The Pioneer: Prioritizing People, Augmenting Human Intelligence
Pattie Maes is the Germeshausen Professor of Media Arts and Sciences and the Director of the Fluid Interfaces research group at the MIT Media Lab. Her research lies at the intersection of human-computer interaction and artificial intelligence with a focus on applications in health, wellbeing, and learning.
In 1988 Pattie Maes departed Brussels, Belgium to do postdoctoral research at MIT’s Artificial Intelligence Laboratory (now the Computer Science and Artificial Intelligence Laboratory). Here, she worked side-by-side with luminaries in the field like Marvin Minsky, and was particularly drawn to the work of Rodney Brooks.
The consummate contrarian, Brooks eschewed traditional thinking on AI and robotics, which presupposed that a machine must first be fed and then comprehend a complete picture of the environment in which it would be deployed. Brooks rethought the paradigm, introducing a bottom-up, behavior-based approach. Rather than trying to teach robots everything about the world before they entered it, he built robots that could perform basic tasks and learn from their environments using rudimentary vision systems and very little computation.
Asked to explain her affinity for Brooks’ line of research, Maes says, “I admired that Brooks was disrupting robotics research by pursuing a radically different, very simple approach that was motivated by how animals demonstrate seemingly intelligent behavior while just following innate reflexive behaviors.”
Maes helped bring Brooks’ machines to life in the shape of his inspiration: insects, which he noted were short on neurons but capable of deftly interacting with their environments. Her contributions were impressive enough to earn her a visiting professorship at the Institute.
A year later, Maes joined the faculty at the MIT Media Lab, and in 1991 she founded the Software Agents research group. Recalling the transition, Maes says, “Rather than trying to model intelligence in machines, I was more interested in helping people become smarter. The Media Lab tends to emphasize the human and societal impact of emerging technologies; here we try to think hard about how cutting-edge technologies might improve people's lives.”
Rather than trying to model intelligence in machines, I was more interested in helping people become smarter
Heading up her first research group, Maes began developing intelligent software agents to help people navigate their lives, which were increasingly entwined with the abundance and complexity of the burgeoning World Wide Web. One of her former junior researchers, John Maeda (current VP of Design and Artificial Intelligence at Microsoft), recently called Maes “the pioneer of the agents revolution.”
Her early work at the Media Lab played a significant role in the expansion of e-commerce and was a foundational step in the development of social media networks. As an outgrowth of her work with the Software Agents group, Maes co-founded Firefly Networks, Inc., an MIT-connected start-up with a platform that enabled personalized recommendations and interactions between websites and their visitors while opening the door to the notion of online communities. This was in 1994, well before Mark Zuckerberg had even dreamed up Facebook’s lurid predecessor, Facesmash. Maes also co-founded Open Ratings, Inc., which helped B2B companies track and identify performance patterns of suppliers while predicting and improving decision-making processes.
In fact, Maes’ research has spawned numerous entrepreneurial endeavors: “Not because we planned from day one to start a company,” she explains, “but because we came up with a novel idea, there was a need for it, and we wanted to make it deployable.” Pattie Maes and her ilk at MIT are a big part of the reason Kendall Square evolved into an innovation hub with global impact. “With MIT being so motivated to solve real-world problems in innovative ways, it naturally leads to a lot of entrepreneurship and industry collaboration.”
Today, Maes leads the Fluid Interfaces group at the Media Lab, where she devotes much of her time to the topic of cognitive enhancement, or how wearable, immersive, and brain-computer interface systems can actively assist people with issues including memory, attention, learning, decision making, communication, wellbeing, and sleep.
Her people-focused approach to technology is reflected in projects that explore, among other things, how the devices we carry with us might help an aging population live independently in their homes for longer. For example, she and her research group have developed a wearable, voice-based memory assistant called MemPal. The device logs user actions in real time without storing image data, thereby preserving privacy. The prototype is essentially a camera that a user wears around their neck—though Maes notes that it could potentially be built into glasses.
The system takes constant pictures of what the user does with their hands, interprets those pictures with a computer vision multimodal AI model, processes them locally to maintain privacy, and creates a description of the actions. Rather than saving the pictures, it maintains the descriptions, creating a log of actions taken by the individual throughout the day. Misplace your keys, medication, and reading glasses: ask MemPal. The system also allows users or their caregivers to input “if-then” rules to set reminders. For example, MemPal can prompt you to take a particular medication (or remind you that you already took your medication earlier in the day), lock your door at a certain time, or check to make sure you’ve turned off the stove.
Maes and her team at the Fluid Interfaces research group are also working to develop AI systems that encourage users to develop critical thinking skills. We already rely on AI systems like Siri and Alexa, and the advent of generative AI systems like ChatGPT has ushered in what some believe is a new era on the road to AI ubiquity. Ask Alexa, Siri, or ChatGPT a question and they spit out an answer. But these systems are prone to presenting fallacy as fact in a manner that seems logical. Furthermore, research has demonstrated that over-reliance on AI degrades critical reasoning.
Maes and the Media Lab, in conjunction with colleagues at Columbia University, recently conducted a study with 204 people to explore the novel idea of AI-framed Questioning, and how it impacts a user’s ability to discern the logical validity of socially divisive statements. Their findings were presented in a peer-reviewed paper, “Don’t Just Tell Me, Ask Me: AI Systems that Intelligently Frame Explanations as Questions Improve Human Logical Discernment Accuracy over Causal AI Explanations.”
As AI development has increasingly shifted from academia to industry, Maes continues to focus on designing AI that benefits human beings. She believes that academia’s role at this stage is to advocate for the user and that we need to design integration of AI into our lives carefully to ensure that people will be better off in an era of pervasive AI use. “We create these tools that are ultimately intended to improve our lives and our ability to control our world—it’s what sets people apart from other animals. But I think we can do a lot better with digital technology now and in the future,” she says. “We need to design AI to maximize human agency and flourishing.”