Will Machines Replace Human Teachers?

Could children get just as much value from a well-designed piece of tutor software?
Could children get just as much value from a well-designed piece of tutor software?

In this week's video, we offered a few conclusions about the future of education. One of them was that formal education and the student-teacher relationship aren't going to be replaced by technology anytime soon. (Or at least they shouldn't be - but there's no accounting for dystopia.) I want to develop this idea a little bit.

I agree that there's a lot of potential for technological supplementation of the education process. One example would be the Web.

Good information itself used to be rare. This seems kind of weird to us now, but just think about it: There was a time when if you wanted to learn about, say, medicine, you actually had to go to an expert or a school to collect pieces of information about the subject. There was no Internet, you probably had no access to books on the subject, and much of the professional knowledge wasn't even written down. You literally couldn't get the facts you needed about the human body and the treatment of disease, unless you physically went to a place of education - or did primary research yourself.

Now information is everywhere, and because of libraries and the Web, most anything you can learn you can potentially learn for free. It's obvious that the Web is a great resource - though my brief time as a college instructor convinced me that even the most tech-savvy of first-year college students usually need training in how to use the Web judiciously for research.

Another technological intrusion on traditional education is the idea of tutor software. Imagine that a student is working out a series of math problems - but instead of solving them on paper, she solves them on the surface of a smart tablet that's running an advanced personalized learning app. As she makes her way through the assignment, the software tracks her work, noticing the types of problems that are giving her the most trouble, any consistent mistakes she's making on the way to the solution, etc. Software like this has the potential to attack bad habits before they set in, by giving feedback on the student's work in real time. Now, a human teacher can do all of this too, but the advantage of a technological solution is that it isn't limited by time and space. There's only one teacher, and that teacher can only work closely with one student at a time. Technology can really help fill in the gaps in situations like this.

However, there are some people who think it's going to go way beyond this - that technology will soon replace flesh-and-blood teachers, and we'll get all of our learning done in, I don't know, glowing education pods that zap our brains with knowledge.

I'd like to toss up a few thoughts about why I think human teachers with human brains are essentially indispensable:

The student-teacher relationship helps the student learn how to interact with authority. This is NOT merely to say that it helps train kids to do what the boss tells them to do. Though it does do that, and learning how to follow instructions actually does kind of matter. What's more important is that it creates an important conversational model: When you're talking about a subject, and somebody in the room knows more than you do about that subject, you should stop talking and listen to what that person has to say. More than that, you should ask questions of that person when you don't understand. Which leads us to the next point...

Students with a human teacher get to practice asking questions. This is actually a more difficult skill than you might think. I remember when I was in second grade, seeing a classmate lash out at our teacher while we were doing in-class work on math. The little guy was stuck on something. He stood at the teacher's desk as she tried to help him, but he kept getting more and more frustrated, and eventually yelled at her, "You're not helping me!" But it obviously wasn't her fault - the problem was that he was having trouble explaining to her what it was he didn't understand. He didn't know the best way to ask for help. By working through that tough moment, he was getting practice in the human equivalent of using the "advanced search" option on a search engine: limiting variables, narrowing the scope of the question until he finally figured out how to get the information or advice that he needed.

The student-teacher relationship accommodates Socratic learning in a way that couldn't be achieved through technology without a sci-fi level of artificial intelligence. The Socratic method is - and I will fight to the death on this one - absolutely the best way to learn. In brief, it's a teaching model in which the instructor causes the students to arrive at the point of the lesson themselves, by posing questions to the students and leading them to make their own discoveries. I can attest to this having been both a student and as a teacher: Students get much, much more out of a lesson when they arrive at conclusions themselves instead of simply being told. They remember the contents of the lesson better, and they walk away with more intellectual confidence and usually more interest in the subject.

Human teachers can be funny. Straight-up: Most of the classes I remember best from college, the ones in which I deeply internalized lessons and changed my own thinking, were the classes that induced laughter.

Can a machine tutor actually be funny? Believe it or not, simulating a good sense of humor is one of the hardest problems in AI. This NYT article from January shares a few earnest attempts from a piece of joke-generating software called STANDUP (System to Augment Non-Speakers' Dialogue Using Puns). Here ya go:

Q: What do you get when you cross a fragrance with an actor?

A: A smell Gibson.


Q: What do you call a fish tank that has a horn?

A: A goldfish bull.

I leave you to make your own conclusions.

At the college level, human professors demonstrate how an expert in any given subject behaves. If you're aspiring to become a bacteriologist, you should learn not only the subject matter that bacteriologists focus on, but the conversational modes of modern bacteriology. How do experts talk about their subject? What kinds of questions do they find interesting? In this way, one of the key aspects of a formal education is not only the creation of a knowledge base, but a kind of intellectual socialization into an expert community.

Human teachers can nurture creativity. I can see how a solid piece of tutor software could help a student solve math problems or even develop grammar or vocabulary. But that's because these exercises involve pre-programmed values and machine-readable logic. In other words, software is very helpful in developing sequential, deterministic skills. But what about the parts of education that ask students to make leaps of thought that are unpredictable? Contrary to popular myth, Albert Einstein was extremely good at math, and it's feasible that he could have attained scary excellence in rote computation, geometrical proofs, algebra, etc., just as well if he had been trained on adaptive tutor software. But Einstein's real contribution to modern physics wasn't in his computational ability; it was in his creativity. The original formulation of special and general relativity required not just the competent execution of familiar steps, but unpredictable innovations in thought. These creative shifts in thinking cannot be taught systematically, but seem to be more the product of inspiration by exciting work and challenging human peers - or teachers.

This is just a quick round-up, and there are probably lots of other ways that human teachers are irreplaceable, barring exponential advances in artificial intelligence. Disagree with any of these? Or can you think of any others?