AI Tutors, Inequality, and the New Education Production Function

Could AI be a new tool to increase equity in education, allowing students from marginalized or low-income backgrounds to compete with students who enjoy greater resources and privilege?  Educators are bracing themselves for the upcoming impacts of AI, with many seeing it both as a powerful tool and as a negative disruptor.  Popular AI sites like ChatGPT, Grok, Claude, and Google Gemini allow both teachers and students to generate lengthy, polished responses to queries, including generating images and video  clips, very quickly.  

Educators can use AI to rapidly generate lesson plans, instructional slides, plan activities, and conduct research.  Students can use AI to find sources of information more quickly than ever before…or write the entire paper for them.  For both educators and students, is AI helping them access information and use it more effectively, or replacing their ability to think critically?

History of Technological Evolutions in Public Education

AI is not the first major technological innovation to reshape education and student learning.  Over the past generation, several computer advancements have given students increasing access to educational material.  Around 2000, the rise of high-speed cable Internet allowed students to reliably access online information and communication tools.  Soon afterward, educators began utilizing their own digital platforms to host educational materials, deliver assignments and assessments, and record student grades and submitted artifacts.  

Learning Management Systems (LMS)

In 1997, just as high-speed broadband Internet emerged, Blackboard hit the scene as the first LMS for education (as opposed to corporate training).  Other LMS software quickly came to rival it, such as Moodle, Canvas, and Google Classroom.  Colleges and public K-12 schools could have instructors post class materials online, and LMS software regularly updated to allow for more interactive options.  Today, most classes in all schools have an LMS component, with students expected to log in and complete some work digitally.  Many colleges have courses today that are entirely online and conducted via LMS, with instructors able to post videos and conduct interactive video lectures or conferences with students thousands of miles away.

One-to-One Devices

Of course, many students were not able to take advantage of the convenience of LMS-posted learning materials due to lack of computers at home.  During the 2020 Covid pandemic, most school districts were able to access vast government resources to provide students their own (one-to-one) computer devices, such as iPads or Chromebooks.  Five years after the pandemic, most school districts in economically developed areas allow each student their own computer device to access classes online.

Virtual Instruction

The immediate need for one-to-one devices was the fact that the Covid pandemic was forcing education to be done remotely.  Virtual instruction was almost universal between March and May of 2020, requiring students to get all of their education online.  Out of necessity, both software and teaching styles rapidly evolved to make virtual instruction more palatable for students of all age levels.  Today’s educators are far more familiar with the tools needed for successful virtual instruction, such as shorter, faster-paced video clips, than those of previous generations.  

AI in Education: A Complement or a Substitute for Quality Teaching?

Previously, the field of economics automatically considered technological advancements to increase production capacity.  In terms of physical output, such as consumer goods, this made sense: new equipment operated faster and more efficiently.  However, education and developing human capital is a service with many subjective components and outputs.  Human learners will not respond proportionally to all changes.  Many observers have questioned whether new technology in education, including AI, is a beneficial complement or a questionable substitute for high-quality teaching.

Complement

AI can benefit learners when used as a complement to existing high-quality education, allowing teachers and students to access additional information or complete rote tasks more quickly.  For example, AI can more quickly comb through reams of data and present factual results than regular search engines, helping students get the answers they need faster.  It can also be used to help students generate images and sounds needed for creative projects.

Substitute

Unfortunately, AI can harm learners when used as a substitute for the learning process.  Many students use AI to commit plagiarism or cheat on assignments, as some AI software can quickly generate entire research papers in mere minutes.  Students can plug increasingly complex questions into AI and receive the correct answer directly, not having to go to an information source and do the work of learning and extrapolating.  In the long run, this can result in students who no longer know how to extrapolate information themselves and are totally reliant on AI to simply provide answers.

The Matthew Effect

So, does AI help low-income students compete more effectively against those with ample resources?  As with previous technological innovations in education, ranging from LMS-run MOOCs (Massive Open Online Courses) to one-to-one devices, there is likely not to be a major gain to low-income students versus their peers with AI.  These earlier innovations were heralded as ways to level the playing field by giving poor students equal access to information, but largely failed to deliver.  Instead, research found that these new tech innovations in education mostly benefited students who were already middle- or upper-class, exemplifying the Matthew effect.

Created by sociologist Robert K. Merton in 1968, the Matthew effect states that those who are already advantaged in terms of resources tend to automatically accumulate more resources from innovations.  Essentially, the “rich get richer.”  In education, in regard to technological innovations, this was found accurate due to wealthier students’ abilities to more effectively utilize advancements like MOOCs and one-to-one devices than their low-income peers.  For example, wealthy students had more free time with which to take online classes and had high-speed Internet with which to use their school-provided device.  Many low-income students, by contrast, struggled with scheduling conflicts (such as part-time jobs after school) and lack of Internet access in the home.

Policy Consideration of AI in Public Education

If the Matthew effect suggests that AI will actually widen the education gap by allowing wealthy students to more quickly reap the academic gains than their low-income peers, what policies could be put in place to help keep things equitable?

Target AI and Tech Tutorials Toward Lower-Income and At-Risk Students

One possibility would be to target classes that teach educational technology (ed tech) skills to low-income or at-risk students.  In middle and high school, counselors could steer students into these courses to receive instruction on how to use the ed tech tools that will be crucial in higher education and the white collar workplace.  This would help these students more ably compete with their wealthier schoolmates, who often receive this type of education at home, informally, from tech-savvy parents.  

Require Students to Handwrite Assignments in Class to Avoid AI Cheating

Some educators have gone back to old-fashioned basics to eliminate AI cheating, such as requiring students to write by hand while in the classroom.  This may seem counterintuitive in our tech-obsessed world, but it does force students to utilize their critical thinking skills instead of relying on AI to spit out a fully-formed answer.  Reducing the use of technology in education can also reduce the technological advantages enjoyed by wealthier students, such as faster Internet and computers at home.  If students are relying on paper books and writing by hand, one’s ownership of a more expensive computer is less relevant.