Edunudging: the future of learning? 

Arguably, the most important goal of education is to support students to achieve their full potential, both as learners and as humans overall. Yet, each and every day students face a number of behavioral barriers that interfere with learning choices, such as ingrained cognitive biases, procrastination, or fear of failure. When faced with these challenges, even the most motivated learners will struggle to succeed academically.

As I alluded to with the thirsty horse example, small changes in someone’s “choice environment” can help promote desired behaviors. Each time we make a decision about something, our thought processes are influenced by how choices are presented to us. This is called choice architecture. If we want to subtly alter people’s behaviors to make better choices—without actually limiting their choices—then we can use nudges to redesign this choice architecture.^1,2 Importantly, nudges don’t involve coercion but simply provide gentle encouragement to guide individuals toward desired behaviors by leveraging aspects of human psychology. 

We’ve been nudging students to make better educational decisions for over a decade now, a phenomenon that Educational Science and Sociology professors Mathias Decuypere and Sigrid Hartong refer to as “edunudging.”³ Yet, despite increasing interest in the application of behavioral insights in learning spaces, we still know relatively little about nudging in education; only about four percent of all nudge research relates to education,⁴ with most of that focused on higher education. The limited evidence that we do have available, however, suggests that nudging has the potential to be an effective tool for promoting a range of educational outcomes—such as increasing engagement in online study, student enrollment, and grade attainment—that is, if it is done the right way. 

One of the main challenges in applying nudges in educational contexts is distinguishing between end goals and the behaviors that lead to them. Sometimes we can be impatient and want to see immediate results rather than looking at the underlying processes needed to achieve those results. Anyone who’s ever tried to lose a few pounds, for example, will know that you won’t reach your weight loss goal in a week just because you went to the gym a couple of times and chose the salad option on your night out. Achieving the end goal takes time, and will only happen if you address the underlying behaviors around eating, exercising, and other factors such as stress. 

Most attempts to apply behavioral insights to education contexts have used a goal-focused approach, where nudges primarily aim to improve end goals (such as better grades or higher attendance) rather than encourage the behaviors that eventually lead to those end goals. Take, for instance, a study conducted by behavioral economists Ghazala Azmat and Nagore Iriberri.⁵ For an entire year, high school students in Spain were shown their grades relative to their peers, leading to a 5 percent increase in student attainment. When the nudge was removed, the positive effect vanished. Why? Because the behaviors linked to the students’ performance were never considered, only the end goal. 

The main problem with a goal-driven approach to nudging is that it isn’t sustainable; while you might be successful in achieving desired goals for a limited time, the underlying behaviors haven’t changed and are going to be repeated in the future. Moreover, you also run the risk of achieving end goals through undesirable cognitive processes or unwanted behaviors. Better grades don’t necessarily mean that students are more engaged in their studies—some may have resorted to other options, such as cheating or using essay writing services. 

When applied properly, nudge theory uses cognitive processes to gradually create a behavioral change which can eventually help to reach the intended end goal. As economists Mette Damgaard and Helena Nielsen emphasize, knowing which behavior to target is crucial for successful nudging in education.⁶ But what happens when you are faced with edunudging hundreds, perhaps thousands, of students, all of whom have different behaviors and end goals? For that, we need EdTech and the power of big data and machine learning. 

Contemporary edunudge research and application seek to maximize the benefits and resources of digital environments—such as data tracking and analytics, big data, and machine learning—to optimize nudges. In 2020, for example, the UK’s innovation agency, NESTA, published their “incomplete guide” to Applying Behavioural Insights in EdTech. In this, NESTA unveiled a framework for using machine learning in conjunction with behavioral science to create more personalized interventions.⁷ The result of marrying behavioral science with EdTech? Hypernudging.⁸

In comparison to analog and digital nudges (such as optimizing the placement of healthy food choices in school cafeterias or sending parents text reminders), hypernudges are networked, continuously updated, dynamic, more persuasive, and less obtrusive. Continuous data collection, large databases, and predictive modeling produce increasingly individualized nudges that are adapted to the ever-changing behaviors of specific individuals. Hypernudges aren’t unique to the world of EdTech; they can be found in various fields, such as self-tracking technologies, product design, politics, and healthcare—basically any situation in which massive data sets can be used to influence behavior. 

Some examples of data-driven, interactive hypernudging in education include:

• Gamification: Gamification is the use of game elements in non-game contexts to encourage certain behaviors. When these elements are embedded within EdTech applications, they can serve as effective nudges to increase students’ engagement and persistence. By providing immediate feedback and fostering personal satisfaction, features including badges, rewards, progress indicators, and social competitions can encourage desired learning behaviors. For instance, the Learning Management System (LMS) called Classcraft uses self-determination theory as a framework to incorporate game elements into non-game educational contexts. 
• Data-Driven Decision-Making: EdTech platforms equipped with analytics capabilities can generate actionable insights for educators and administrators to implement nudges at a large scale. By analyzing student performance data, identifying patterns, and predicting intervention outcomes, educators can design more effective nudges to support student success and optimize learning. Open-source learning analytics tools, such as Google Analytics and Moodle Analytics, are used extensively in higher education contexts to capture and act upon data about students’ engagement, success, and satisfaction.⁹ 
• Adaptive Learning Systems: Adaptive learning platforms often leverage behavioral insights and predictive analytics to deliver targeted nudges that guide students toward effective learning strategies and resources tailored to their needs. Adaptive learning systems can track data such as student progress, engagement, and performance, and use these insights to provide personalized learning experiences. A well-known example is LinkedIn Learning which leverages algorithms to match students with appropriate coursework and suggest new courses based on what they’ve already studied. 

Let’s see how these approaches might work in a real-life example. Imagine you’re a university lecturer preparing to teach a highly important topic with a large graded assignment at the end of the semester. You want to make sure that your students both enjoy the material and academically succeed by getting the assignment done. Using a learning analytics tool, you discover that students are most engaged when receiving immediate feedback rather than waiting weeks to receive marks on their end-of-term papers. This insight leads you to try a learning management system that provides students with progress indicators as they work through the course material. By tapping into the students’ innate competitiveness and their desire for instant gratification, you successfully manage to end the semester with all students passing the course and a set of impressive final assignments!

The overall objective of hypernudging is to achieve personalized edunudging. In education, personalization is about recognizing that each student is an individual whose learning experiences should be tailored to their unique needs and challenges. To achieve just this, Georgia State University is using big data and edunudging to solve the problem of college dropouts.¹⁰ In the United States, 40 percent of students who start college don’t finish in at least six years, with some never receiving their degrees at all. Since 2012, Georgia State has been feeding student data into a predictive analytics system to figure out who is likely to succeed in college and how to keep them enrolled. 

The main emphasis of the university’s program is to ensure that students choose a major that is well-suited to their skills and talents; this entails taking classes in the correct sequence, as well as not filling their schedule with classes they don’t need. Academic counselors are assigned 250+ students to monitor using a predictive system that continuously identifies students with a “risk level” of either green (low risk), yellow (medium risk), or red (high risk). Based on this categorization, personalized nudges and data-driven decision-guidance techniques are then sent to the student. 

However, an unexpected level of personalization has also emerged from the system. After receiving an email, students are invited to an appointment with their counselor to further explain the data. Rather than being told what to do by their advisor, students are encouraged to use the data to make their own decisions about the next steps in their learning. This ‘self-nudging’ is intended to lead not only to the students’ academic success but overall changes in the way they make choices about their education.

So what’s been the result of Georgia State’s algorithm-led hypernudging? So far, so good: between 2011 and 2018, the university’s six-year graduation rate rose from 48 percent to 55 percent. However, despite its increasing use across institutions around the globe, the jury is still out as to whether or not personalized edunudging actually changes behaviors in the long term. 

Researchers at the Vrije Universiteit in Amsterdam, for example, conducted a study to see if it was possible to increase student participation in an online proctored (software-monitored) formative statistics exam by using digital nudging.¹¹ In the study, students in the intervention group received targeted email nudges based on their self-recorded motivation and perceived ability, while students in the control group were sent plain email nudges with just the required information to perform the intended behavior. The researchers found that the tailored nudges were no more effective in increasing student engagement with the assessment than the plain, non-personalized nudges. 

What can we learn from comparing the results of the intervention in Georgia with the findings of the study in Amsterdam? That the number one design principle for edunudging is behavior before end goals.

It would be naïve to think that in today’s society, edunudging is applied solely to help students achieve their full potential. Hypernudging in particular has been criticized for its entanglement in state-driven goals to maximize policy efficiency and return on investments. As with the case at Georgia State University, hypernudging students to be academically successful is not only about what’s best for the young learners; it’s also about lessening the USA’s $1.5 trillion student loan burden and promoting value for money. 

Some have even suggested that hypernudging will gradually turn education actors, such as teachers, into “classroom nudge operatives” who spend less time teaching and more time implementing nudges.¹² These concerns can be read as part of wider global anxieties surrounding the extent to which machine-learning technologies will eventually replace human professions. 

Two further important considerations arise from these interventions. First, do schools, education providers, and commercial companies have the mandate to decide which behaviors are desirable for our offspring? And second, should these entities be actively attempting to alter students’ behaviors without thoroughly consulting parents and caregivers? 

Rather than viewing students as autonomous individuals who are finding their place in the world through learning, edunudging and hypernudging might push students to behave according to prescribed desired behaviors. These issues are part of wider ethical debates around whether nudging undermines personal autonomy, human dignity, and sustainable well-being.^13,14 If nudges are already questionable when applied to adults, it becomes even more problematic when working with young people who have less experience in critical decision-making.

I want to go back to the horse for one moment. Earlier, I suggested two nudge interventions—more water troughs and the addition of apple juice—to encourage the horse to adopt the desired behavior of drinking more water. But we never established the end goal; was it the overall happiness and well-being of the horse or to have a healthy, well-hydrated horse that will go on to win races and earn its owners lots of money? 

Increasingly, it’s hard to tell who really benefits from edunudging and it’s important not to lose sight of the intended beneficiaries of education: students. For edunudging to truly remain relevant to education moving forward, we need to thoroughly reflect on whose behavior we’re changing and to what end.