In August 2011, 5,000 Arizona State University (ASU) freshmen walked into windowless computer labs to encounter their new math instructor: a "robot tutor in the sky" that promised to read their minds and personalize their learning. No pilot program. No gradual rollout. Just a massive experiment in adaptive learning technology that would eventually become one of edtech's most cautionary tales.
This is the story of how ASU's partnership with Knewton, once valued at over $180 million, collapsed spectacularly, leaving behind frustrated students, resistant faculty, and critical lessons about the risks of implementing unproven technology at scale. It's a story that unfolds against the backdrop of Phoenix's dramatic boom-bust cycle, severe state budget cuts, and a university's desperate search for innovation in the face of financial crisis.
To understand why ASU rushed into the Knewton experiment, we need to understand the financial earthquake that hit Arizona's higher education system after 2008. The numbers are staggering: Arizona made the deepest cuts to higher education funding of any state following the financial crisis, slashing per-student funding by 54.3% between 2008 and 2018.
For ASU specifically, this meant an $88 million cut in 2008 alone, 18% of the university's base state budget. The human cost was immediate: over 550 staff positions eliminated, 200 faculty associate positions cut, and 48 academic programs shuttered. Faculty endured 10-15 day furloughs, including the university president. The state, which once provided the majority of ASU's funding, now contributes less than 9% of its total budget.
Meanwhile, Phoenix itself was reeling from the housing crash. After years of explosive growth, with home prices doubling between 2001 and 2006, the city became ground zero for the foreclosure crisis. Over 100,000 Arizona homes went into foreclosure in 2008, with Phoenix seeing 8,000-10,000 foreclosure notices per month compared to the normal 200-300. Property values plummeted 56%, the third worst rate nationally.
Enter Michael Crow, ASU's iconoclastic president since 2002, who had transformed the former "party school" into his vision of a "New American University." Faced with devastating budget cuts and soaring enrollment (which paradoxically grew from 67,082 to 70,440 students between 2008-2010), Crow saw technology as the answer. As one administrator put it, "We have used new approaches to teach mathematics with technology and individual pacing, improving learning and decreasing costs."
Knewton arrived on the scene with extraordinary promises. Founded in 2008 by former Kaplan executive Jose Ferreira, the company claimed its adaptive learning platform could create "individual, psychometric profiles that would presume to say, with statistical authority, what students know and how they learn." Ferreira wasn't shy about the company's ambitions, boasting at a Department of Education event in 2012: "We literally know everything about what you know and how you learn best, everything. We have five orders of magnitude more data about you than Google has."
The pitch to ASU was seductive: for $100 per student per course (passed on as a textbook replacement fee), Knewton would transform remedial math education. The system would break down courses into discrete concepts, continuously assess student understanding, and create personalized learning paths. Students could progress at their own pace, potentially finishing courses weeks early. Faculty would receive real-time dashboards showing exactly where each student struggled.
Philip Regier, ASU's Online Dean and Knewton's primary champion, articulated the vision with missionary zeal: "It's the Swiss cheese effect. You can't have a big hole in your knowledge. If you get a C, you know 70 percent of the material for one course. But the missing 30 is likely to be important to passing the next course."
What happened next violated every principle of responsible technology implementation. In August 2011, ASU deployed Knewton to 5,000 students without any pilot phase. The courses affected included MAT110 (remedial math), MAT117 (College Algebra), and MAT142 (College Mathematics). The remedial course was broken down into 52 discrete concepts that students had to master sequentially.
Wayne Raskind, ASU's Director of Mathematics at the time, learned about the implementation in the most shocking way possible. "I found out about the Knewton contract by reading the student newspaper while at a conference in New Orleans," he recalled. "The contract was signed very quickly and without a lot of consultation, at least on my part."
The lack of faculty consultation wasn't just a procedural oversight, it was symptomatic of a deeper problem. Faculty members, who would need to completely transform their teaching methods, were given minimal training and support. Some instructors thrived in the new "flipped classroom" model, walking around with iPads to coach individual students. Others struggled, with some having "almost half their students off-track," according to instructor Irene Bloom.
The initial results seemed promising. ASU reported that pass rates in remedial math jumped from 66% to 75%. The university generated positive press coverage, and administrators began planning expansions to "macro and microeconomics, psychology, biology, chemistry and physics." Regier even envisioned "an entire degree adaptively."
But beneath the surface, cracks were already showing.
The true impact of the Knewton experiment is best understood through the eyes of the students who lived it. Take Naomi, a struggling student in a yellow Sun Devils shirt, sitting in the MAT110 remedial math course. Her experience reveals the dark side of "personalized" learning:
"I think this is worse," she said, comparing Knewton to traditional instruction. "Because we have to do this on our own. In another class they would tell you what to do... But right now? I'm five lessons behind, and it's on my own, too, you know? Fifty-two lessons."
The isolation was crushing. In a traditional class, Naomi might have raised her hand, asked a neighbor, or caught the instructor after class. In the Knewton system, she faced her computer screen alone, falling further behind with each passing day. "It's embarrassing," she whispered. "They say this math is easy. To me it's not. I'm not good at it."
Not every student struggled. Maddy, a first-year student with thick-framed glasses, excelled in the system, finishing eight lessons ahead with a 97% average. "It makes you actually do it," she said. "There's no way you can actually not do it." But even her success came with a telling detail, when asked what she'd do with the extra time from finishing early, her answer was simple: "Sleep. Just sleep."
The variability in outcomes was extreme. In the same course, pass rates ranged from 33% to 100% depending on the instructor. Al Boggess, Raskind's successor as Math Director, acknowledged: "Some instructors adapted to the paradigm better than others, and so therefore that may have reflected itself in the different pass rates."
By 2019, Knewton's grand experiment was over. The company that had raised over $180 million in venture capital was sold to publisher John Wiley & Sons for a reported $10-25 million—a fraction of its peak valuation. ASU had quietly moved on, replacing Knewton with McGraw-Hill's ALEKS platform.
The failure had multiple causes:
The Knewton experiment can't be understood without the broader context of Phoenix's dramatic demographic and economic shifts. During the early 2000s housing boom, Phoenix experienced explosive growth, with the metropolitan area permitting 63,000 new homes in 2004 alone, far exceeding the 37,000 permits in the much larger Los Angeles area.
This growth brought new residents seeking affordable alternatives to California's high costs, swelling ASU's enrollment and creating demand for scalable education solutions. But when the housing bubble burst in 2008, Phoenix became one of the hardest-hit cities nationally. The economic devastation created both budget pressures and a flood of working adults seeking career transitions through education.
Today, Phoenix has reinvented itself as "Silicon Desert," with major tech companies like TSMC investing $40 billion in semiconductor manufacturing and LG Energy Solution building a $1.39 billion battery gigafactory. This transformation has created new demands for STEM education that ASU continues to scramble to meet, though now with more caution about unproven technologies.
The Knewton failure offers critical lessons for educational institutions considering adaptive learning or any transformative technology:
The regulatory environment has evolved significantly since Knewton's heyday. New federal oversight of edtech companies handling student data, enhanced FERPA protections, and the Department of Education's 2023 AI guidelines requiring "humans in the loop" would likely prevent some of Knewton's worst excesses.
The search for Maricopa County connections revealed limited direct relevance of Supervisor Clint Hickman to education technology policy, though county-level governance does impact educational infrastructure and funding decisions that create pressure for technological solutions.
For institutions still interested in adaptive learning, open-source alternatives now exist that offer transparency and community control:
These alternatives typically cost 60-80% less than proprietary systems while providing transparent algorithms that educators can inspect and modify.
Not everyone believes adaptive learning can deliver on its promises. Education critic Audrey Watters argues that these systems represent recycled 1950s behaviorism, reducing human learning to algorithmic processes. "Personalization," she contends, really means "predictive" learning based on extensive data mining that treats students as experimental subjects.
Recent research supports some skepticism. A comprehensive meta-analysis found only 59% of adaptive learning studies showed improved academic performance, with just 36% demonstrating increased student engagement. Many studies suffer from small sample sizes, lack of independent assessment, and funding from the companies whose products they evaluate.
The theoretical foundations also face criticism. Adaptive learning often focuses on correct/incorrect responses, potentially discouraging exploration and creative thinking. Despite claims of personalization, systems impose uniform pathways that may not accommodate diverse learning styles or cultural contexts.
The ASU-Knewton story isn't just about one failed partnership. It's about what happens when financial desperation meets Silicon Valley hubris, when the pressure to innovate overwhelms the responsibility to validate, and when students become guinea pigs in someone else's vision of the future.
Today, ASU continues to innovate, ranking #1 in U.S. News's "Most Innovative Schools" for eight consecutive years. The university now uses different adaptive learning platforms, implemented more carefully with faculty input. The scars from Knewton have taught valuable lessons about the difference between innovation and experimentation on live subjects.
For educational leaders facing similar pressures, whether from budget cuts, enrollment challenges, or the latest AI promises, the Knewton story offers a clear message: Innovation is essential, but it must be grounded in evidence, implemented with care, and always keep student success as the north star. The "robot tutor in the sky" may have crashed to earth, but the dream of personalized learning continues, hopefully now tempered with the wisdom that comes from spectacular failure.
The next time a vendor promises to revolutionize education with algorithms that know "everything" about your students, remember the 5,000 ASU freshmen who walked into those windowless computer labs in 2011. They deserved better than to be test subjects. Your students do too.