A Harvard Dropout’s Plan to Fix College Admissions With Video Games
(Bloomberg Businessweek) -- In statistical terms, this is the golden age of American higher education. More than 1 in 3 Americans has at least a bachelor’s degree, the most ever. Almost 70 percent of high school seniors graduating this spring will go to college in the fall, compared with about half during the mid-1970s.
The benefits of all that education, however, are highly uneven. The campuses of elite colleges remain disproportionately populated by the rich. At selective universities—ones that admit fewer than half of applicants—3 out of 4 students come from the richest quartile of families. According to Opportunity Insights, a research group led by Harvard economist Raj Chetty, children from families in the top 1 percent of income distribution are 77 times more likely to attend an Ivy-plus school—Ivy League plus Duke, MIT, Stanford, and the University of Chicago—than those from the bottom 20 percent.
Put another way: Higher education in America is a racket.
On March 12, just as millions of nervous 12th graders were about to find out where they’ll be spending the next four years, the FBI announced the arrests of 50 people—including two Hollywood actresses, the co-chairman of a prominent global law firm, and the former chief executive officer of Pimco—in a scandal that exposed a culture of fraud at the heart of the college-admissions process. The FBI investigation, called Operation Varsity Blues, found that wealthy Americans are no longer buying spots for their children the old-fashioned way, with seven-figure donations, or finagling them through family legacies and social connections; they’re actively conspiring with criminal fixers, coaches, and college officials to cheat, lie, and bribe their way in, too. As Andrew Lelling, the U.S. attorney for the District of Massachusetts, put it in a press conference, “The case is about the widening corruption of elite college admissions through the steady application of wealth combined with fraud.”
Getting a college degree has long been integral to the mythic promise of American opportunity. Yet for millions, it’s become exactly that, a myth—and a very expensive myth at that. The average student leaves school carrying $30,000 in debt. More than 40 percent of students who enter college fail to earn a degree within six years, and many of them wind up in the workforce lacking the credentials and practical skills required to get ahead. The U.S. system of higher education isn’t the main source of economic inequality in America. But it’s almost certainly making things worse.
A 27-year-old entrepreneur who dropped out of Harvard, Rebecca Kantar, has a plan to fix it. The American obsession with college admissions, she says, benefits the wealthiest and highest-achieving students, while leaving the vast majority ill-qualified for the jobs of the future. She says a big part of the problem is the avalanche of standardized tests students take from kindergarten through high school, a $10 billion industry that drives much of what’s taught in the classroom. At the top of the pyramid sit the SAT and ACT, the generations-old multiple-choice tests that still help to determine who gains entry to top colleges and universities.
In Kantar’s view, those tests reveal little, if anything, about whether a student has the cognitive skills essential for success beyond college. As the FBI’s investigation reveals, the SAT and ACT can also be gamed: The mastermind of the scheme had parents petition for their kids to take the tests in largely unsupervised settings, then submitted fake scores on their behalf. “The system has coalesced around things that work for at most 30 percent or so of kids,” Kantar says. “They don’t work for the rest.”
Kantar is the founder of Imbellus Inc., a startup in Los Angeles that aims to reinvent testing and, in the process, challenge the received wisdom about what students are expected to learn. The digital assessments Imbellus has developed resemble video games. Placing users in a simulated natural environment, they present test takers with a series of tasks, all the while capturing the decision-making process used to complete them. And because each simulation delivers a unique user experience, they’re intended to be cheatproof.
Since coming up with the idea for the company four years ago, Kantar has raised more than $23.5 million in funding, hired a dozen Ph.D.s, and persuaded the consulting giant McKinsey and Co., and a few others, to work with Imbellus to create game-based tests that measure prospective employees’ decision-making, adaptability, and critical thinking. She argues that by harnessing advances in computing power, artificial intelligence, and data science, her assessments can deliver a quantitative picture of how a worker thinks. But her goals go beyond providing corporate America with a sharper hiring tool: “That is a problem. We do try to address it,” she says. “But it’s not the problem.”
Kantar’s premise is that huge numbers of American students lack the competencies required in an age of automation, because the country’s schools are failing to provide them with the proper preparation. “It’s not an aptitude problem—it’s a practice problem. They aren’t practicing the right kind of thinking.” In her view, expanding economic opportunity is impossible without transforming the way big institutions test for and evaluate student potential. “If you want to change the default settings in the system,” she says, “you’ve got to start at the top.”
Kantar stands 5-foot-4, with straight brown hair that falls almost to her waist. She speaks in bursts of increasing velocity, as if she’s in a hurry. She grew up in Newton, Mass., an affluent suburb of Boston that churns out high achievers. Kantar’s parents own a construction company that specializes in environmentally sustainable development, which they run out of the bottom floor of their elegant five-bedroom home.
From the start, Kantar’s interests were mostly extracurricular. She sewed her own clothes, played the trumpet, took up stained glass and pottery, and sold her handmade creations at friends’ bar mitzvahs. In junior high she started taking Mandarin—she slipped worksheets into waterproof folders so she could practice in the shower—and earned a grant from the city council to stage a Chinese-language production of Cinderella. In high school, Kantar helped create Minga, a student-run charity dedicated to raising awareness about the child sex trade. The group raised $100,000 in five years, with Kantar leading a half-dozen other teens on a 40-city speaking tour. At 18 she gave her first TED Talk. “I really enjoyed thinking about how complex the problem was, how many different pieces were involved,” she says, over a plate of pasta near her parents’ home in Newton. “That’s what I learned about myself: I had a propensity for thinking about complex systems dynamics.”
But she had little patience for formal education. “She didn’t enjoy her classes,” says her father, Jonathan, “but she did take them seriously. And she’s very competitive.” She devoted considerable time to tutoring her younger brother, Josh, who has an undiagnosed developmental disability. “I’ve always believed Josh is capable of more things than most people would assume, if he were taught those things in the right way. And as a little kid I put in a lot of energy figuring out what is the right way.”
As a high school senior, she was accepted to Harvard, Princeton, and Yale, and was offered a full scholarship to Duke. She chose Harvard. “I wouldn’t say I was excited when I started, but I recognized why it was important to try it. After my first semester, I was like, ‘Yeah, no. Done.’ ” Her parents allowed her to move back home but insisted she stay in school.
By that time, she’d written a business plan and gotten seed funding for her first company, BrightCo, a network of socially minded young entrepreneurs who provided brand advice to large corporations. When Harvard rejected her proposal to create an interdisciplinary major called Leadership and Organizations, she decided to drop out. “For us, it was drilled into our heads that a four-year degree from a high-quality university was your ticket to success,” says her mother, Ruth. “But there was no stopping her. It wasn’t just ‘I want to quit because this is too hard’ or ‘I don’t want to do this anymore.’ She just finally convinced us that this was the right thing for her.”
Kantar sold BrightCo to the expert-advisory company Gerson Lehrman Group Inc., moved to New York, and began plotting to disrupt the U.S. education system. She initially thought of designing an alternative college curriculum focused on work-oriented, project-based learning and selling it to elite universities. “It was the most fabulous nonstarter I’d ever encountered,” says Jeff Brenzel, a former dean of undergraduate admissions at Yale, who met Kantar in 2014. “These schools were not going to outsource half of their undergraduate program to Rebecca Kantar.” She decided to shift her focus away from what college students learned on campus to how they got there.
In 2018 more than 2 million students took the SAT and 1.9 million took the ACT. Kantar argues that these standardized tests exacerbate inequality in two ways. Most obviously, they give an advantage to wealthier students who can pay for tutoring and test-preparation courses—or for fake scores, as was the case in Operation Varsity Blues. The other effect is more pernicious. At least as early as high school, classroom instruction is geared to boost kids’ performance on college-admissions tests. But those tests measure what students already know, not the qualities employers and economists say they need to thrive in the future: problem-solving, critical reasoning, collaboration, creativity, empathy. “I’m interested in introducing tests that, hopefully, impose standards that shape curriculum in a way that’s better than tests that are shaping it now,” Kantar says. “It’s less about who does and does not get into Harvard. Yeah, that matters. It’s a topic. But it’s secondary to changing the default settings of the education-to-employment system so that it works better for all kids.”
Kantar advocates project-based learning, rather than content mastery, and pushing students to apply their knowledge outside of the classroom. She doesn’t necessarily support a German-style system, in which a student is placed in either a baccalaureate or vocational track before entering high school. Rather, she’s arguing for creating standards that force schools to prioritize teaching students how to think for themselves. “The nature of human intelligence required in even the most elite jobs is very different than what it was 30 years ago. If you look at any job across the spectrum, whether it’s a blue-collar job or a white-collar job, the thinking skills involved are getting harder, not easier,” she says. “My point is not to reconnect education to work so that we pump kids out of college into factory jobs. It’s about schools’ focusing a little less on one specific set of information and a little bit more on the thinking faculties needed to be an adult.”
Brenzel, the former Yale admissions dean, says the SAT and ACT “have become essentially what Rebecca believes: a measure of an important but very narrowly defined cognitive skill set. But there’s been no alternative.” The College Board—the nonprofit consortium of schools that owns the SAT and which generated more than $1 billion in revenue in 2017—has over the years introduced changes to the test, in response to accusations of bias in its questions. But it’s largely resisted altering the basic format: a timed, multiple-choice test of math and literacy skills, administered in a proctored setting on a scheduled date.
The SAT remains a useful tool for predicting whether students can handle their first year of college, says Jack Buckley, a former senior vice president of the College Board, who joined Imbellus in January as its president and chief scientist. But in the wake of Operation Varsity Blues, the folly of using such an easily manipulated test for high-stakes evaluation has never been more apparent. “There are a lot of people hungry for the system to change,” Buckley says. “But the College Board is a membership organization where the key members are institutions of higher education. They can’t get too far ahead of what higher ed wants. And getting higher ed to change is hard.” In an emailed statement, College Board spokesman Zachary Goldberg says its research demonstrates that SAT scores “improve the ability to predict college performance above high school GPA alone.” As for reports of cheating, Goldberg says “The College Board has significantly increased our test security efforts and resources in recent years to combat theft and organized cheating.” These steps include “producing more test content, banning and collecting cell phones, employing lock boxes, conducting data-driven analyses of test taker behaviors, and enhancing security measures at test centers.”
In 2016, Kantar moved to Los Angeles with the beginning of the idea that would become Imbellus. She studied Tesla Inc.’s impact on the car industry. “What would it take to disrupt the big tests in this way? My answer was: a lot of time, a lot of people, and a lot of money. And I felt that, you know, those are things that you can acquire.” Her first hire was Richard Wainess, a 65-year-old educational psychologist at UCLA’s National Center for Research on Evaluation, Standards, and Student Testing. A veteran of the entertainment and gaming industries, Wainess focused his research on how to create a reliable test that’s as engaging and immersive as a video game. “Until now, no one had done it successfully,” he says. “They either sacrificed game for test or they sacrificed test for game.” After his first meeting with Kantar, Wainess told his wife, “I’ve just met Bill Gates, Mark Zuckerberg, and Steve Jobs.”
They began assembling an eclectic team: psychometricians, 3D animators, video game designers, theoretical physicists. Kantar named the company Imbellus, after a kind of betta fish that doesn’t swim in schools. She decided its games would be ecologically themed, the idea being that any test taker, regardless of background, could grasp the laws of the natural world. To prevent cheating, Kantar insisted that no two test takers experience the game in the same way, which required devising tens of thousands of variations of each scenario, each with a uniform level of difficulty. It took a year to create the first prototype. “Very few people understand what a hard-science problem building a high-stakes assessment is,” says Kantar, who compares it to drug development. “Each one of these costs many millions of dollars to build, and we’ve only raised $25 million. So you don’t get that many swings at bat.”
Kantar realized that finding a receptive audience for her tests also required challenging employers’ fixation on undergraduate credentials. It makes little financial sense for a student who wants a career as, say, a high-end electrician or coder to go to a four-year college and amass hundreds of thousands in debt. The trouble is that companies “overindex” for bachelor’s degrees when making hiring decisions, even for positions that don’t require a B.A. That’s in part because they lack the time and resources to assess a candidate’s skills in any other way. If Imbellus could show that its approach to testing could help companies identify talent they would otherwise have missed, it would be easier to persuade the educational establishment to start giving the assessment to students. Kantar just needed a partner willing to allow its recruitment process to be used as a laboratory for her product. She found one: McKinsey.
Over the past 18 months, about 5,000 McKinsey job candidates in 20 countries have taken the Imbellus assessment, alongside the company’s traditional multiple-choice exam. For about half of those recruits, their performance on the Imbellus test is a factor in determining whether they receive an in-person interview. Presiyana Karastoyanova, a 22-year-old Bulgarian who’s studying for a master’s in finance at Imperial College London, took both tests after applying for a job at McKinsey in November. She prepared extensively for the paper-and-pencil test, but had no idea what to expect when she was handed a laptop with the Imbellus simulation on it. “I thought I’d just improvise.”
After she logged in, an animation of a lush tropical island appeared on the screen. She was presented with three different scenarios, each one depicting a natural environment under stress. In one part of the island, she had to devise a plan to save native species from an impending natural disaster; in another, she was instructed to create a coral reef ecosystem that could withstand elevated levels of toxicity. She navigated crystalline scenes of wild animals, fish, plants, mountains, and ocean waters. “I became totally immersed. I forgot the world around me,” she says. Karastoyanova completed the tasks in half an hour and felt a rush of adrenaline, as if she’d just played a video game. (She made it through two rounds of interviews and was offered a job in December.)
McKinsey plans to double the number of candidates taking the Imbellus assessment by the end of 2019 and even begin allowing some to take it on their home computer. Keith McNulty, the company’s head of people analytics and measurement, says the number will grow “significantly over time.” Early data suggest that a candidate’s performance on the Imbellus problem-solving simulation is a slightly stronger indicator of whether she’ll be hired than her scores on McKinsey’s traditional test—though Kantar says those results are not “earth-shatteringly groundbreaking.” And that’s still a long way from proving those people will be more successful on the job. “We kind of need companies who at this stage are interested in taking the pretty long-term view. Which is, they’re eventually going to run out of talent in their pipeline who are fit for the nature of cognition their work requires if they don’t invest in reorienting school toward teaching to these deep-thinking skills.”
Imbellus’s headquarters in Culver City sits in a cluster of sleekly renovated warehouses. When I visited on a balmy morning in late January, Kantar’s black Tesla was parked in front. The office’s glass garage doors were pulled open, and a row of potted bamboo trees gave the place the feel of a boutique hotel lounge. While a group of Imbellus employees sipped coffee in the open-air kitchen, Kantar chased her Dalmatian, Nala, across the concrete floor.
The company has developed six game-based simulations, in collaboration with McKinsey and its other corporate partners. (Citing confidentiality agreements, Kantar declined to identify them.) Imbellus’s software captures and analyzes every keystroke a player makes while going through the simulation, to arrive at both a “product score” and a “process score,” which Imbellus generates within two hours of the game’s completion. Erica Snow, the head cognitive scientist, says, “We’re not just interested in whether you got it right in the end. Cognition is dynamic—so we’re also incredibly interested in how you got there. The goal is not the same as in scoring a multiple-choice test. We want to know: How did you make the choices you made? When you made errors, how did you correct them?”
The company uses artificial intelligence to make the virtual environments and animals look indistinguishable from the real thing. “You want to get people in a flow state while they’re taking the test, so that they’re functioning at their highest capacity,” Kantar says. “You never want the art to distract. If I gave your kids a game that looked like Pac-Man from the ’80s, they’re going to be like, ‘Why is the screen broken?’ It doesn’t look like what they’re used to seeing.”
Kantar says she was rejected by “something like 50” venture capital firms during a fundraising round last year, before securing her first major investment from Owl Ventures, an educational-technology-focused fund. VCs are skeptical about her in part because, she says, “I don’t look the part of someone who would be running a hard-science company. There are no mental models that they can hang on to in order for me to make sense. And Elizabeth Holmes certainly didn’t help my case.”
Comparisons to Holmes, the disgraced ex-CEO of the defunct blood-testing company Theranos—another dazzlingly bright, twentysomething female college dropout—are unavoidable. “People even say I look like her,” Kantar says, sheepishly. Jeff Hunter, a former Bridgewater Associates executive who’s advised Kantar, says, “When you’ve got a young female entrepreneur who is selling a big vision, you’ve got a lot people who are, like, is this smoke and mirrors? Is this going to be something where no one’s really peeling back the layers and looking behind the curtain and saying, ‘Is this real?’ But she’s making it work. There’s a real thing there.” Kantar has surrounded herself with women—including Imbellus’s chief operating officer and all four of its board members—but says doing so wasn’t a conscious decision.
“It’s something we’ve discussed, how much less scrutiny we would face if we were men, how much more implicit trust you’d get. But there’s nothing you can do about it, and talking about it makes you sound kind of silly,” says Meredith Perry, founder of the wireless technology startup UBeam and a close friend. “Rebecca’s not easily influenced by anyone at all. She’s not like some lucky kid. Everything she does is planned and thought out and very strategic.”
Kantar plans to begin giving the Imbellus test to high schoolers later this year. She also says, without going into many details, that the company is bidding to take part in a federally funded national assessment that will test 100,000 students in 2022.
If they gain traction, tests such as Imbellus’s could, over time, help reduce the fixation on four-year college degrees by bolstering alternative paths to employment. Kantar says that “for more than 50 percent of kids, college is net bad.” Her goal isn’t to make higher education obsolete but to convince colleges they’re selecting students based on standards that no longer make sense: “We’re establishing that simulation-based assessments are a better medium for getting at deep-thinking skills than multiple-choice tests,” she says. “We’re using that to go back to colleges and say, ‘Hey, we’re already testing 70 percent of your graduates. Let us run our tests alongside the SAT, the ACT, and APs.’ If we can build a test that definitively measures skills that matter more for work and better predict outcomes that really matter on a longitudinal basis, like, why keep what we have now?”
For all her ambition, Kantar is clear-eyed about the prospects for top-to-bottom reform of the U.S. educational system. “I don’t think my test is a silver bullet. A lot of school districts have a lot of problems beyond the assessments they’re using. The under-resourced schools are still going to struggle. But I think that over a generation, hopefully, I can inch toward something that’s more relevant to adulthood than what we have today. The schools who start way behind are still going to be relatively behind in hitting that North Star. But hopefully in their preparation for it, their kids are going to be left a little better off.”
Will they? “My worry is that the kind of assessment she wants to build is going to take us in the wrong direction—away from college prep and toward something even fuzzier and less rigorous and cognitively challenging than what we have today,” says Michael Petrilli, president of the Thomas B. Fordham Institute, a think tank focused on promoting standards-based reform. He’s met with Kantar. “My sense is that she doesn’t see a lot of value in teaching traditional academic skills, and I disagree.”
To that, Kantar responds, “I’m not saying that kids don’t need to know history or math or biology. They do. But my thought is, can you move the North Star of the system a bit? You’re still going to biology class and history class, but in those courses there’s a little less focus on specific modules of curriculum and more focus on practicing the thinking that’s required for work and not just college. What I’m trying to do is to reconnect K-12 education with the world of work and the reality of being an adult.”
She goes on: “I just want people to know that we won’t stop. At some point, it’s going to work, whether it takes 5 years, or 20 years, or 50 years and whether it means doing it alone or doing it with others. I’m really pretty sure that my initial mission and research was right—that this testing has to change to see the rest of the dominoes fall.”
To contact the editor responsible for this story: Jim Aley at email@example.com, Jillian Goodman
©2019 Bloomberg L.P.