Is AI Finally a Way to Reduce Higher Ed Costs?

New technology has often changed work. The Industrial Revolution did not eliminate weavers, brewers and farmers, but it did change core aspects of their work while increasing productivity and efficiency. For those who could master the new machines, wages increased, and for those who owned the machines, costs went down and the volume of production went up. More recently, countless innovations—calculators, email, the internet, spellcheckers, spreadsheets, citation software and learning management systems—have all changed how higher education faculty and staff work, but they have done little to increase salaries or lower student costs. Supporting these technologies, of course, added costs which may have cancelled out any savings.

Today, professional jobs that were immune to earlier shifts in physical labor are already seeing disruptions from artificial intelligence (AI). Indeed, one new research study found that AI could already “match or exceed human accuracy in determining legal issues” (at a significantly lower cost), concluding “these results are not just statistics—they signal a seismic shift in legal practice.” Similarly, a Google study has found that software bug fixes can now be identified and fixed more quickly using AI.

Jobs consist of groups of tasks. AI can’t do everything better, but almost every job has some tasks that AI can already do better. The ability to outsource tedious tasks (either to others or machines, like a spellchecker) often makes us happier and more efficient, and it is no surprise the same has already been demonstrated when we use AI well.

Most Popular Stories

Most Popular

Despite the promises and hype of previous technology, television, satellites, the internet, online courses and MOOCs all failed to lower higher education costs. All of them made it easier and faster to distribute content, but none of them helped us scale the feedback and relationships that form the core of higher education. Faculty labor has remained central and impossible to scale.

AI offers different possibilities.

1. AI can relieve faculty from some tedious tasks and do it at scale.

Here are some (perhaps tedious) tasks with (much abbreviated) prompts that might allow AI to give you back some time and even make you happier or better at your job. Recognize that you will get significantly better results with a more advanced (i.e., paid) AI (GPT-4, Claude 3 or Gemini Advanced) and that many of these require an AI connected to the internet (like the free Copilot or Perplexity).

Course Design

• Provide me with three sample policies I might include on my syllabus regarding how AI might be used by students in my X course.
• How could I make this syllabus/assignment more inclusive?

Content and Pedagogy

• Find me five relevant videos appropriate for college students on A that are B minutes in length and give me a 75-word summary for each that includes its content, reliability and source.
• Provide five ideas for how to introduce college students at X university to topic Y using examples or analogies they will find relevant.

Grading

• Create a rubric in table form to assess the learning in this assignment using these learning outcomes. List criteria in the first column and then provide descriptions in subsequent columns for poor, fair, good and excellent performance.
• Provide grades and feedback for these essays. Use my rubric, previously graded essays and samples of my feedback to calibrate your feedback to write and grade in my voice.

Assessment and Accreditation

• Evaluate these essays using rubric A and assess what percent of essays meet the B standard.
• Write my departmental accreditation report using these formats, guidelines and data.

Research

• Create a literature search/bibliography/list of articles on topic A, using methodology B, with sample size C that disputes claim D. (Consensus and Elicit are examples of AI that search only the published articles in the Semantic Scholar database to give you verifiable results.)
• Who are the other major figures in this field who might be potential reviewers of this article? What work of theirs should I be sure to cite?

Grant Writing

• Read these emails/strategic goals and advise me how to make my funding request more compelling to my provost.
• Transform this research brief into a compelling proposal for the X foundation using its format and guidelines.

Job Search

• Pretend you are faculty member X on this search committee for a new dean. Read the uploaded position description, my cover letter and CV. How might the committee or X react to my materials? List missing elements and suggest ways for me to improve my application.
• Act like professor Y on this search committee. Help me prepare for my interview by using the attached materials and ask me a series of potential questions that will challenge me. Include questions with inaccurate information and require me to correct you with real data.

Student Support

• Write a reference letter for student A using these materials. Begin by reading my previous letters and describing the elements of my voice and form.
• Act like student X using these details and context. Help me practice for a conversation with X about Z.

Administration

• Produce three spreadsheets with three different versions of the department schedule using this data on faculty preferences, room availability, student demand and X.
• Write a kind and personal 75-word letter of rejection for authors whose manuscript has been rejected by the Journal of X.

More context, process and iteration will greatly improve the results of these prompts. For example, a larger set of essays with your previous feedback is critical for better results and for training AI (and for greater privacy and security, you can download an opensource AI to run and fine-tune on your laptop). There are a rapidly increasing host of bots that will do specialized tasks and these functions are already creeping into your LMS. But what will you do with the time that you were spending grading or writing reports?

2. AI can improve relationships.

AI presents opportunities not just to reduce costs or improve efficiency. Consider this scenario:

A student has come into your office needing an approval to take an overload. Assuming you remember your password and which system to use, you can probably complete this task, but do you remember the student’s hometown? Now imagine, you have an AI assistant that recognizes the student and immediately organizes the relevant information and prompts you to ask about the student’s dog. The AI listens to your conversation and prompts you that the student will need Statistics 101 to complete her major or might like the new basketball club. These are among the opportunities that are now ahead of us.

In emerging studies regarding AI as an assistant, it is the least experienced who benefited the most from AI support. That is hardly a surprise, but neither is the complaint that faculty, whose job requirement entails five years of mostly solitary confinement, are not responsive enough, or trained enough, to build the relationships we now know are so vital for student success. As one researcher put it: AI is “like Grammarly for empathy.”

AI can now transcribe, organize and fact-check meeting notes while also showing who talks or interrupts the most (a feature now in Teams and Zoom!). An AI companion gets all of your references but can also determine when a student is in distress. This strange and creepy new world is here, but who better than faculty to help us understand and test where it can be useful or harmful? Further, AI is already reducing costs in a plethora of sectors, and faculty need to be on the front lines of discovering if and how AI might help us combat high tuition.

There are many tasks where AI will never be able to replace humans, and we should focus on them. Faculty have rushed to say that human teaching will never be replaced with AI. True—and the need for human faculty will still be the primary driver of costs in higher education. Still, AI may finally be the technology that gives faculty more time and more assistance for the most important educational and relational tasks.

For now, try some of the tasks and prompts above, but when you don’t get results you like, don’t assume AI is useless. Try to ask a better question with more detail, complexity and specification (and perhaps pay for a significantly better AI).

The abacus, thesaurus and computers have all provided aids to human thinking, but none of them replaced the educational priorities of questioning assumptions and asking better questions. AI will change how humans think and has the potential to strengthen the ways we interact. Like the internet, it will change some of the skills we need, but it won’t change the need for thinking with this new technology.