Using Artificial Intelligence to Navigate the New Challenges of College and Career

March 9, 2021
Chatbot

By Mike Meotti and Drew Magliozzi

The COVID-19 pandemic has put a spotlight on the fact that American workers at all levels are at a crossroads due to the relentless improvement of technology. The World Economic Forum in 2020 found that, amid “the largest experiment in mass remote-working in history,” more than half of businesses have accelerated automation in response to the crisis. Even before the outbreak of COVID-19, the path to economic opportunity in the United States was, in many ways, less clear than ever as workers’ careers have taken increasingly non-traditional routes with technology carving new tracks and creating new ways of working.

Technology has always been a disruptive force. But can it also be employed as an accelerant? How can we help current and incoming students, recent graduates, and working learners navigate a fragmented landscape of education and work?

Against a backdrop of unprecedented volatility, policymakers and technologists must harness technology and utilize some of the very forces that are causing this upheaval to build new routes to economic mobility for all students.

While Artificial Intelligence (AI) and related technologies are often seen as disruptors, these same technology-based tools can redefine what’s possible and guide students through the challenges of college and careers.

Defining the Problem: Three Inflection Points

As we see it, there are three primary inflection points on the journey to and through college. We’ve made the focus of our work to address a set of pervasive, pernicious challenges that are keeping too many people from accessing the sort of educational opportunities that can be a path to economic mobility for all. AI is a powerful tool to address these very challenges.

  • College Access. We’re focused on addressing the increasing inequities around college access, including so-called “summer melt,” in which promising students apply and are accepted to college, but never show up in the fall. Each year, summer melt affects an estimated 10 to 20 percent of the more than 2 million students who plan to attend college -- the rates are higher, sometimes up to 40 percent higher, among low-income and first-generation students. In addition, today’s students are making key decisions about their education and careers amid an unprecedented pandemic that has already taken its toll on college access, particularly for students of color.
  • College Completion. Unfortunately, simply enrolling in college is only the beginning of a postsecondary journey that is too often bumpy for students who don’t receive enough support. As a result, we’re also focused on challenges with persistence and completion, and working to ensure that colleges and universities have the resources to help students not just get in but stay in and graduate. At open-admissions colleges, which are open to virtually all applicants with a high school diploma or GED certificate, the persistence rate is just 62 percent. And at any institution, open-admissions or not, completion rates are worse for low-income and first-generation students than for their peers from higher-income or more-educated households -- leading to persistent and substantial disparities in graduation rates between white students and their Black and Latinx peers.
  • Career Entry. Of course, the journey doesn’t end at college graduation either. For recent college graduates,underemployment has become a “pervasive problem,” according to the Strada Institute for the Future of Work. Researchers found that 43 percent of college graduates in 2018 were underemployed in their first job -- that is, they were employed in a job that didn’t require the skills and credentials they had earned, which has dangerous downstream effects for their long-term career prospects. At the same time, automation threatensworkers in both blue- and white-collar professions: a 2017 white paper from economists at McKinsey & Company predicted that roughly 60 percent of jobs might see about 30 percent of skills automated. An earlier paper from Oxford University concluded that AI, robotics and other developments would likely fully eliminate 47 percent of jobs. Recent surveys of business leaders suggest that the COVID-19 pandemic is hastening this process, with about six in 10 businesses saying they need to develop “new processes” in response to the realities of the pandemic. And the Education Trust found that focusing on racial gaps in college graduation rates may actually mask even bigger disparities in post-college success for Black and Latinx students, as they navigate a labor market that is rife with discrimination and systemic bias.

In short, the path to and through college is not an easy one -- but if history is any guide, the degree will only become more important as automation and technological change reshape the labor market and demand constant retraining. In the wake of the last recession, for instance, workers with college degrees were spared from the worst declines in employment and wages, while those with less postsecondary education experience bore the brunt of the economic downturn. Our determination to ensure that more students don’t just enroll in college, but graduate and enter into fulfilling careers has been the catalyst for us to look to AI for ways to address these challenges.

AI in Education

A decade ago, we would have had to exponentially increase the number of college counselors in schools to exponentially increase the number of first-generation college students. With technology, we can have a bigger impact at a smaller cost, elevating the people who do the work and finding ways to put them in the highest-leverage situations.

Over the past few years, colleges have quietly rolled out chatbots to help students navigate areas as diverse as admissions and financial aid, IT support, academic advising, and career services. One such AI developer, AdmitHub (soon to be known as Mainstay), works with more than a hundred universities today to improve access and retention. And we’ve begun to see the impact that this technology can have on the increasingly intractable challenges facing students on their path to and through college.

AdmitHub’s work primarily employs AI using chatbots, artificially intelligent agents such as Siri and Alexa, which provide a user interface for AI and non-AI systems that uses spoken or typed speech. Washington state is leveraging the benefits of AI to pioneer something new and different for students preparing to enter college: a digital infrastructure that wraps around in-person supports and affordably fills in gaps. It promises to be a national model. More complex agents, like the ones we’ve developed in Washington state, use Natural Language Processing (NLP) to parse human language and learn from previous conversations to improve their accuracy.

AdmitHub and Washington Student Achievement Council created Otterbot, a free texting service built to support 10,000 high school seniors who have signed up for the College Bound Scholarship and ensure that Washington state students have access to helpful financial aid resources any time they need support.

Unlike in most high schools, where overworked college counselors each must serve hundreds of students face-to-face each year, college bound students have access to Otterbot 24 hours a day, seven days a week, enabling them to receive periodic messages with need-to-know financial aid information, resources, suggestions, dates, and deadlines.

When a student gets a text from Otter, they reply directly to the message. Students can also query Otter by text at any time. If the app receives a question it can’t answer, it will send the message to a financial aid expert or other college admission staff member from the Washington Student Achievement Council.

While it’s still early, results from the program suggest that it’s making a difference. In the first year after Otterbot’s launch, students who interacted more frequently or received more messages from the chatbot were also more likely to have better outcomes (e.g., higher high school graduation and Free Application for Federal Student Aid (FAFSA) completion rates). More specifically, among students who asked Otterbot five or more questions, over 86% graduated from high school and over 60% completed the FAFSA.

At Georgia State University, researchers found similar results from a randomized controlled trial: incoming students intending to attend the University were 3.3 percent more likely to begin their fall semester there if they used a smartphone application that sent them text messages reminding them of application deadlines. That translated into an estimated 21 percent reduction in “summer melt,” in which students accepted for admission in the spring fail to enroll by the fall. The impact of the app was similar to other interventions, researchers found, but with a smaller burden on staff time and budgets.

Georgia State’s app also utilizes a chatbot that answers “guiding questions” automatically -- it has learned to provide better answers over time and help students with pre-enrollment processes, such as navigating financial aid.

Research has found that students are more willing to be candid with chatbots, especially when discussing personal issues as diverse as health, finances, or immigration status.

College Completion

Due in part to the rise of degree inflation, millions of jobs that didn’t previously require college degrees now do -- which creates even more barriers for the Black and Latinx students and workers who are disproportionately less likely to have earned a four-year degree.

The opportunity to employ AI for student success doesn’t end once a student has enrolled. Using text messaging, chatbots can reach students on devices they already carry, enabling institutions to provide important information and remind them of key deadlines, like financial aid filings, that can make the difference in their ability to return to campus the following semester.

Research suggests that applying AI to student retention and completion can also be effective. Georgia State University found that ongoing students who used the university's chatbot were 3 percent more likely to re-enroll than those who did not use the chatbot. Chatbot users had higher rates of the FAFSA filing and registration, and the greatest gains were among low-income and first-generation students -- the same students who most struggle with the enrollment process.

AI is also improving instruction in ways that boost engagement, retention, and success. Packback, an inquiry-driven online discussion platform, uses AI to spark college students’ curiosity and motivate them to ask their own open-ended questions. Designed in partnership with college faculty and administrators, Packback’s technology helps instructors moderate class discussions, give feedback, and grade participation automatically. The platform enables students to receive feedback on their posts and questions in real time, while also freeing up time for instructors to focus on more substantive topics and support. And forthcoming research conducted at 10 colleges around the country suggests that the use of this platform has led to increased engagement and improved academic outcomes.

At Rensselaer Polytechnic Institute, The Mandarin Project, an experimental collaboration between the university and IBM, allows students to immerse themselves in Chinese culture without leaving a classroom. A “cognitive immersive room" with a 360-degree, floor-to-ceiling screen allows students to practice Mandarin with AI-powered animated characters in simulations that make students feel as if they are in China.

At Stanford University, researchers have developed an AI-powered chatbot called QuizBot that uses natural-language conversations to help students review coursework. The tool effectively serves as an improved flashcard application that has led to a 20 percent improvement in retention, the researchers said recently. While students may need to spend more time with a conversational agent than with flashcards, they found, students preferred the AI chatbot to flashcards, especially in more informal review settings.

It’s worth noting that these are just a few of the myriad examples of emerging technology designed to help students enter and persist in higher education. Other examples that demonstrate the application of technology for college access and retention include:

InsideTrack uses coaching to improve retention and graduation rates for students across all demographics, backed by predictive analytics technology that helps staff and advisors know which students to engage, when, and how to address their most pressing challenges. A 2011 review by Stanford professors found that it improved retention and completion rates by as much as 15 percent.

RaiseMe enables students to research colleges and access “micro-scholarships” before they attend. Students earn achievements through good course grades, club involvement, sports, volunteer activities, and the like. For each achievement, students are awarded more that can be applied to the cost of attending college.

Strive for College uses a sophisticated algorithm to match students with free, one-to-one virtual mentoring to help them navigate the college and financial aid application process. For those already enrolled in college, it matches them with a mentor who supports them toward graduation and career opportunities.

Career Entry

Even with improved outcomes, students need guidance on the right coursework for their degrees. To that end, WGU Labs, a non-profit founded by Western Governors University, has been working with researchers from Carnegie Mellon University to develop a tool to help guide students into coursework that supports the degree program and career they’re seeking. Preliminary findings suggest that the tool helps students make better academic and career decisions.

Along with the others, this example demonstrates the ways that artificial intelligence can help to improve outcomes and expand access within the current education system -- what you might call repaving existing roads to economic mobility. But the reality is that the roads themselves may be outdated, and in need of a new system altogether.

Ultimately, it should be no surprise that a system designed to meet the needs of a rapidly industrializing country at the turn of the century is imperfectly equipped to meet the needs of an increasingly global, digital world. But as the pace of technological change accelerates and the pandemic throws economies into turmoil, demand is increasing for new approaches that break down the “traditional” silos of school, college, and careers.

In the face of these challenges, organizations like Get Schooled and SkillUp are tapping AI to help illuminate learning-to-earning paths, providing on-demand advice to students and job-seekers, and helping match skills and interests with open jobs and career tracks. On the other side of the equation, emerging companies like FutureFit.ai are providing similar services to employers, who can use AI to help their own workers navigate the increasingly dynamic labor market.

Wait, but…

It’s not hard to paint a rosy picture of AI’s potential as an engine to drive better connectivity between education, work, and economic opportunity. But it’s also important to consider what we give up in exchange for a more accurate navigation system.

By this point, we’re all used to Google knowing everything about us. There’s no better way to understand real-time traffic patterns, or tell riders when the subway is going to arrive, than to be constantly sending and receiving data from every car on the road or train on the track. Technology, much of it powered by artificial intelligence, is following us wherever we go and whatever decisions we make. Does it know too much?

The truth is, of course, that it’s never black and white. Artificial intelligence has rightly been the source of significant skepticism throughout the education world -- from its role in student data privacy to the invasiveness associated with online test proctoring and the opportunity cost of not having human interaction in spaces that may be uniquely human and collaborative. Accusations of algorithmic bias have brought to the forefront a very real, and very dangerous side effect of our enthusiasm for technology: if not designed intentionally, and with empathy, it runs the risk of perpetuating -- if not exacerbating -- the endemic biases that are the foundations of inequity in our society.

What does an ethical, empathetic, student-focused approach to AI look like in practice? It’s a question being considered by a growing number of companies, and the answers are still taking shape. But our work helping facilitate communication between schools and students may point to some promising practices, including these:

  • Be transparent. Take data security seriously. Be up-front about both the potential risks inherent in any new technology, as well as the clear steps being taken to address those risks, which goes a long way not only to just assuaging users but also to cultivating a company culture that prioritizes responsibility and integrity.
  • Know your limits. AI is a vitally important tool for communicating key information -- like filing deadlines, reminders, and admissions tips -- to thousands of students at once. But it can’t do everything. Constant monitoring by human advisors, and the ability to quickly elevate serious issues to those advisors, is as important as continuing to advance the capabilities of the technology itself.
  • Strike the right tone. The DC-based think tank New America’s groundbreaking report “How You Say It Matters” should be mandatory reading for any college (or institution of any kind) hoping to communicate using artificial intelligence. The report’s thesis stresses the importance of both encouragement and clarity: “Be sure to balance a positive tone with a realistic approach so that students are both encouraged to continue their education but are not set up with false hopes for their success.” In our case, that’s meant a lot of coaching with Washington’s statewide AI tools to ensure that they are striking the right balance of clear guidance and positivity.
  • Measure what matters. When implementing a new tool or technology, institutions should set clear metrics to ensure that both the institution and the technology developer are held accountable for results. Whether the goal is to help more students enroll, boost persistence and completion, or increase rates of FAFSA filing, setting objectives, and regularly measuring progress will ensure that you aren’t just implementing technology for its own sake. Georgia State’s randomized controlled trials are an example of an institution’s intensive efforts to ensure that technology was being applied with measurable and distinct results.

Conclusion

It may be fair to say that we’re in an interim period, one that has been the subject of far less apocalyptic speculation, wherein AI and other new technologies are beginning to demonstrate their utility and potential when applied to solve pressing societal problems. That’s what we’re beginning to see in Washington state and at colleges around the country: artificial intelligence that, far from eliminating the role of human workers, actually supports their work and enables them to help more people more effectively. In a world that is -- often rightfully -- skeptical about the transformative impact of technology in society, we hope examples like these can cut a more hopeful path forward with many opportunities to help people move from learning to earning faster and more effectively.

 

About the Authors:

MikeMeotti

Mike Meotti is the executive director of the Washington Student Achievement Council (WSAC), the state’s higher education agency. He has served as commissioner of the Connecticut Department of Higher Education and executive vice president of the Connecticut Board of Regents for Higher Education.

DrewMagliozzi

Drew Magliozzi is co-founder and CEO of AdmitHub, a ground-breaking communication platform that uses AI to help students navigate college admissions, enrollment, and completion processes.

 

Authors Meotti and Magliozzi have collaborated on a Washington state effort to apply emerging technologies such as artificial intelligence (AI) and machine learning to the college application and attendance processes. Together they have decades of experience in higher education and college admissions policy work and share a belief that emerging technologies have the potential to revolutionize how people navigate the increasingly complex path to college and the opportunities it unlocks.

 

 

***********************************************************************************************************

What Is AI?

While it’s not a new field, Artificial Intelligence (AI) is young, relatively speaking. Born in 1955, it took shape when the computer scientist John McCarthy, along with John Nash, Marvin Minsky, and others, proposed what they conceived of as a summer research project at Dartmouth. The group envisioned building a machine that might behave “in ways that would be called intelligent if a human were so behaving.”

Sixty-five years later, AI has not only passed various Turing Tests to behave like an intelligent human. It has utterly smashed them. From AI-powered chess apps to well-known efforts by Google to challenge the world’s best Go players to the jaw-dropping possibilities of OpenAI’s text generation, AI has proved to be limited only by designers’ daring imaginations.

AI typically falls into two broad categories: “Narrow” AI (or ANI) is programmed to perform a single task, such as predicting the weather or playing chess, and “General” AI (or AGI) can successfully perform any intellectual task that a human being can. For now, however, AGI remains in the realm of science fiction, even as ANI becomes increasingly sophisticated.

Typically, AI refers to a type of machine learning that resembles what we consider human intelligence. Machine learning is the process by which systems, accessing data, automatically learn and improve from experience without being explicitly programmed.

The most relevant applications of AI in education work stem from what’s called deep learning, a subset of machine learning which allows for things like accurate speech and facial recognition, by employing networks capable of learning from data that is unstructured or unlabeled. Computers parse accurate meaning from human conversation by means of natural language processing (NLP). Using contextual clues, NLP helps machines make sense of what humans are trying to say -- for example, parsing the difference between trying to reach the accounting department and finding out the requirements for the accounting major.

While deep learning is unique in that it can “teach itself” using fairly unstructured data, we have also been able to make use of supervised learning, in which programmers map training data (such as historical student records) to a single correct output, such as whether a student graduated on time. Then an algorithm finds the patterns and mathematical relationships that connect the inputs and outputs so it can predict what the output will most likely be for inputs it has never seen.

Bringing deep learning and supervised learning together is what enables us to both identify the barriers to college access and completion and also engage with students in ways that can remove those barriers.

 

magliozzi_meotti-using-ai-to-navigate-new-challenges-of-college-and-career.pdf280 KB
See also: Education