The local name for the Philadelphia neighborhood of Kensington is “the Badlands,” and with good reason. Pockmarked with empty lots and burned-out row houses, the area has an unemployment rate of 29 percent and a poverty rate of 90 percent. Just a few miles to the northwest, the genteel neighborhood of Chestnut Hill seems to belong to a different universe. Here, educated professionals shop the boutiques along Germantown Avenue and return home to gracious stone and brick houses, the average price of which hovers above $400,000.

Within these very different communities, however, are two places remarkably similar in the resources they provide: the local public libraries. Each has been retooled with banks of new computers, the latest software and speedy Internet access. Susan B. Neuman, a professor of early childhood and literacy education at NYU, and Donna C. Celano, an assistant professor of communication at LaSalle University in Philadelphia, spent hundreds of hours in the Chestnut Hill and Badlands libraries, watching how patrons used the books and computers on offer.

The two were especially interested in how the introduction of computers might “level the playing field” for the neighborhoods’ young people, children of “concentrated affluence” and “concentrated poverty.” They undertook their observations in a hopeful frame of mind: “Given the wizardry of these machines and their ability to support children’s self-teaching,” they wondered, “might we begin to see a closing of the opportunity gap?”

Many hours of observation and analysis later, Neuman and Celanano were forced to acknowledge a radically different outcome: “The very tool designed to level the playing field is, in fact, un-leveling it,” they wrote in a 2012 book based on their Philadelphia library study. With the spread of educational technology, they predicted, “the not-so-small disparities in skills for children of affluence and children of poverty are about to get even larger.”

Neuman and Celano are not the only researchers to reach this surprising and distressing conclusion. While technology has often been hailed as the great equalizer of educational opportunity, a growing body of evidence indicates that in many cases, tech is actually having the opposite effect: It is increasing the gap between rich and poor, between whites and minorities, and between the school-ready and the less-prepared.

This is not a story of the familiar “digital divide”—a lack of access to technology for poor and minority children. This has to do, rather, with a phenomenon Neuman and Celano observed again and again in the two libraries: Granted access to technology, affluent kids and poor kids use tech differently. They select different programs and features, engage in different types of mental activity, and come away with different kinds of knowledge and experience.

The unleveling impact of technology also has to do with a phenomenon known as the “Matthew Effect”: the tendency for early advantages to multiply over time. Sociologist Robert Merton coined the term in 1968, making reference to a line in the gospel of Matthew (“for whosoever hath, to him shall be given, and he shall have more abundance: but whosoever hath not, from him shall be taken away even that he hath”).

In a paper published in 1986, psychologist Keith Stanovich applied the Matthew Effect to reading. He showed that children who get off to a strong early start with reading acquire more vocabulary words and more background knowledge, which in turn makes reading easier and more enjoyable, leading them to read still more: a virtuous cycle of achievement. Children who struggle early on with reading fail to acquire vocabulary and knowledge, find reading even more difficult as a result, and consequently do it less: a dispiriting downward spiral.

Now researchers are beginning to document a digital Matthew Effect, in which the already advantaged gain more from technology than do the less fortunate. As with books and reading, the most-knowledgeable, most-experienced, and most-supported students are those in the best position to use computers to leap further ahead. For example: In the Technology Immersion Pilot, a $20 million project carried out in Texas public schools beginning in 2003, laptops were randomly assigned to middle school students. The benefit of owning one of these computers, researchers later determined, was significantly greater for those students whose test scores were high to begin with.

Some studies of the introduction of technology have found an overall negative effect on academic achievement—and in these cases, poor students’ performance suffers more than that of their richer peers. In an article to be published next month in the journal Economic Inquiry, for example, Duke University economist Jacob Vigdor and co-authors Helen Ladd and Erika Martinez report their analysis of what happened when high-speed Internet service was rolled out across North Carolina: Math and reading test scores of the state’s public school students went down in each region as broadband was introduced, and this negative impact was greatest among economically disadvantaged students. Dousing the hope that spreading technology will engender growing equality, the authors write: “Reliable evidence points to the conclusion that broadening student access to home computers or home Internet service would widen, not narrow, achievement gaps.”

Why would improved access to the Internet harm the academic performance of poor students in particular? Vigdor and his colleagues speculate that “this may occur because student computer use is more effectively monitored and channeled toward productive ends in more affluent homes.” This is, in fact, exactly the dynamic Susan Neuman and Donna Celano saw playing out in the libraries they monitored. At the Chestnut Hill library, they found, young visitors to the computer area were almost always accompanied by a parent or grandparent. Adults positioned themselves close to the children and close to the screen, offering a stream of questions and suggestions. Kids were steered away from games and toward educational programs emphasizing letters, numbers, and shapes. When the children became confused or frustrated, the grown-ups guided them to a solution.

The Badlands library boasted computers and software identical to Chestnut Hill’s, but here, children manipulated the computers on their own, while accompanying adults watched silently or remained in other areas of the library altogether. Lacking the “scaffolding” provided by the Chestnut Hill parents, the Badlands kids clicked around frenetically, rarely staying with one program for long. Older children figured out how to use the programs as games; younger children became discouraged and banged on the keyboard or wandered away.

These different patterns of use had quantifiable effects on the children’s educational experiences, Neuman and Celano showed. Chestnut Hill preschoolers encountered twice as many written words on computer screens as did Badlands children; the more affluent toddlers received 17 times as much adult attention while using the library’s computers as did their less privileged counterparts. The researchers documented differences among older kids as well: Chestnut Hill “tweens,” or 10- to 13-year-olds, spent five times as long reading informational text on computers as did Badlands tweens, who tended to gravitate toward online games and entertainment. When Badlands tweens did seek out information on the Web, it was related to their homework only 9 percent of the time, while 39 percent of the Chestnut Hill tweens’ information searches were homework-related.

Research is finding other differences in how economically disadvantaged children use technology. Some evidence suggests, for example, that schools in low-income neighborhoods are more apt to employ computers for drill and practice sessions than for creative or innovative projects. Poor children also bring less knowledge to their encounters with computers. Crucially, the comparatively rich background knowledge possessed by high-income students is not only about technology itself, but about everything in the wide world beyond one’s neighborhood. Not only are affluent kids more likely to know how to Google; they’re more likely to know what to Google for.

Slogans like “one laptop per child” and “one-to-one computing” evoke an appealingly egalitarian vision: If every child has a computer, every child is starting off on equal footing. But though the sameness of the hardware may feel satisfyingly fair, it is superficial. A computer in the hands of a disadvantaged child is in an important sense not the same thing as a computer in the hands of a child of privilege.

The focus of educators, politicians, and philanthropists on differences in access to technology has obscured another problem: what some call “the second digital divide,” or differences in the use of technology. Access to adequate equipment and reliable high-speed connections remains a concern, of course. But improving the way that technology is employed in learning is an even bigger and more important issue. Addressing it would require a focus on people: training teachers, librarians, parents and children themselves to use computers effectively. It would require a focus on practices: what one researcher has called the dynamic “social envelope” that surrounds the hunks of plastic and silicon on our desks. And it would require a focus on knowledge: background knowledge that is both broad and deep. (The Common Core standards, with their focus on building broad background knowledge, may be education’s most significant contribution to true computer literacy.)

It would take all this to begin to “level the playing field” for America’s students—far more than a bank of computers in a library, or even one laptop per child.

This story was produced by The Hechinger Report, a nonprofit, nonpartisan education-news outlet based at Teachers College, Columbia University.