In her "free time" Dr. Ming explores the possible applications for "augmented cognition" through technologies like Google Glass and is in the process of developing a predictive model of diabetes, that would allow patients to better manage blood glucose levels. Dr. Ming also sits on the board of Our Family Coalition, a group focused on supporting LGBT families and is an advocate for the issues of both LGBT inclusion and gender equality in technology. Dr. Ming remains a visiting scholar at UC Berkeley’s Redwood Center for Theoretical Neuroscience and was a junior fellow at Stanford’s Mind, Brain & Computation Center. Dr. Ming holds a Ph.D. from Carnegie Mellon.
EXCLUSIVE NEW LEARNING TIMES INTERVIEW
Question: How did your educational trajectory (background) affect your current work?
Answer: My education was less a trajectory and more a walkabout. The most foundational experiences came when I was least successful. In elementary and high school I was a terrible student, often taking F's and D's into the last weeks of class from a profound unwillingness to do the work. My sophomore year in high school I was kicked out of all of my honors classes for the rather minor infraction of failing them. Everyone became obsessed with worries about my transcript and my chances of attending college. Did they actually think that I would find a passion for chemistry homework outside the honors class? That I would nap less frequently during lectures? With a forged letter from my parents in hand, I reversed these decisions and spent the rest of the year daydreaming about limiting reagents and statistical mechanics while completing just enough lab reports to get by. I loved and still love science. I didn't care about my grades, shortsighted though that may have been, but I was passionate about understanding. The inability of my otherwise excellent teachers to distinguish and appropriately support those qualities has had a huge influence on my learning sciences research and my business philosophy.
Of course, my teachers weren't completely off. Standardized testing lifted me into UC San Diego, but I was soon paying tuition on classes I was not attending and accumulating a powerful collection of awful grades. I eventually dropped out. In nearly any plausible story, that would have been the end of my academic career. I worked (and didn't) for several years. Given what I do in life now as theoretical neuroscientist and educational technologist, it seems like such a waste that my potential was never realized over 15 years of school. That this all could have ended then. Today everything I do is driven by thoughts of how many life stories start like mine but never have the post-intermission transformation I experienced.
Circumstance and boundless parental support allowed me to finish my undergrad seven years later. For reasons that largely elude me, it was all so easy. My passion for understanding was finally paired with drive to have an impact. I chose cognitive science over economics on a coin toss and completed my undergrad in a year. During that time I discovered two new passions as a research assistant at the Machine Perception Lab: machine learning and educational technology. We were involved in a CIA-sponsored competition to read facial expressions off of free-form video. I began to think of how this sort of technology could monitor learners, adjusting computer-based learning (still in it's childhood) to the attentional, emotional and cognitive states of the learner. When I moved on to graduate school at Carnegie Mellon I saw these same ideas realized in the cognitive models and cognitive tutors of groups like Carnegie Learning. While I developed algorithms with an eye on cognitive neuroprosthetics for my dissertation, I continually daydreamed about bringing the same sorts of unstructured data modeling to cognitive tutoring systems.
Question: What professional experiences have been most formative to your current work?
Answer: My first startup, Augniscient, was meant to be an EdTech company. As my co-founders and I went out to raise money, however, we quickly found that investor enthusiasm for our technology did not extend to enthusiasm for education (and definitely not for the education business market). Seemingly inevitably, we veered out of education and applied our "cognitive models from unstructured data" to less noble goals, like modeling social gamers. Though it was a mistake to abandon my passion, I learned a fundamental lesson about building products: they must be in harmony with their users. EdTech is filled with "brilliant" products, which fail to achieve their potential because they do not treat instructors and learners as partners. Every education conference I've attended has been filled with the big dreams of researchers that treat teachers as an afterthought (at best) and also numerous educators whose interests remain unalterably focused on their immediate frustrations and resent "outside interference." I've seen this play out with nurses and doctors, recruiters, engineers, and so many others. The customer focus that drives the product design and marketing world is too often missing from high tech research and product design for education, and it plays a large role in the cyclic boom and bust of educational technology. Hybrid and flipped classrooms, MOOCs, personalized learning... none of them will live up to their potential if they are antagonistic to actual educators, focused only on "disrupting" and not supporting.
Question: How do you hope your work will change the learning landscape?
Answer: Though we have many short- and medium-term goals at Socos, our true hopes reside in our outsized ambitions: personalized learning for all, ubiquitous yet non-intrusive technology, and no more tests. The vision of Socos is that teachers run their classrooms however they like using whatever curriculum they and their school deems appropriate. Our hard job is to capture as much data from that classroom as possible and make meaning from it, returning contextual insight and recommendations to both the learner and the instructor. By inferring the conceptual understanding of students, using everything from homework to questions to free-form discussions, we have been able to predict their performance in their courses and even their specific performance on individual assessments. I can think of nothing better than technology that gets out of the way of learners while also obviating the need for tests. If someday we can reliably predictively assess students in real time using all of their classroom activities, then why waste time with biased, high-stakes tests. So much of educational technology is now focused on taking up class time and delivering explicit assessment; our more human approach ironically requires much more technological and computational sophistication, but I can think of little else I would be proud to accomplish.
Question:What broad trends do you think will have the most impact on learning in the years ahead?
Answer: The major trends in edtech that we all hear about rarely excite me. I met recently with the director of a major European education foundation. He had been conducting a tour of the US, including the White House and Dept. of Education, the Gates Foundation, Microsoft, Google, Pearson and so on. He came to learn about the cutting edge of educational technology, the latest new thing. What he found was that while everyone here was focused on new and better approaches, they were all talking about the exact same new and better approaches. This sort of lock step is almost the definition of a boom economy and the bust, while not inevitable, might soon follow if we don't diversify and maintain a constructively skeptical mind.
The results of MOOCs have been quite mixed so far. Our enthusiasm for them is understandable, but has anyone really studied the educational impact of massive scale social network effects. A single-lecture-fits-all approach might scale well, but truly personalized learning will require a marketplace for a diversity of learning experiences and technology that can serve up the right experience for the right learner and then assess its impact.
Similarly, massive scale discussions rely on very small numbers of peer-tutors; this is part of the beautiful scaling properties of social networks. But what happens when that role of tutoring peers is largely dedicated to the tiniest handful of "super-star" students. The experience of learning by teaching might became a rare experience in MOOC forums. Combating this, various technologies that can spontaneously aggregate small peer networks from enormous "classes" will become increasingly important.
Finally, I am very broadly excited about the potential of unstructured data in education. From video and audio capture and automated analysis to processing text and even notes and images for meaning, the trend away from structured data can have tremendous impact. I've done extensive modeling of Facebook data, for example, and a single free-form post is worth vastly more than a "like." In the context of education, the idea the we want essays rather than multiple choice hardly needs arguing, but the reality of the time needed to consume and assess that work severely undermines it's ideal value. The power to turn free-form projects and writings by students into real time insight can transform the economics of the classroom.
Question:What are you currently working on & what is your next big project?
Answer: While I am ridiculously over-extended -- somehow believing I can follow all of my passions and be a mom too -- I have three principal projects right now with educational relevance. First, at Gild my job is to take a vision of holistic "assessment" of students and apply it to the world of professional reputation and recruiting. The biases and waste of the recruiting world parallel that of education. A bad day for the interview or interviewee means a missed job opportunity. Over-valued signals like the school you attended dominate hiring descriptions. My research suggests that for every highly credentialed candidate for a given job opening, there are 10-100 equally qualified candidates without similar "Ivy League" credentials. But in a risk-averse job market, signals such as one's social network and academic history play a hugely outsized role in recruiting decisions. Gild uses a "big data" approach to predict best fits between a database of six million professional software developers and job openings. Our guiding principle is meritocracy; no talent should ever be wasted because a candidate lacks the "right" degree or didn't know the "right" people or simply because they were quiet and self-effacing. Our next steps are to use a large probabilistic graphical model to map professional skills and recommend new skills for candidates to learn which will maximize their chance of landing a dream job.
Of course, the world can only be so meritocratic if we don't begin earlier in the education process. At Socos we have begun supporting the White House Office of Science and Technology in their push to increase the graduation rates in STEM majors from underrepresented group. These students might be performing equitably with their peers but have much higher dropout rates throughout their academic careers. While we are looking at building models to predict disengagement in students and flag them for intervention, fundamentally I believe these are meta-cognitive issues, such as perspective taking and grit. Importantly, I also deeply believe these are changeable. We can change students' skills for self-reflection. We can build in them a belief that their work will produce successful outcomes. Designing metacognitive training into personalized learning systems is a long-term investment, but I will always choose a student with grit over one with brittle knowledge.
Finally, I am excited about a new project we have begun recently in partnership with Pathbrite, a company with it's roots in ePortfolios and a vision to extend far beyond. The opportunity to pair our cognitive modeling technology with their platform has been one of our first opportunities to go beyond subsets of student data and truly build a career-spanning model. By powering their predictive learning analytics we can provide a sense of context and perspective to students, peer matching in online learning, actionable intervention recommendations to instructors, and even global feedback to administrators to assess changes on a daily basis.
Image: Courtesy Vivienne Ming