For years, the drive to collect data on which to base nearly every decision at colleges and universities—and equally at every corner of our economy—has been a national obsession. Gripped by its results, many find its conclusions sacrosanct, unquestioned, as if its word is handed down by digital gods. 

“Numbers don’t lie,” they say.   

But the foundations of our confidence are often based on shaky input, sources that may be suspect, gathered by researchers whose beliefs and backgrounds heavily influence questions they ask and citations they ignore. Some call this failure institutional bias, which influences results unintentionally or may even be mendacious.

As a former dean at NYU, I’ve often sat on academic panels that initiate data collection, designing questionnaires to uncover college trends. Other members are usually drawn from the same elite pool—scholars, faculty and administrators. Some are old friends whom I meet from time to time at one academic event or another—coffee at a conference in Orlando, delivering a joint scholarly paper, joining one another on an academic board, recommending a colleague for an award. So comfortable are we with each other, our thoughts often channel pretty much along similar lines. 

Only in retrospect have I come to realize that crucial decisions on which questions to ask and which to omit may not cover the interests of those we aren’t close to. While there is often a good mix of men and women, others are often missing or poorly represented. The essential voice of students is often not heard. Others are faculty and staff from lower-status academic disciplines. And, crucially, as often happens in the U.S. and elsewhere, members of ethnic and racial minorities—Black, Hispanic, South Asian and others, as well as people with disabilities—are frequently uninvited.

My colleagues, all serious and committed academics, work with determination to generate high-quality results, hoping to meet the needs of faculty and students across the nation for solid answers. Their feeling of academic responsibility is of the highest standard. But what essential data might not get reported because the right questions are not asked? What datasets might be generated if students played a central role? What questions would Black faculty raise? Or Hispanic and other ethnic minorities? What questions are suppressed because none of us have the background or experience of those who are outside our circle?

During a decade as online dean, each month I’d attend institutionwide discussions, exploring best ways of moving forward with digital education. A couple of dozen expert colleagues, drawn from various aspects of the school—teaching and learning, technology, finance, marketing, among others—sat around a giant, polished conference table in a handsome campus room overlooking a leafy park. Our discussions covered a long list of topics that colleges and universities must consider about online programs—support services, course design, digital tools, advising, tuition, plus a cascade of others. In all those years, students were never invited, and I don’t recall ever seeing a Black colleague pull up a chair to join us at our table.

Just as with my experience in other academic settings, I now wonder how thorough we were in our deliberations, recognizing that no student and not one Black colleague was present. What issues never got raised? What programs never got considered? What training was never suggested? What critical investments were never made? What questions never came to mind because our academic and life experiences were divorced from those who were excluded? 

How much faith should you have in data generated at your institution when insights from ignored colleagues are missing from the ranks of those who decide what we must know and what we may never learn?

On a wider higher ed landscape, unintended bias can result in serious consequences. For example, an article published recently in the journal of the American Educational Research Association reveals that widely used algorithms at many colleges to evaluate college applicant potential can block African American and other minorities from being accepted because the tools used are riddled with racist historical data. 

“These models are significantly more likely to predict failure for Black and Hispanic students who actually succeeded and less likely to predict success for students who failed compared to White and Asian groups,” researchers found. 

When data collectors lack diversity, results often suffer from massive blind spots. High-status groups often design questions and build data collections that reflect their own goals, not those of the wider student body or broader faculty. 

Serious moves by the current administration banning diversity training and courses and prohibiting identity-based hiring, among other brutal actions, are central to the present assault on higher ed. Yet, with the deepest of ironies and even acknowledging official deliberate ignorance, years of close study reveal that ethnic diversity is required to achieve high-quality scholarship. One major paper (Nature Communications 9, 5163 [2018]), aimed at gauging the effects of diversity, analyzed more than nine million papers and six million scientists to uncover the relationship between research impact and ethnic diversity among other key elements—discipline, gender, affiliation and academic age.

“Remarkably,” the study concludes, “of the classes considered, ethnic diversity had the strongest correlation with scientific impact.”

As AI moves relentlessly forward, those who plug in sources that underlie answers everyone is looking for must be aware that crucial information may be missing. With AI occupying so much of our thought, basic questions about who collects inputs occupy the far-distant wings of our national drama. Doubts raised by critics who claim that materials feeding giant digital repositories can be falsely misleading are not at the center of what concerns us.

Instead, we worry about whether AI has feelings, when it might totally outsmart us, how it might invade our privacy, compromise the environment, spread deepfakes and, scariest of all, how it might inadvertently put an end to all humankind. In the meantime, we must be vigilant about who collects data at our colleges and universities and who, shamefully, might be left out. 

Robert Ubell is vice dean emeritus of online learning at New York University’s Tandon School of Engineering and former editor of the New York Academy of Sciences monthly, The Sciences, and American publisher of Nature. A collection of his essays on virtual education, Staying Online: How to Navigate Digital Higher Education, is published by Routledge.

Next Story

More from Learning Innovation