyone is aware of how much they draw in their daily life. We may draw our thoughts or workshop ideas with sketches,” she says. “Our tool aims to emulate that process, making multimodal language models more useful in helping us visually express ideas.”

SketchAgent teaches these models to draw stroke-by-stroke without training on any data — instead, the researchers developed a “sketching language” in which a sketch is translated into a numbered sequence of strokes on a grid. The system was given an example of how things like a house would be drawn, with each stroke labeled according to what it represented — such as the seventh stroke being a rectangle labeled as a “front door” — to help the model generalize to new concepts.

Vinker wrote the paper alongside three CSAIL affiliates — postdoc Tamar Rott Shaham, undergraduate researcher Alex Zhao, and MIT Professor Antonio Torralba — as well as Stanford University Research Fellow Kristine Zheng and Assistant Professor Judith Ellen Fan. They’ll present their work at the 2025 Conference on Computer Vision and Pattern Recognition (CVPR) this month.

Assessing AI’s sketching abilities

While text-to-image models such as DALL-E 3 can create intriguing drawings, they lack a crucial component of sketching: the spontaneous, creative process where each stroke can impact the overall design. On the other hand, SketchAgent’s drawings are modeled as a sequence of strokes, appearing more natural and fluid, like human sketches.

Prior works have mimicked this process, too, but they trained their models on human-drawn datasets, which are often limited in scale and diversity. SketchAgent uses pre-trained language models instead, which are knowledgeable about many concepts, but don’t know how to sketch. When the researchers taught language models this process, SketchAgent began to sketch diverse concepts it hadn’t explicitly trained on.

Still, Vinker and her colleagues wanted to see if SketchAgent was actively working with humans on the sketching process, or if it was working independently of its drawing partner. The team tested their system in collaboration mode, where a human and a language model work toward drawing a particular concept in tandem. Removing SketchAgent’s contributions revealed that their tool’s strokes were essential to the final drawing. In a drawing of a sailboat, for instance, removing the artificial strokes representing a mast made the overall sketch unrecognizable.

In another experiment, CSAIL and Stanford researchers plugged different multimodal language models into SketchAgent to see which could create the most recognizable sketches. Their default backbone model, Claude 3.5 Sonnet, generated the most human-like vector graphics (essentially text-based files that can be converted into high-resolution images). It outperformed models like GPT-4o and Claude 3 Opus.

“The fact that Claude 3.5 Sonnet outperformed other models like GPT-4o and Claude 3 Opus suggests that this model processes and generates visual-related information differently,” says co-author Tamar Rott Shaham.

She adds that SketchAgent could become a helpful interface for collaborating with AI models beyond standard, text-based communication. “As models advance in understanding and generating other modalities, like sketches, they open up new ways for users to express ideas and receive responses that feel more intuitive and human-like,” says Shaham. “This could significantly enrich interactions, making AI more accessible and versatile.”

While SketchAgent’s drawing prowess is promising, it can’t make professional sketches yet. It renders simple representations of concepts using stick figures and doodles, but struggles to doodle things like logos, sentences, complex creatures like unicorns and cows, and specific human figures.

At times, their model also misunderstood users’ intentions in collaborative drawings, like when SketchAgent drew a bunny with two heads. According to Vinker, this may be because the model breaks down each task into smaller steps (also called “Chain of Thought” reasoning). When working with humans, the model creates a drawing plan, potentially misinterpreting which part of that outline a human is contributing to. The researchers could possibly refine these drawing skills by training on synthetic data from diffusion models.

Additionally, SketchAgent often requires a few rounds of prompting to generate human-like doodles. In the future, the team aims to make it easier to interact and sketch with multimodal language models, including refining their interface. 

Still, the tool suggests AI could draw diverse concepts the way humans do, with step-by-step human-AI collaboration that results in more aligned final designs.

This work was supported, in part, by the U.S. National Science Foundation, a Hoffman-Yee Grant from the Stanford Institute for Human-Centered AI, the Hyundai Motor Co., the U.S. Army Research Laboratory, the Zuckerman STEM Leadership Program, and a Viterbi Fellowship.

The Hertz Foundation announced that it has awarded fellowships to eight MIT affiliates. The prestigious award provides each recipient with five years of doctoral-level research funding (up to a total of $250,000), which gives them an unusual measure of independence in their graduate work to pursue groundbreaking research.

The MIT-affiliated awardees are Matthew Caren ’25; April Qiu Cheng ’24; Arav Karighattam, who begins his PhD at the Institute this fall; Benjamin Lou ’25; Isabelle A. Quaye ’22, MNG ’24; Albert Qin ’24; Ananthan Sadagopan ’24; and Gianfranco (Franco) Yee ’24.

“Hertz Fellows embody the promise of future scientific breakthroughs, major engineering achievements and thought leadership that is vital to our future,” said Stephen Fantone, chair of the Hertz Foundation board of directors and president and CEO of Optikos Corp., in the announcement. “The newest recipients will direct research teams, serve in leadership positions in our government and take the helm of major corporations and startups that impact our communities and the world.”

In addition to funding, fellows receive access to Hertz Foundation programs throughout their lives, including events, mentoring, and networking. They join the ranks of over 1,300 former Hertz Fellows since the fellowship was established in 1963 who are leaders and scholars in a range of technology, science, and engineering fields. Former fellows have contributed to breakthroughs in such areas as advanced medical therapies, computational systems used by billions of people daily, global defense networks, and the recent launch of the James Webb Space Telescope.

This year’s MIT recipients are among a total of 19 Hertz Foundation Fellows scholars selected from across the United States.

Matthew Caren ’25 studied electrical engineering and computer science, mathematics, and music at MIT. His research focuses on computational models of how people use their voices to communicate sound at the Computer Science and Artificial Intelligence Lab (CSAIL) and interpretable real-time machine listening systems at the MIT Music Technology Lab. He spent several summers developing large language model systems and bioinformatics algorithms at Apple and a year researching expressive digital instruments at Stanford University’s Center for Computer Research in Music and Acoustics. He chaired the MIT Schwarzman College of Computing Undergraduate Advisory Group, where he led undergraduate committees on interdisciplinary computing AI and was a founding member of the MIT Voxel Lab for music and arts technology. In addition, Caren has invented novel instruments used by Grammy-winning musicians on international stages. He plans to pursue a doctorate at Stanford.

April Qiu Cheng ’24 majored in physics at MIT, graduating in just three years. Their research focused on black hole phenomenology, gravitational-wave inference, and the use of fast radio bursts as a statistical probe of large-scale structure. They received numerous awards, including an MIT Outstanding Undergraduate Research Award, the MIT Barrett Prize, the Astronaut Scholarship, and the Princeton President’s Fellowship. Cheng contributed to the physics department community by serving as vice president of advocacy for Undergraduate Women in Physics and as the undergraduate representative on the Physics Values Committee. In addition, they have participated in various science outreach programs for middle and high school students. Since graduating, they have been a Fulbright Fellow at the Max Planck Institute for Gravitational Physics, where they have been studying gravitational-wave cosmology. Cheng will begin a doctorate in astrophysics at Princeton in the fall.

Arav Karighattam was home schooled, and by age 14 had completed most of the undergraduate and graduate courses in physics and mathematics at the University of California at Davis. He graduated from Harvard University in 2024 with a bachelor’s degree in mathematics and will attend MIT to pursue a PhD, also in mathematics. Karighattam is fascinated by algebraic number theory and arithmetic geometry and seeks to understand the mysteries underlying the structure of solutions to Diophantine equations. He also wants to apply his mathematical skills to mitigating climate change and biodiversity loss. At a recent conference at MIT titled “Mordell’s Conjecture 100 Years Later,” Karighattam distinguished himself as the youngest speaker to present a paper among graduate students, postdocs, and faculty members.

Benjamin Lou ’25 graduated from MIT in May with a BS in physics and is interested in finding connections between fundamental truths of the universe. One of his research projects applies symplectic techniques to understand the nature of precision measurements using quantum states of light. Another is about geometrically unifying several theorems in quantum mechanics using the Prüfer transformation. For his work, Lou was honored with the Barry Goldwater Scholarship. Lou will pursue his doctorate at MIT, where he plans to work on unifying quantum mechanics and gravity, with an eye toward uncovering experimentally testable predictions. Living with the debilitating disease spinal muscular atrophy, which causes severe, full-body weakness and makes scratchwork unfeasible, Lou has developed a unique learning style emphasizing mental visualization. He also co-founded and helped lead the MIT Assistive Technology Club, dedicated to empowering those with disabilities using creative technologies. He is working on a robotic self-feeding device for those who cannot eat independently.

Isabelle A. Quaye ’22, MNG ’24 studied electrical engineering and computer science as an undergraduate at MIT, with a minor in economics. She was awarded competitive fellowships and scholarships from Hyundai, Intel, D. E. Shaw, and Palantir, and received the Albert G. Hill Prize, given to juniors and seniors who have maintained high academic standards and have made continued contributions to improving the quality of life for underrepresented students at MIT. While obtaining her master’s degree at MIT, she focused on theoretical computer science and systems. She is currently a software engineer at Apple, where she continues to develop frameworks that harness intelligence from data to improve systems and processes. Quaye also believes in contributing to the advancement of science and technology through teaching and has volunteered in summer programs to teach programming and informatics to high school students in the United States and Ghana.

Albert Qin ’24 majored in physics and mathematics at MIT. He also pursued an interest in biology, researching single-molecule approaches to study transcription factor diffusion in living cells and studying the cell circuits that control animal development. His dual interests have motivated him to find common ground between physics and biological fields. Inspired by his MIT undergraduate advisors, he hopes to become a teacher and mentor for aspiring young scientists. Qin is currently pursuing a PhD at Princeton University, addressing questions about the behavior of neural networks — both artificial and biological — using a variety of approaches and ideas from physics and neuroscience.

Ananthan Sadagopan ’24 is currently pursuing a doctorate in biological and biomedical science at Harvard University, focusing on chemical biology and the development of new therapeutic strategies for intractable diseases. He earned his BS at MIT in chemistry and biology in three years and led projects characterizing somatic perturbations of X chromosome inactivation in cancer, developing machine learning tools for cancer dependency prediction, using small molecules for targeted protein relocalization and creating a generalizable strategy to drug the most mutated gene in cancer (TP53). He published as the first author in top journals, such as Cell, during his undergraduate career. He also holds patents related to his work on cancer dependency prediction and drugging TP53. While at the Institute, he served as president of the Chemistry Undergraduate Association, winning both the First-Year and Senior Chemistry Achievement Awards, and was head of the events committee for the MIT Science Olympiad.

Gianfranco (Franco) Yee ’24 majored in biological engineering at MIT, conducting research in the Manalis Lab on chemical gradients in the gut microenvironment and helping to develop a novel gut-on-a-chip platform for culturing organoids under these gradients. His senior thesis extended this work to the microbiome, investigating host-microbe interactions linked to intestinal inflammation and metabolic disorders. Yee also earned a concentration in education at MIT, and is committed to increasing access to STEM resources in underserved communities. He co-founded Momentum AI, an educational outreach program that teaches computer science to high school students across Greater Boston. The inaugural program served nearly 100 students and included remote outreach efforts in Ukraine and China. Yee has also worked with MIT Amphibious Achievement and the MIT Office of Engineering Outreach Programs. He currently attends Gerstner Sloan Kettering Graduate School, where he plans to leverage the gut microbiome and immune system to develop innovative therapeutic treatments.

Former Hertz Fellows include two Nobel laureates; recipients of 11 Breakthrough Prizes and three MacArthur Foundation “genius awards;” and winners of the Turing Award, the Fields Medal, the National Medal of Technology, the National Medal of Science, and the Wall Street Journal Technology Innovation Award. In addition, 54 are members of the National Academies of Sciences, Engineering and Medicine, and 40 are fellows of the American Association for the Advancement of Science. Hertz Fellows hold over 3,000 patents, have founded more than 375 companies, and have created hundreds of thousands of science and technology jobs.

Every year, thousands of students take courses that teach them how to deploy artificial intelligence models that can help doctors diagnose disease and determine appropriate treatments. However, many of these courses omit a key element: training students to detect flaws in the training data used to develop the models.

Leo Anthony Celi, a senior research scientist at MIT’s Institute for Medical Engineering and Science, a physician at Beth Israel Deaconess Medical Center, and an associate professor at Harvard Medical School, has documented these shortcomings in a new paper and hopes to persuade course developers to teach students to more thoroughly evaluate their data before incorporating it into their models. Many previous studies have found that models trained mostly on clinical data from white males don’t work well when applied to people from other groups. Here, Celi describes the impact of such bias and how educators might address it in their teachings about AI models.

Q: How does bias get into these datasets, and how can these shortcomings be addressed?

A: Any problems in the data will be baked into any modeling of the data. In the past we have described instruments and devices that don’t work well across individuals. As one example, we found that pulse oximeters overestimate oxygen levels for people of color, because there weren’t enough people of color enrolled in the clinical trials of the devices. We remind our students that medical devices and equipment are optimized on healthy young males. They were never optimized for an 80-year-old woman with heart failure, and yet we use them for those purposes. And the FDA does not require that a device work well on this diverse of a population that we will be using it on. All they need is proof that it works on healthy subjects.

Additionally, the electronic health record system is in no shape to be used as the building blocks of AI. Those records were not designed to be a learning system, and for that reason, you have to be really careful about using electronic health records. The electronic health record system is to be replaced, but that’s not going to happen anytime soon, so we need to be smarter. We need to be more creative about using the data that we have now, no matter how bad they are, in building algorithms.

One promising avenue that we are exploring is the development of a transformer model of numeric electronic health record data, including but not limited to laboratory test results. Modeling the underlying relationship between the laboratory tests, the vital signs and the treatments can mitigate the effect of missing data as a result of social determinants of health and provider implicit biases.

Q: Why is it important for courses in AI to cover the sources of potential bias? What did you find when you analyzed such courses’ content?

A: Our course at MIT started in 2016, and at some point we realized that we were encouraging people to race to build models that are overfitted to some statistical measure of model performance, when in fact the data that we’re using is rife with problems that people are not aware of. At that time, we were wondering: How common is this problem?

Our suspicion was that if you looked at the courses where the syllabus is available online, or the online courses, that none of them even bothers to tell the students that they should be paranoid about the data. And true enough, when we looked at the different online courses, it’s all about building the model. How do you build the model? How do you visualize the data? We found that of 11 courses we reviewed, only five included sections on bias in datasets, and only two contained any significant discussion of bias.

That said, we cannot discount the value of these courses. I’ve heard lots of stories where people self-study based on these online courses, but at the same time, given how influential they are, how impactful they are, we need to really double down on requiring them to teach the right skillsets, as more and more people are drawn to this AI multiverse. It’s important for people to really equip themselves with the agency to be able to work with AI. We’re hoping that this paper will shine a spotlight on this huge gap in the way we teach AI now to our students.

Q: What kind of content should course developers be incorporating?

A: One, giving them a checklist of questions in the beginning. Where did this data came from? Who were the observers? Who were the doctors and nurses who collected the data? And then learn a little bit about the landscape of those institutions. If it’s an ICU database, they need to ask who makes it to the ICU, and who doesn’t make it to the ICU, because that already introduces a sampling selection bias. If all the minority patients don’t even get admitted to the ICU because they cannot reach the ICU in time, then the models are not going to work for them. Truly, to me, 50 percent of the course content should really be understanding the data, if not more, because the modeling itself is easy once you understand the data.

Since 2014, the MIT Critical Data consortium has been organizing datathons (data “hackathons”) around the world. At these gatherings, doctors, nurses, other health care workers, and data scientists get together to comb through databases and try to examine health and disease in the local context. Textbooks and journal papers present diseases based on observations and trials involving a narrow demographic typically from countries with resources for research. 

Our main objective now, what we want to teach them, is critical thinking skills. And the main ingredient for critical thinking is bringing together people with different backgrounds.

You cannot teach critical thinking in a room full of CEOs or in a room full of doctors. The environment is just not there. When we have datathons, we don’t even have to teach them how do you do critical thinking. As soon as you bring the right mix of people — and it’s not just coming from different backgrounds but from different generations — you don’t even have to tell them how to think critically. It just happens. The environment is right for that kind of thinking. So, we now tell our participants and our students, please, please do not start building any model unless you truly understand how the data came about, which patients made it into the database, what devices were used to measure, and are those devices consistently accurate across individuals?

When we have events around the world, we encourage them to look for data sets that are local, so that they are relevant. There’s resistance because they know that they will discover how bad their data sets are. We say that that’s fine. This is how you fix that. If you don’t know how bad they are, you’re going to continue collecting them in a very bad manner and they’re useless. You have to acknowledge that you’re not going to get it right the first time, and that’s perfectly fine. MIMIC (the Medical Information Marked for Intensive Care database built at Beth Israel Deaconess Medical Center) took a decade before we had a decent schema, and we only have a decent schema because people were telling us how bad MIMIC was.

We may not have the answers to all of these questions, but we can evoke something in people that helps them realize that there are so many problems in the data. I’m always thrilled to look at the blog posts from people who attended a datathon, who say that their world has changed. Now they’re more excited about the field because they realize the immense potential, but also the immense risk of harm if they don’t do this correctly.