Artificial Intelligence Applied to Education

In Why Online Education Works I wrote:

The future of online education is adaptive assessment, not for testing, but for learning. Incorrect answers are not random but betray specific assumptions and patterns of thought. Analysis of answers, therefore, can be used to guide students to exactly that lecture that needs to be reviewed and understood to achieve mastery of the material. Computer-adaptive testing will thus become computer-adaptive learning.

Computer-adaptive learning will be as if every student has their own professor on demand—much more personalized than one professor teaching 500 students or even 50 students. In his novel Diamond Age, science fiction author Neal Stephenson describes a Young Lady’s Illustrated Primer, an interactive book that can answer a learner’s questions with specific information and also teach young children with allegories tuned to the child’s environment and experience. In short, something like an iPad combining Siri, Watson, and the gaming technology behind an online world like Skyrim. Surprisingly, the computer will make learning less standardized and robotic.

In other words, the adaptive textbook will read you as you read it. The NYTimes has a good piece discussing recent advances in this area including Bakpax which reads student handwriting and grades answers. Furthermore:

Today, learning algorithms uncover patterns in large pools of data about how students have performed on material in the past and optimize teaching strategies accordingly. They adapt to the student’s performance as the student interacts with the system.

Studies show that these systems can raise student performance well beyond the level of conventional classes and even beyond the level achieved by students who receive instruction from human tutors. A.I. tutors perform better, in part, because a computer is more patient and often more insightful.

…Still more transformational applications are being developed that could revolutionize education altogether. Acuitus, a Silicon Valley start-up, has drawn on lessons learned over the past 50 years in education — cognitive psychology, social psychology, computer science, linguistics and artificial intelligence — to create a digital tutor that it claims can train experts in months rather than years.

Acuitus’s system was originally funded by the Defense Department’s Defense Advanced Research Projects Agency for training Navy information technology specialists. John Newkirk, the company’s co-founder and chief executive, said Acuitus focused on teaching concepts and understanding.

The company has taught nearly 1,000 students with its course on information technology and is in the prototype stage for a system that will teach algebra. Dr. Newkirk said the underlying A.I. technology was content-agnostic and could be used to teach the full range of STEM subjects.

Dr. Newkirk likens A.I.-powered education today to the Wright brothers’ early exhibition flights — proof that it can be done, but far from what it will be a decade or two from now.

See also my piece with Tyler, the Industrial Organization of Online Education and, of course, check out our textbook Modern Principles of Economics which isn’t using AI yet but the course management system combines excellent videos with flexible computerized assessment and grading.



Now we just need to help kids with poor study habits.

My wife volunteers as a tutor for junior high students, along with others. To a person, she says, the kids are smart enough to do the work. But, because their families move from one rental apartment to another, one neighborhood to another, or from one parent to the other, the kids are very disorganized. And, their friends change from move to move.

Learning occurs in an environment.

One wonders why these bad environments don’t seem to prevent kids learning intricate details about, for example, sports.

One also wonders why the children of poor Asian immigrants in NYC systematically academically outperform the children of non-poor non-immigrant families. I suspect a more stable residential address is not the reason.

One wonders why these bad environments don’t seem to prevent kids learning intricate details about, for example, sports.

Because they care about those other things, and they don't care much about a lot of what they are supposed to learn in school. Which isn't all that surprising, because very little of it will be of use to them. (Except perhaps in a bootstrapping sense: "When will I ever use this?" "In the course you'll be required to take next year.")


I can see that you are unaware of the research, so here is a link:

Studies include long range study of Danish children.

These studies do not even match the frequency of moving these kids have.

Your question is eternal, or at least as old as 1880

"Charles Francis Adams, Jr., ...; and adds, "The imitative or memorizing faculties only are cultivated, and little or no attention is paid to the thinking or reflective powers. Indeed it may almost be said that a child of any originality or with individual characteristics is looked upon as wholly out of place in a public school. ... To skate is as difficult as to write; probably more difficult. Yet in spite of hard teaching in the one case and no teaching in the other, the boy can skate beautifully, and he cannot write his native tongue at all."*

* "Scientific Common-school Education." Harper's Magazine,
November, 1880"

I'm not altogether surprised that studies show students learn better from AIs than from human tutors, but I suspect that selection bias might account for much of it.

In the time that I spent as a math and science tutor, both to junior-high students and to college underclassmen, I found that a majority of my sessions involved a student who wanted me to do their homework quickly and with the minimum of fuss, so that they could go do something more interesting. The question I quickly came to dread was "Couldn't you just show me how to do the problem?"

Since AIs cleary wouldn't respond to that kind of badgering, they'd draw fewer visits from that kind of student. They'd seem more effective as teachers, just because they were working with a better class of students.

The one thing that schools, and so-called educators, won't teach students is how to do their job, i.e., study. They hector and badger, but do not train the vocational skill.

I thought this was just because it hadn't been discovered but then I found 'How to Study and Teaching How to Study (1909) by F. M. McMurry, Professor of Elementary Education, Teachers College, Columbia University'. It was illuminating. It was a oft cited book up until the 1920s, then it was seemingly lost to the so-called educators. If you read this early work, you will see the 'modern' education ideas are just plagiarized from these earlier ideas, either overtly or by re-invention of the wheel.

"True or logical study is not aimless mental activity or a passive reception of ideas only for the sake of having them. It is the vigorous application of the mind to a subject for the satisfaction of a felt need. Instead of being aimless, every portion of of effort put forth is an organic step toward the accomplishment of a specific purpose; instead of being passive, it requires the reaction of the self upon the ideas presented, until they are supplemented, organized, and tentatively judged, so that they are held well in memory. The study of a subject has not reached its end until the guiding purpose has been accomplished and the knowledge has been so assimilated that it has been used in a normal way and has become experience. And, finally, since the danger of submergence of self among so much foreign thought is so great, it is not complete — at least for young students — until precautions for the preservation of individuality have been included."

McMurry recommended starting the vo-tech training of students how to study in 3rd grade. The earlier grades devoted to the vocational skills of reading, writing, arithmetic.

Neal Stephenson deserves a lot of credit. Computers were used to automate drills and tests early on, but with the Young Lady’s Illustrated Primer he leaped far ahead.

It's a classic case of sci-fi creating the dream for technologists to follow.

Diamond Age was written back in 1995. That's the year Netscape went public, well before the dot com boom, and the big build-out of mass education resources.

Is reading "the irrevocable future" according to our preferences any or much different from reading "the dead past" according to our preferences?

Both look startlingly like projections into a single realm of non-existence.

Alex has here revealed his prophetic soul: "computer-adaptive learning will be AS IF . . .".

We are all now compelled to live in this "as if" world, arguably, a brand new form of totalitarianism (from technology henceforth NO ESCAPE can be permitted).

+1, Stephenson does deserve a lot of credit for fleshing out the concept. The concept is far older than what he wrote in 1995. Heinlein wrote about it in the "Moon is a Harsh Mistress" from 1966. But Stephenson sort of fleshed the concept out in a far more detailed way.

The tutoring in Stephenson's novel is NOT simply AI. Another character is an actor/tutor hired to provide the book's voice to the girl being taught. She continues in that role for years, so rather than AI, the novel could be read as (partly) an argument for personal tutoring by another human, perhaps assisted by AI.

Kal El learned using an AI.

Don't give away the gag. For most of the way through the book we are told it is a computer program (written by her dead dad?). The switch to a human behind the mask was a second MacGuffin in a way.

Either Stephenson was expressing pessimism about the sci-fi potential of AI, *or* he realized that if it really was an AI it could scale to many more users than one. The "girl's army" would not have been in sync. That's a different story.

On a larger sense when AI surpasses human intelligence (some decades into the future, hopefully), there wouldn't be much point in humans learning anything, since AI would do it better. Schools and universities would go bankrupt presumably.

Yeah, I think in 15-20 years the whole concept of learning and vocations will probably be quite different. You'll only bother learning what interests you. The introduction of AI-Brain interfaces 20 years after that will mean learning is more akin to Neo's Kung-Fu downloads.

The whole idea of school is that it doesn't matter what interests you. It has been decided that you should learn X and Y and ... That is what you will be tested on and that is what will determine whether you get that certification at the end-- say a high school diploma--which will make a big difference in your future.

The point is if highschool will atill be considered useful in a world where

A high school diploma is not useful because it certifies you have acquired a lot of useful knowledge; you have probably forgotten the vast majority of what your teachers tested you on. It is useful because it certifies that you meet a certain standard of intelligence, conscientiousness, and willingness to follow the rules.

In the future, brains will matter as little as brawn.

Physical strength used to matter before mechanical automation. Now machines outperform everyone. Cognitive abilities still matter today, but within one or two generations, AI will outperform everyone.

And that's how all the brouhaha will go away, all of today's debates and arguments. Meritocracy will be irrelevant because no human will have merit.

AI will still help us learn, in the same way that we move a red laser-pointer dot around so that cats can chase after it.

If we all receive the same stipend per month, within a generation or two we'll be back to inequity.

Interesting to think that ignorance betrays specific assumptions and patterns of thought in terms of incorrect answers.

So that is it. We must become slaves of the machines we created and we once lorded over. Even the holiest human activities must be subjected to machines. Nothing gold can stay.

Don't be such a drama queen. Of course you will be a slave to the machines - but you will like it!

No, I won't.

don't worry, they'll create a Robert Frost AI to write poems to cheer you up

A Robot's Will

I do not care. I won't relent.

Testing, in tech anyways, has been adaptive for several years. It seem like a logical next step for instruction as well.

For people who WANT to learn something, and who will then USE that something, this stuff may be transformative. But for most high school students in most subjects, it will just be another way to learn and forget, memorize enough to pass a test and then let the knowledge decay because you are not using it and don't care about it.

What if students are learning about as much as they can retain?
Then the better approach is to focus on the more valuable knowledge and skills.

What a useful and totally unacceptable idea.

Eh, color me a bit skeptical. I can easily imagine customized learning modules or tutorials helping people learn stuff quicker and perhaps make marginal improvements in how far students progress in a particular discipline, but it's not like it's going to help somebody with a 95 IQ learn differential calculus. Comparisons to the Wright Brothers seems like hype.

Other way around. The humans are teaching the bots.

I'll be retired before anything much like this is rolled out (really, the web based tools I've seen so far are feeble ever for survey level historical subjects). Just imagine how much money Pearson is going to make off of this sort of thing!

It's software based. It can actually roll out to the private market pretty quickly. The period between the first mass market cell phones and the first successful smart phones was less than 15 years.

Granted, the US education system probably would resist any such changes for decades, but that wouldn't stop parents from side stepping the process if the software was cheap for the results.

Just think: by the time one or two cohorts have had their primary educations overseen by AI, they'll be such brainiacs they'll have no need for post-secondary training after ages 13 or 14.

All the money formerly dedicated to "post-secondary training" can then be funneled into AI until our brainiacs can be fully credentialed at ages 7 or 8.

(In other intellectual domains this is called "front-end loading".)

Sounds as if the days of the Chronicle of Higher Education are numbered, also.

You seem to jest, but this will cause some students to pull ahead causing further 'inequity'. There will be much gnashing of teeth and calls for justice.

Sure, it's not hard to imagine how AI can help improve human capital. But I don't see how AI is going to improve my ability to signal, which is what really matters in education.

start-up idea: an AI that predicts the highest status hobbies and fashions for your tween given their age and zip code

True. If anything it might water down the signal a bit. We'll just have to compensate by making it cost more or something.

"But I don't see how AI is going to improve my ability to signal, which is what really matters in education."

HR response: Look for the key phrase "AI tutored" on the resumes, dump the rest.

For all the direct and implicit hits on college above, this quote is pertinent (can be found in a few articles today or yesterday):

“You know, I majored in economics,” Belichick said on "Belichick & Saban: The Art of Coaching." “I’d struggle to say that’s really helped me. I think the biggest thing that I learned in college wasn’t the material, it was how to solve problems and how to think, and how to come up with your own idea and solution to the problem. That’s really what our job is.”

I usually agree with Alex, but AI and online ed are being grossly oversold. They are more about textbook publishers wanting to prevent the sale of used textbooks than they are about true learning.

Correct, Alex seems to have been caught in a time-warp back to 2012 when (way too many) observers thought that MOOCs and other online education technologies were going to take over education.

The key problem is motivation. Those students who are daydreaming in class or looking at their phones? At least they're in a classroom where the entire environment is about learning. Plop an average 18-year old in front of a computer and expect them to learn calculus? They'll spend the time watching cat videos or TikTok.

Programmed instruction has been around for decades. It's a nice approach, as is the use of computers and AI, but so was the really revolutionary educational technology: the book. And as vital and important as books and writing are, they didn't cause the shutdown of schools and classrooms either.

AI learning looks appealing at first blush, with interaction and midcourse adjustments to tease out those experiences. What isn't discussed is how that is subject to manipulation both in terms of content and eventual monetization. Both topics should scare sensible people.

teachers will hate it.

Unfortunately, this still emphasizes what Paul Graham recently advised as 'The Lesson to Unlearn', that is school trains students to pass tests to the detriment of real learning.

"Getting a good grade in a class on x is so different from learning a lot about x that you have to choose one or the other, and you can't blame students if they choose grades. Everyone judges them by their grades —graduate programs, employers, scholarships, even their own parents.

"I liked learning, and I really enjoyed some of the papers and programs I wrote in college. But did I ever, after turning in a paper in some class, sit down and write another just for fun? Of course not. I had things due in other classes. If it ever came to a choice of learning or grades, I chose grades. I hadn't come to college to do badly."

Comments for this post are closed