TOM DACCORD

View Original

News outlets report that teachers are "in danger" of being taken over by robots in the A.I. revolution. Are they right?

Recently, business news outlets have reported that teaching occupations are “the most in danger" of being taken over by robots in the A.I. revolution. Business Insider includes Teachers in its "top 10" list of occupations most likely to be replaced by artificial intelligence. Fortune Magazine is more specific, listing humanities and language teachers as among those most susceptible to being usurped by AI.

Many articles cite a study conducted by researchers from Princeton University that states that education occupations are more likely to be impacted by advances in Large Language Models (LLMs) like ChatGPT than most other professions. Out of the 774 occupations listed in the study, six education occupations are ranked in the top 20% of the most “exposed” to AI, and include Middle School and Secondary School teachers.

However, as the researchers point out, high exposure to AI does not necessarily mean that an occupation will be replaced by AI. In their study, “exposure to AI” could translate into either “substitution” or “augmentation”, depending on various factors. In other words, AI might actually help teachers do their jobs better, rather than replace them.

Moreover, the World Economic Forum (WEF) predicts “large-scale job growth” in education worldwide over the next five years. In its “Future of Jobs Report 2023”, the WEF foresees 10% employment growth between 2023 and 2027 for an education sector that includes librarians, primary school and early childhood teachers, secondary education teachers, special education teachers, and vocational education teachers. For its part, the U.S. Bureau of Labor Statistics expects employment of U.S. high school teachers to grow 5 percent from 2021 to 2031, and employment of kindergarten, elementary, middle school, and special education teachers to grow 4 percent during the same period. If so, employment growth for these teacher categories would be “about as fast as the average” for all occupations during this period.

That said, AI technologies are going to disrupt education considerably. Case in point, ChatGPT created “near panic” in schools and universities when introduced and educators are struggling to understand all the challenges it presents. (And ChatGPT comes in the “first inning” of what will be a long game of AI rollouts and advancements.) Moreover, both the WEF’s and the U.S. Bureau of Labor Statistics’ rosy teacher-jobs predictions are based on research undertaken before the advent of ChatGPT and other new AI technologies. So, their predictions stand on shaky footing.

What we do know is that the organizations polled by the World Economic Forum – including the education sector – said they were planning to incorporate more AI technologies. They estimate that 34% of all jobs-related tasks in their organizations are already being performed by machines and they expect that number to reach 42% by 2027.

So, if 42% of all school-related tasks are being performed by machines in 2027 wouldn’t that mean that teachers are losing their jobs to AI?

Not necessarily.

An Education Week study in 2022 found that a typical U.S. teacher works a median of 54 hours per week, but just 46 percent of their time in the school building is spent teaching. (Across countries, according to the OECD, teachers spend half of their working time in non-teaching activities) Teachers spend non-teaching time on planning lessons, grading, communication and record-keeping, among other things. And AI is able to undertake most of these non-teaching tasks. So, AI may simply save teachers time.

Here are non-teaching tasks that can already be fully undertaken by AI now:

  • Lesson planning and curriculum development (AI can generate lessons and suggest resources based on student data)

    • Example: CoPilot, Curipod, TeacherBot

  • Attendance tracking and reporting

    • Example: GeniusEdu, Attendance AI

  • Grading (for objective assessments such as multiple-choice tests) and assessment record-keeping

    • Example: Quizizz

  • Basic parent-teacher communication (e.g., sending automated messages about student progress)

    • Example: “Help Me Write” feature in Gmail and Docs

  • Content generation (AI can generate handouts, virtual lectures, demonstrations, or multimedia presentations)

    • Example: ChatGPT, beautiful.ai, SlidesGo

  • Monitoring student progress (through analysis of assessment data)

    • Example: Schoolytics

The AI systems mentioned all include some degree of human oversight, but taken together they can save teachers considerable non-teaching time. For many teachers, the tasks listed above easily comprise between 34%-42% of their time at work each week, if not more.

Many teachers would point out that grading subjective assignments, like open-ended questions and essays, is among their most time-consuming tasks, and it’s not included in the list above. Imagine a teacher who is teaching four classes of 25 students each and collects 100 two-page essays from students. If the teacher spends 2-3 minutes each page reading and writing feedback that teacher would likely spend between 6 and 10 hours grading those (short) student essays. That’s a considerable amount of non-teaching time. And it doesn’t take into account other types of assessment that the teacher might be undertaking simultaneously.

But AI is making great strides in grading and feedback systems for subjective assessments that can save teachers considerable time. These AI systems function as a teacher’s assistant and can be trained to grade subjective assessments. Some systems are guided by a teacher-created rubric for the assignment and/or samples of teacher feedback and can also be geared to specific curriculum requirements and teacher priorities. One such system is EnlightenAI and at the moment it’s completely free. EnlightenAI provides the teacher the opportunity to edit and validate any scores and feedback that its system generates and it works in tandem with Google Classroom.

In all, a prediction that 42% of all school-related tasks will be performed by AI by 2027 is not unreasonable. But that does not necessarily mean that robots will be teaching our children any time soon.

Or, does it?

Consider how AI is mastering all the curriculum content taught in schools and universities. For instance, GPT-4 has achieved “stunning results,” on standardized educational assessments, including the LSAT (88th percentile), SAT (Reading & Writing 93rd, Math 89th) and various APs (a ‘5’ in nine different AP exams). In addition, The Organisation for Economic Co-operation and Development (OECD) predicts that AI will soon completely master its test of adult literacy, numeracy, and problem-solving skills.

AI can also create effective educational content and deliver personalized lessons. AI systems like CoPilot can create entire units of educational lessons and activities. AI can also create entire multimedia presentations from a few inputted words. Moreover, AI can provide individualized learning experiences that address various student strengths and weaknesses and heighten both comprehension and engagement. These systems can also tailor the pace of content delivery and the style of teaching to the student needs. Imagine robots that can explain concepts in clear and concise ways that are easy for students to understand and provide useful examples and analogies to ensure that each student is learning. All the while, they adapt to each student's learning style and pace, providing personalized instruction and feedback.

Teachers point out that human-to-human interaction is at the core of learning. In an Education Week survey of K-12 teachers, 84 percent disagreed with the idea that student learning would likely improve if AI-powered robots were working with teachers as classroom assistants. In addition, greater than 90 percent disagreed that learning would improve where “chronically low-performing” human teachers were replaced by AI robots. These attitudes underscore a persistent concern that robots could undermine the foundational bonds between students and educators that are crucial for effective learning. They reflect a commonly held belief amongst educators that robots will “never” replace teachers since human interaction is vital to the educational experience.

This last sentiment, while noble, strikes me as wishful thinking.

For one, the American public certainly views teachers as ripe for automation. A 2019 PEW Research Center report indicated that Americans viewed teachers as the sixth most likely profession to be at risk of automation, following fast-food workers, insurance claims processors, software engineers, legal clerks, and construction workers — in that order. Furthermore, in the 2019 Phi Delta Kappan (PDK) Poll of the Public's Attitudes Toward Public Schools, revealed 60% of American adults believed that AI and robotics would have a significant impact on education, and a full 30% supported the idea of actually replacing teachers with robots.

Furthermore, students and their parents are increasingly open to being taught by machines. A recent study found that students who had experienced being tutored by both AI and humans overwhelmingly (85%) said ChatGPT was more effective. In the study, 39% percent of high school and college students surveyed said they have fully replaced tutoring sessions with ChatGPT, while 30% of parents of students surveyed say they have done the same for their children. In addition, a 2021 Frontiers in Education study found that students were “generally excited and positive” about the idea of being taught by robots.

And AI is increasingly able to understand humans. MIT and other institutions have made great progress in ”Emotion AI,” a subset of artificial intelligence that measures, understands, and reacts to human emotions. Robots are increasingly adept at interpreting faces and movements. They can recognize how voice inflections correlate with stress or anger and indetify subtleties in micro-expressions. And AI is making progress in reading our minds. Scientists at the University of Austin recently leveraged AI to translate the private thoughts of human subjects by analyzing scans that measure the flow of blood to different parts of the brain.

So, in the years ahead, couldn’t a machine be more adept than a human at preparing students for a standardized test, say an AP exam? Machines can analyze vast amounts of exam data, and quickly, and provide critical answers to test preparation questions: What questions are asked? How frequently? How are they phrased? Which topics appear most often? When do they appear? How much time should be spent studying on each individual topic? How should I go about studying each topic? What are the most effective strategies for preparing for the exam? What are the highest-scored responses? How are they constructed?—and so on. If curricula and instruction is formulaic, automation is easier. And what if robots could also identify what we understand and what we don’t understand by looking at us and listening to us?

There are challenges to the teaching profession ahead.

In the end, the question of whether teachers will be replaced by robots is too often framed as an “either-or” proposition, teachers or robots. The more likely scenario is that teachers will be forging collaborative partnerships with AI. AI will support educators by providing personalized learning content and recommendations, analyzing student performance data, and offering insights into optimal teaching strategies. Teachers will use this information to tailor their lessons and better address individual student needs, while still maintaining the human connection and empathy critical to fostering a supportive learning environment. By combining the analytical capabilities of AI with the relational skills of human teachers, humans and AI together can create a more effective and engaging educational experience for students. The question then becomes whether teachers are ready, willing, and able to partner with AI.

See this form in the original post