Harvard ALI Social Impact Review

View Original

AI Can Make Schools More Human, But Only If Schools Prioritize Relationship Metrics

Despite promises of efficiency leading to more focus on relationships between and among teachers and students, AI is on track to dramatically worsen student isolation – unless schools reconfigure their organizational models and metrics to elevate, rather than diminish, human relationships. 

Where promise and reality diverge

When ChatGPT went viral last winter, more than a few analysts wagered that AI tools could once and for all liberate teachers from the long list of vital but mundane administrative tasks that keep them from spending more time with students. Some estimates suggest that up to a staggering 40 percent of educators’ days are spent on activities like lesson planning, grading, and maintaining student records.

Seasoned experts have begun to write about the benefits recouped time could unlock.

Education venture capitalist Jennifer Carolan, who started her career as a teacher, wrote “My hope is that the proliferation of AI tutors will end busywork, reporting, hours of grading, and ultimately elevate the human social learning experience to something considered valuable, special and to be protected.” Teach For America founder Wendy Kopp, writing for the World Economic Forum, observed “By automating routine administrative tasks, AI could help streamline teacher workflows, giving them more time to build relationships with students and foster their learning and development.”

These arguments aren’t entirely new: a 2020 McKinsey report forecasting the rise of AI in education staked a similar claim: “[AI] will enable teachers to foster one-on-one relationships with students, encourage self-regulation and perseverance, and help students collaborate with each other.”

At the core of these predictions is an earnest hope that with more time, teachers can focus on the human side of education: deepening relationships with and among their students.

That focus is sorely needed. Decades of research have shown that relationships are foundational to healthy development and learning. And data overwhelmingly shows that stronger teacher-student relationships contribute to everything from better grades to higher college enrollment rates.

But despite all that’s known about the power of connection, the promise and reality diverge: social connection in schools appears to be on the decline. During the 2022-23 school year, for example, student perceptions that their teachers were making an effort to understand their lives outside of school fell to a new low of just 22 percent of students.

It’s hardly surprising, then, that the AI revolution is spurring hopes that with more time, teacher-student relationships can blossom.

What gets measured gets done

Unfortunately, appealing as the proposition may be, it rests on the faulty assumption that schools are designed to optimize for connection in the first place. While most educators would wholeheartedly agree that relationships matter, schools rarely measure students’ connections – with educators, peers, or community members – with regularity or rigor.

Without any such metrics in place, relationships, valuable as they are, cannot be prioritized alongside learning outcomes. Instead, connections are treated as but one of many contributing factors – along with text books, technology tools, and the like – that schools enlist in an effort to keep students on track and boost their test scores. But those connections are also one of the most cost-intensive inputs. In turn, as educators rationally seek the most efficient route to increasing learning outcomes, they are incentivized to look for shortcuts that require less interaction, not more.

In other words, schools’ organizational models routinely leave students’ access to social capital – relationships and the resources they offer – to mere chance.

That gamble is most evident when it comes to teacher-student connections. Teachers are not evaluated on their ability to forge deep connections with students. Whether students bond with their teachers or other staff often hinges on them opting into electives or clubs. And little time in the schedule is dedicated to forming relationships rather than receiving instruction. Having the same teacher for multiple years is a proven method for strengthening bonds but often happens by accident.

Moreover, schools orient inwards: with little to no data on whom their students already know, it’s only by luck or individual initiative that educators know who their students know beyond immediate families and caregivers.

Harnessing AI’s efficiencies in classrooms that rely exclusively on traditional attendance and academic measures of success hardly guarantees that schools will successfully use newfound time to deepen connections.

Schools’ missing metrics: relationship data

In 2021, my colleague Dr. Mahnaz Charania and I began to look for ways that schools could capture these missing metrics. We set out to find examples of schools and youth-serving nonprofits that were filling this blindspot, measuring students’ social capital as its own programmatic outcome. We cataloged our findings in The missing metrics: emerging practices for measuring students’ relationships and networks. An updated version of the report came out this spring.

Measuring relationships can sound impersonal and impractical. How can something as nuanced and mysterious as human connection be distilled down to an indicator or survey item? Networks, moreover, are dynamic and ever changing. Depending on the challenges or opportunities confronting a young person on any given day, different people may be more or less critical to their success.

To address these complexities, we looked for examples of student network measures across four dimensions that we selected based on empirical research on why and how relationships drive healthy development, learning, and access to opportunity:

  • the quantity of relationships in students’ networks,

  • the quality of those relationships,

  • the structure of students’ networks, and

  • students’ ability to mobilize networks.

The first three dimensions reflect broader sociological, economic, and political science research on the ways in which social capital shapes young people’s access to both critical supports and new opportunities. The framework’s fourth dimension reflects broader youth development and social and emotional learning research, measuring the ways in which young people's skills and mindsets shape how they build, maintain, and activate relationships.

Diversity of students’ networks undergirds all four of these dimensions. Diversity can refer to different racial, ethnic, and socioeconomic backgrounds represented in students’ networks, as well as to the array of expertise, interests, and professional experiences. Fostering diverse networks not only offers young people access to opportunities that may be beyond their reach, but also the opportunity to mutually build emotional and cultural competencies as they maintain relationships over time.

Social capital metrics in action

A host of early innovators across K-12, postsecondary, and workforce development are making important strides toward purposefully measuring students’ relationships across all four dimensions. In turn, they’re able to keep an accurate – and more equitable – pulse on the connectedness of their students.

Relationship mapping is one emerging strategy to capture baseline data on the number and nature of relationships in students’ lives.

For example, the Making Caring Common Project at the Harvard Graduate School of Education has created a tool to help K-12 schools visually map the relationships between students and staff. In schools using the “Relationship Mapping Strategy,” faculty and school staff are presented with student rosters and asked to identify the students with whom they feel they have a strong connection. Students can likewise generate a list of connections and identify the faculty and staff with whom they feel they have strong connections. Schools can then work to ensure that every student possesses at least one – but ideally many – positive and stable relationships at school. This approach prevents schools from leaving students’ access to connections to chance and positions leaders to identify, early on, those who may need additional supports. From there, schools can repeat mapping exercises at regular intervals to understand if, and to what extent, students are increasing and deepening relationships.

Another mapping approach we identified aims to put relationship data not only in front of teachers but also directly into the hands of students. iCouldBe, a virtual mentoring program, connects high school students to online mentors who guide them through a college and career curriculum. The curriculum combines network-mapping with a series of activities called “quests” that prompt students to identify and forge connections based on their academic and career interests. Many quests nudge students to build offline relationships with teachers at their school or members of their community. At each juncture, students identify these additional connections, which are, in turn, added to personalized network maps on the iCouldBe app. As a result, iCouldBe can keep up-to-date information on the number of connections students are forging throughout the course of their program, and students can visualize the growing web of relationships at their disposal.

How – and how often – schools collect social capital data inevitably impacts their ability to act on it. As much as multi-dimensional data paints a richer picture of students’ connections, it comes at its own cost: the time it takes to survey students and aggregate and analyze their responses.

That’s led some programs to embrace brief surveys – sometimes called ‘exit tickets’ or ‘pulse surveys’ – to gain real-time understanding of whether relationships are or aren’t thriving within their programs. For example, Union Capital Boston (UCB), a nonprofit focused on community development, aims to help both young people and adults access economic opportunity through civic engagement. UCB offers participants access to this form of social capital through frequent in-person “Network Nights” that include, among other activities, a “Marketplace” where participants can request or offer help. UCB uses surveys to measure participants’ access to what the program dubs both “social supports” and “social leverage.” Through Network Night exit ticket surveys, UCB asks participants about the nature of the networking experience and the extent to which exchanges or reciprocity took place, including: “What were your emotions tonight at Network Night? (Happy, Shy, Lonely, Inspired, Bored)” and “Did you participate in Marketplace tonight?”

A few schools are taking a similar tack, in particular to understand how students are forging relationships with mentors in the course of high school internships. Using an app developed by Big Picture Learning called ImBlaze, one school program, Ken-Ton Big Picture, used the app’s dynamic survey functionality to ask students three questions repeatedly over the course of a year to gauge their comfort and level of connection to their on-site mentor.

A dynamic approach can mitigate lengthy surveys and also yield more reliable, actionable relationship data.

Although few of the survey items and measures that we documented have been validated, there are also more evidence-based measures designed for schools and youth-serving nonprofits starting to emerge. For example, with support from the Bill & Melinda Gates Foundation, the Search Institute, a nonprofit research organization focused on youth development, created a free assessment called the Social Capital Assessment + Learning for Equity (SCALE) Measures. Learning measurement platforms, such as Hello Insight, have also begun integrating social capital and peer network measurements within their social-emotional learning assessments.

The AI revolution should prompt schools to take a closer look at measures like these. There’s ample reason to be hopeful about the power of AI to personalize – and humanize – education by unlocking more time for weary educators. But realizing that possibility will require schools to better organize around relationship-oriented measurement.

Taken together, these examples and tools addressing schools’ missing metrics mark a promising frontier for schools to improve: ensuring that teacher and staff time, in fact, goes back into the human side of school.

Taking the long view: safeguarding against loneliness

Looking ahead, these metrics aren’t just critical to ensuring that schools redirect time back into connection. Keeping a close eye on relationship outcomes may also prove our best safeguard against AI cannibalizing students’ authentic connections altogether.

Young people today are already in the throes of what Surgeon General Vivek Murthy has dubbed a troubling loneliness epidemic, brought on in no small part by countless hours spent online. AI could take that epidemic to a new extreme.

AI tools like ChatGPT can simulate relationships. The bot builds rapport through conversation, and can nurture frictionless connection and intimacy with a machine. This marks a sharp departure from previous generations of social media technology.

Young people growing up amidst AI may be more tempted than ever to turn to technology not merely for entertainment or distraction, but for a deep – yet simulated – sense of connection.

Although strategic philanthropy advisor and AI expert John Bailey has written about the enormous potential of AI in schools, he’s also cautious about the allure of simulated relationships. “Kids will want the affirming relationships that they can have with their AI system,” Bailey said in an interview. “That sounds like science fiction until you experience the technology.”

Others are echoing Bailey’s concern. As researcher Daniel Cox has written, today’s AI tools, with all their promise of productivity, are already starting to exact a human cost, especially on Gen Z. Turning to AI for connection can jumpstart a vicious cycle; disconnection breeds more disconnection. “When we spend less time with each other, we lose practice in getting along in shared spaces,” Cox wrote. “This is why AI is such a poor substitute for real-world interactions. We need to spend more time with each other.”

What does all this mean for our schools? Although AI assistants could prove a breakthrough time saver for teachers, AI’s less productive applications could be an engine for disconnection among young people themselves. To avoid that dystopia, schools – and arguably myriad other youth-serving institutions like community centers, summer programs, and extracurricular providers – will need metrics that measure the extent to which young people are building and maintaining authentic, human connections.

We shouldn’t wait on lagging data to show up in warnings like the Surgeon General’s. Proactively starting to measure young people’s relationships can keep us honest about the positive impact that AI is – or isn’t – having in their lives. Those metrics can help us to know, at more regular intervals, whether young people’s networks are growing or contracting; whether the quality of their human connections is deepening or deteriorating; and whether their muscle to interact with the peers and adults in their lives is strengthening or atrophying.

Without data points like these, we run the risk not only of perpetuating the flawed assumption that time freed up by AI is reinvested into time connecting, but also of relinquishing connection itself to machines.

Since ChatGPT’s dramatic rise, education debates have fallen into two familiar camps. Advocates are heralding the immense potential of AI to disrupt staid structures in schools. Skeptics are calling for the outright ban of tools in schools to protect against cheating, privacy violations, and isolation. What both sides miss is that how AI impacts our young people’s lives has far less to do with chatbots, and far more with the metrics by which schools measure their success.

By adopting more rigorous and regular relationship measures, schools can ensure that they are ultimately enlisting AI in ways that make them more human – and make their students more whole.  


About the Author:

Julia Freeland Fisher is the director of education research at the Clayton Christensen Institute and author of the book Who You Know: Unlocking Innovations that Expand Students’ Networks. She researches students' access to and ability to mobilize peer, mentor, and industry connections. She holds a BA from Princeton University and a JD from Yale Law School.