The Center for Responsible AI
Where data science and social science unite for the common good
The evolution and wider use of artificial intelligence (AI) is creating an ethical crisis in computer science that strikes at the core of who we are. Institute Associate Professor of Computer Science and Engineering Julia Stoyanovich, who co-founded the Center for Responsible AI and is a member of the Center for Data Science, explains that when we ask AI to move beyond games, like chess or Go, in which the rules are the same irrespective of a player’s gender, race, or abilities, and look for it to perform tasks that allocate resources or predict social outcomes — such as deciding who gets a job or a loan or which sidewalks in a city should be fixed first — we quickly discover that embedded in the data are social, political, and cultural biases that distort results, because the human programmers inputting the data can have conscious and unconscious biases of their own. She’s dedicated to rooting out that bias and training a new generation of engineers about the social implications of the technology they build. AI can’t really be used effectively unless it’s also used responsibly, she says — and good, unbiased data is key
Stoyanovich and her colleagues aren’t focused only on NYU Tandon students or on the technical and computer science aspects of Responsible AI, either. The Center’s other efforts span policy and public education initiatives through collaboration at the hyper-local and global level, including:
- Mounting a transatlantic panel series open to the public to explore how AI works, the impact it has, the pitfalls it can pose, and the framework needed to ensure its responsible use across sectors from medicine to finance to art.
- Working with the Queens Public library and P2PU to develop a free, five-week course and related public video content that gives citizens a “101” on AI, and empowers them to advocate for policies that prevent its abuses.
- Creating a widely available (and highly engaging) comic book series, “Mirror, Mirror,” with the help of the Center’s artist-in-residence, Falaah Arif Khan, to spread the word about the importance of using data wisely.
Making it an award-winning year
The National Science Foundation (NSF) CAREER Awards are given in support of early-career faculty members who have the potential to serve as academic role models and to lead advances in the mission of their department or organization. This past year, two more researchers joined the over 50% of NYU Tandon’s junior faculty members who hold CAREER Awards or similar young investigator honors. They’re both tackling the task of creating better, more efficient AI systems, each from their own angle.
- Deep learning (DL) technology is now used increasingly in physics, medicine, and chemistry, and for applications like image, speech, and video recognition; image segmentation; and natural language processing. Assistant Professor of Electrical and Computer Engineering Anna Choromanska was honored for research focusing on new, more efficient ways of training DL models — a process that typically consumes resources, time, and money, compromising the progress of public and private sectors that rely on DL, and limiting its deployment in new applications. Choromanska’s project aims to overcome these issues by describing universal properties of DL systems that hold across a variety of DL models and data sets, thus making possible a new generation of DL training strategies that are efficient, accurate, and scalable.
- Assistant Professor of Computer Science and Engineering Christopher Musco was honored for his focus on new ways of processing and analyzing massive data sets from such sources as scientific simulations, urban data, and web content. Typically, datadriven discovery and decision making for science, engineering, and industry requires enormous computational effort, making the process both costly and time consuming. The goal of Musco’s work is to develop new algorithms capable of efficiently processing the world’s largest datasets without the need for the world’s largest supercomputers. To achieve this, Musco and his team are employing a powerful algorithmic technique known as “matrix sketching,” the purpose of which is to quickly compress a large dataset (represented as a matrix of numbers) down to its most essential information by eliminating redundancy and noise.
Additionally, this year, Yann LeCun — Professor of Electrical and Computer Engineering, Silver Professor of Computer Science at Courant, and Professor at the NYU Center for Data Science and NYU Center for Neural Science — was elected to the prestigious National Academy of Sciences. An ACM Turing Award Laureate and a member of the National Academy of Engineering, LeCun is widely celebrated for his seminal breakthroughs in artificial intelligence, specifically deep learning and convolutional neural networks — the very foundation of modern computer vision, speech recognition, speech synthesis, image synthesis, and natural language processing. Without his work, it’s safe to say, physicians might not be using AI to accurately diagnose disease, autonomous vehicles might not be a reality, and Siri and Alexa might not be so responsive to your voice commands.
We make data work across disciplines
Maurizio Porfiri is an NYU Tandon Institute Professor with appointments in the Departments of Mechanical and Aerospace Engineering, Biomedical Engineering, and Civil and Urban Engineering, and as a core faculty member at the Center for Urban Science and Progress (CUSP). Even with the premise that it takes multidisciplinary expertise to solve important problems, it’s an impressive array for just one engineer. Perhaps even more impressive is the wide variety of research that Porfiri — who is currently principal investigator or co-principal investigator on nine National Science Foundation-funded projects — undertakes. Named to Popular Science’s “Brilliant 10” list for his work with biologically inspired aquatic robots, he has also delved into issues involving gun violence, smart materials, and COVID-19, among other topics that might not seem to have much in common on the surface. Dig a little deeper, Porfiri asserts, and the commonality becomes clearer.
How do animals behave? How do people reach consensus? How does a pandemic spread? What happens when particles collide? These all involve dynamic, complex systems, and that’s the underlying layer, he explains. Once you decide upon the questions you want to answer — whatever those questions may be — you need to study the interactions taking place. It’s easy to see, then, why Porfiri’s lab at NYU Tandon is called the Dynamical Systems Laboratory.
Once Porfiri — the recipient of Tandon’s Excellence in Research Award — begins exploring an issue, his work involves analyzing time-series data to come up with inferences, modeling systems in order to extrapolate information, and examining the control problems that emerge. His novel methods for measuring spatial dependencies makes even small data act BIG.
Among some of the NSF-funded projects he has undertaken at the intersections of data and health, materials, and social behavior:
- Engaging stroke patients and others with upper-limb weakness in interesting citizen-science initiatives that can be integrated into their otherwise tedious physical therapy regimes with the use of haptic joysticks to increase compliance and improve the experience of telerehabilitation
- Building a mathematical model specifically for the city’s unique social and transportation structures using data from New Rochelle — the New York suburb first seriously afflicted by COVID-19 — to help health and government leaders make smart, up-to-date testing and contact-tracing decisions
- Disentangling causes from effects in coordinated swimming among large schools of fish in order to contribute a mathematically principled, experimentally grounded approach to quantifying information flow in groups of dynamical systems
- Determining how the use of artificially intelligent tools in the medical field affects physician work practices and the expert-patient relationship
- Advancing the understanding of the sometimes surprising causal relationships among potentially contributing factors to firearm-related harm, such as prevalence of gun purchases, state legislation, media exposure, and perceptions of firearm safety
- Predicting how the cascading effects of migration from flooded areas of Bangladesh will ultimately affect 1.3 million people across the country by 2050 — work that has implications for coastal populations worldwide
- Studying the remarkable structural properties of the Venus flower basket sponge (E. aspergillum) to gain insight into how the organism’s latticework of holes and ridges influences the hydrodynamics of seawater — work that could lead to advanced designs for buildings, bridges, marine vehicles, and anything that must respond safely to forces imposed by the flow of air or water