Hacker News new | comments | ask | show | jobs | submit login
Why Teachers Won't Be Replaced By Software (trinket.io)
23 points by eah13 on Feb 14, 2014 | hide | past | web | favorite | 60 comments

This "fire bad teachers" thing is stupid.

You can't compare a teacher that has a class of 15 native english speaking, upper middle class, 2 parent house holds kids to a teacher with a class of 32 english as a second langurage, poor, hungry, cold, kids that are children of drug addicts, drunks, refugees especially using test written in English.

Mark said something like poor comunities have a problem with bad teachers. WTF? How about poor students have a problem learning regardless of who the teacher is.

software could work for the motivated learner. But so would an abacus or a pen and paper.

>>>You can't compare a teacher that has a class of 15 native english speaking, upper middle class, 2 parent house holds kids to a teacher with a class of 32 english as a second langurage, poor, hungry, cold, kids that are children of drug addicts, drunks, refugees especially using test written in English.

Nor should you, but hopefully no one is seriously suggesting comparing those two groups directly. What you can do is compare teachers to their peers in similar schools, and see whose students improve more over a given year. If an average student in Teacher A's class goes from reading at level x to x+2 in the course of a year, that's better than going from x to x+1.

This is making the assumption that a sample size of 30 kids in one school year doesn't have anomalies that will skew the average and that the testing for reading levels such as the DRAs are foolproof?

To your second point, I wouldn't test only reading levels - I'd test whatever we want/expect students to improve in. Reading, writing, math, social skills, happiness, physical fitness, whatever. And sure, we need good solid methods for testing these things, so let's get started on it.

To your first point...maybe 30 kids isn't a big enough sample. I'm not suggesting this be implemented tomorrow as the absolute standard on which to judge teachers. Maybe they need to be observed over 3-5 years, maybe we can try to control/adjust for other variables. What bothers me is the idea that because measuring teacher skill/success is difficult, we shouldn't try at all and just use seniority (or absolute test scores) as the single deciding factor.

There is a rather large contingent of people in America who suggest exactly that. Broadly, they are generally conservatives who believe that "government" and "evil" are the same thing.

I spent 9 years in some of the worst public schools in America until I was able to escape to a private school.

I think you completely underrate the very real problem of bad teachers in poor cities. This is not just teachers merely being unable to teach kids with all manner of problems - there are many teachers who are actively making things worse, in many cases being both verbally and physically abusive to the students.

The very real problem of teacher evaluation does not in anyway obviate the fact that many teachers in poor communities are more horrible than most would dare imagine.

The only solution to negligent teachers is visibility into these behaviors. This also seems like a way that the conditions could be improved: showing them off.

"poor students have a problem learning regardless of who the teacher is."

That just isn't true. Most school districts don't even get rid of the teachers who are so horrible they shouldn't be allowed near children not to mention the ones that just don't care and kick their feet up and let the kids run wild.

You should check out the documentary "Waiting for Superman". http://www.takepart.com/waiting-for-superman It is on Netflix.

As a former high school teacher in a low-income community, I think that movie only told a small portion of the issues of education. There are certainly teachers that aren't the greatest, but we're doing very little to entice the best teachers to teach in the most difficult places, and even less to make sure it's sustainable. Most of the teachers I knew did what they could at a pace that would allow themselves not to completely burn-out. And a most of the burnouts started as energetic young teachers.

Not dealing with problem employees and allowing them to just continue muddling along sends a strong signal to the best people that their work isn't valued and tends to push them to leave. It's a huge part of the problem; people don't like to talk about firing people because it is socially uncomfortable but in the end it's better for both the workplace and the worker instead of just ignoring it and institutionalizing incompetence.

If you look at the vast majority of MOOCs, they frankly kind of suck. Usually they're just a collection of poorly edited videos with maybe a forum where students can comment. They're on a fixed timeline where assignments need to be done linearly at set intervals, just the same as you would get if you were taking the course in person.

Compare that to something more dynamic like Khan Academy, where the subjects all build on each other like a directed graph. Students learn at their own pace, and gamification techniques get them to come back for more. Subject mastery isn't based on tests, but is constantly checked using heuristics which can determine whether a student needs help in a particular area. Oh, and grades get thrown out the window since everyone achieves complete mastery in the subject.

If you think of replacing teachers with a traditional Udacity style MOOC, yeah, I can't really see it happening. The only benefit they offer is that you don't have to be at the course in person. Could something like Khan Academy replace teachers? Clearly not entirely, however there are so many more benefits to learning this way, that it certainly is in the realm of possibility.

at the same time, MOOCs haven't even really started yet. Think about the first video games compared to what we have now.

Now think about the MOOCs we have now, and imagine what the 8th generation of MOOCs will look like. I can't imagine that. But I can imagine that it will possibly be more effective than the average teacher out there.

One of the greatest things about a MOOC is that your teacher can be a feynman caliber instructor. For a student that is willing to put in a lot of effort and has the desire to learn, what is going to better for you, a year's worth of pre-recorded feynman, or an average physics teacher? that might be up for debate, but personally, i'm going with feynman every single time.

> Feynman caliber instructor

Feynman was brilliant, and I have a set of both his lectures for freshman and the ones he did on gravitation. But he was not a gifted teacher. He was a gifted learner. Those of us who can follow him can see an amazing mind at work. But gifted teachers will look into students minds and help them build new understandings.

The feynman lectures were aimed at the smartest students at caltech to keep them interested in physics. You have to consider who the audience was. If Feynman made a series of lectures for physics 101 where the majority of students had no physics experience (or a physics 101 for the general public), then he would make the material different. You can't look at one set of materials by Feynman for one group of students and conclude that he isn't a gifted teacher. There is a reason he is called the "Great Explainer." There is a reason people called him the Greatest Science Teacher while he was alive.


Don't get me wrong; as I mentioned above I've found his lectures on both introductory physics and gravitation to be extremely enlightening, as did many of the Caltech students he taught (who were, by the way, just normal intro physics students, albeit at a selective university). His teaching can be highly effective when matched to the proper audience.

But that's just my point: there is no 0 to 100 scale of 'goodness' in teaching. And Feyman, regardless of his brilliance, doesn't sit at the top of hierarchy of teaching excellence. It's about fit between the student and the delivery the instructor. The Feynman lectures in video form can not be the pinnacle of physics instruction because they're fixed, not adaptable to the needs of different learners.

How am I getting you wrong? You said above that "he was not a gifted teacher."

You said "his teaching can be highly effective when matched to the proper audience." I disagree. He famously could teach extremely effectively to any level of audience.

You said Feynman "doesn't sit at the top of hierarchy of teaching excellence." Many people would disagree. Many people consider him the greatest physics teacher ever.

In terms of just the Feynman lectures in video form, Yes, I agree that they are limited. But I brought up Feynman because I believe a feynman caliber instructor alive today with the existing possibilities in software and video would definitely be able to make something very adaptable to many learners and easily more effective than the average teacher.

Teachers weren't replaced by books either. Who on earth thinks that they will be replaced by software?

I learnt from books and was guided by my teachers. I expect my kids will learn from books, videos and software and still be guided by teachers.

And as someone who has spent a fruitless day googling for how to do a relatively simple task I can confidently state that mentoring by an accomplished human beats software any day of the week.

Teachers weren't replaced by books either. Who on earth thinks that they will be replaced by software?

I don't see how that follows. Horse-drawn vehicles weren't replaced by trains, but they still were eventually replaced. I don't know if teachers will be replaced by software, but the fact that they weren't replaced by previous technologies is hardly strong evidence against it.

And as someone who has spent a fruitless day googling for how to do a relatively simple task I can confidently state that mentoring by an accomplished human beats software any day of the week.

Well, let's look at the evidence. Since that event, have you stopped using Google and hired a teacher to follow you around to you can ask her/him stuff?

Assuming you haven't, then clearly Google is a better solution, whether it is due to cost, breadth of information, availability, etc.

Now, of course, learning a syllabus is hardly the same as searching for information on random tasks. But then again, so would any software designed to "replace teachers"; clearly no one will suggest using Google for the task.

Personally, I don't think teachers as a whole will be replaced, but I can see it allowing teachers to take on bigger classes, allowing administrators to cut costs by firing a few.

>>> Personally, I don't think teachers as a whole will be replaced, but I can see it allowing teachers to take on bigger classes, allowing administrators to cut costs by firing a few.

So we agree - teachers won't get replaced :-)

but books allowed teachers to take on bigger classes too - and what happened is more people went to classes and we needed more teachers. Maybe we shall see something similar just less classrooms?

This. Software as augmentation or preparation for human interaction, just as books have been (and are still).

Teachers are being replaced by software.

And the reason we need to fire the bad ones is that, unlike software, you can't just replicate their product wherever you need and let the bad ones wilt away through neglect.

I've worked with various online homework systems in mathematics for five years. They have not improved demonstrably in that time.

I've taken a couple MOOCs, and have not been terribly impressed with the mechanical grading algorithms.

Teachers may be replaced by software, but I don't believe the substitution is of equal for equal.

I was thinking of the internet, rather than a specific scholastic tools.

I believe open teaching creates transparency that will enable us to better identify the "bad" ones and allow them to improve their craft through feedback loops. This openness will generate more and better teachers. Those that simply lecture to masses are likely to be replaced because they don't provide true value to students. (disclosure: I'm co-founder of trinket.io.)

I was a bit confused by what you meant "open" until I read more of your company's blog

|| What is Open Teaching?

|| Open teaching means teaching using open materials and providing open access to those materials on the internet.

I gotta say I agree, I would love to see a future where the teachers are curating the content rather than the assigned curriculum. But those bad teachers, they've really come to rely on their curriculums.

P.S. looking forward to see what you guys come up with!

It's not exactly clear what you mean, or you need to be more specific: a curriculum is curated content.

Good curriculums are covering over a lot of poor teaching.

Writing a good curriculum is hard and it's exceptionally difficult to do that as a lone teacher for a novel course. The results will be variable, and this is where you will the very most awful courses.

I'd say in a situation where a good curriculum exists, teachers need to take the role of coach and motivator.

I couldn't agree more about the difficulty in creating a good curriculum. The big question in our minds becomes whether or not a decentralized and distributed system can create great curricula and reduce the pressure on that lone teacher. I believe the answer is yes and with such a system we end up with better courses.

It was more a comment on teachers and their currently established ways.

The big aspect you can't replace with software (yet) is the element of spotting the student's weakness and being a guide. It's a pretty general statement, and would require a tome to defend, but I think the spirit of the article is close.

Knewton.com claims it spots student weaknesses and adapts to it. And since it can offer full attention for each student it can offer better performance than teachers.

What do you mean by being a guide ?

That's nothing. Pessimizer claims he can step over a skyscraper with a single bound, and that he beat Mike Tyson in a Philippine streetfight. For what it's worth.

Knewton.com has claimed a lot of things and should have stayed in test prep. What they are now is a glorified ed publisher (basically, Pearson's technical arm) with crappy management but a lot of VC cash.

You're talking about a very hard problem. Something like Knewton will never replace human teachers.

And why is replacing human teachers desirable, anyway? If anyone succeeded at it (and no one will) it would probably be bad, on the whole, for society. If no one can make money teaching, the value of knowledge goes down. That would just lead to ignorance and bolster the already obnoxious anti-intellectualism of American society.

Why is it so hard ?

At least in math, spotting areas of difficulty for students based on data on previous exercises and data from other students does seem possible.

Based on my experience working on a couple different online homework systems, it's actually a hard problem to determine what kind of wrong answers belong to what kind of conceptual difficulty, even in mathematics.

These systems rely on humans to curate the conceptual difficulties associated with each problem separately, but those lists are never complete and often just confuse the student further.

Some systems don't even try to do what you're suggesting, and just give students the "right" way to solve the problem at hand after a certain number of attempts. These are by far the worst, as it's painfully easy to game the system.

Even in math, this isn't really the case. THere are numerous reasons a student might arrive at an incorrect answer. I've seen systems that provide specific feedback on certain anticipated errors, but tailoring the remediation to individual students is just as hard as teaching the material successfully in the first place. It's often not just a case of "oh, I did this one step wrong", but a deeper issue of misunderstanding the abstraction. We haven't mastered bridging that gap with real-life teachers and tutors, let alone automated software.

It's probably possible. With the leadership team they have, they won't be up to it.

IMO, this whole discussion is flawed as it assumes that the answer is teachers OR software. This is not a zero sum game.

The answer, like the answer to most things, is likely in the middle. Software including but not limited to lecture content from subject matter experts will enhance and modify the way the classroom interacts allowing local "teachers" to focus on their students.

Everything you can learn in an undergraduate cs course (and so much more) has already been written in a hundred different books all available on amazon for a considerably lower price and shipped right to your door. If information was all that teachers (and educational institutions) offered, they would have been replaced a long time ago.

And I'm saying this as someone who prefers to learn from books. So what else is there? Individualized attention? Community and networking? Human interaction? Motivation and structure? Signaling? Probably some combination.

Any real replacement needs to address all of these; video lectures, digital notes, and online questions are just reinventing what I can already pick up at the book store.

I'd say motivation is a big one. It takes motivation to push yourself to learn from books. I prefer to learn that way myself. But plenty of students don't have it or don't see the value. A teacher or mentor can make a huge difference to students in that situation.

Apropos of nothing, one internet guy's unjustified personal opinions on the future of teaching and software: The one-to-many "lecturing" interaction should be replaced by software, in order to give teachers more free time for the one-on-one interaction with students. One-on-one personalized feedback on progress is extremely valuable to students, and more difficult to automate, so it seems to me the best use of teachers' time.

OP here, and I think I agree with what you've written. Lecturing and textbooks are both means of conveying information. Students can seek out information on their own time, but the act of teaching is helping someone else construct an understanding. That's the key thing software can't do.

No. Teachers will be replaced with mentors, who will help guide the students learning. Or teachers will become these mentors.

However, the One to Many lectures WILL be replaced by Software.

Just Watch.

Agreed that lectures need to die.

I think the thing that is most often missing from educational software (and most online course delivery formats) is the most important thing: motivation.

The teachers that I most fondly remember weren't the ones that taught me the most material, or ensured that I got the highest test scores. They were the teachers who inspired me; the ones who unlocked a deep curiosity and desire for learning inside me.

The current system doesn't care that much about motivation (although it is changing). You take classes largely because you have to. You learn so that you can pass a test.

Openness in teaching is a big piece of the puzzle, but it isn't the end of the road. In software, for example, I would argue that by becoming more open, motivation to improve was increased among developers (I don't want to be the one with the terrible code on Github). Of course, in software, the target of openness and the one who needs to be motivated are the same person; you. This isn't the same in education. Increasing openness among teachers will undoubtedly have an effect, but the customer for education isn't the teacher.

We are essentially circling the problem by saying "students aren't learning, let's change things around with the teachers." I think that teachers are very important, and I definitely don't advocate mass firings. I just really hope that we find a way to rephrase the goal to something more along the lines of "students aren't learning, how can we work with the student?" If this is teaching basic stuff largely through online methods and having teachers spend more time with students, great. If this is increasing transparency surrounding schools, great. Personally, I hope that some combination of better software that automates basic stuff and allows teachers to work one on one with students more often, more information, and more training will do it.

Absolutely- I agree wholeheartedly. Motivation to learn is the raw material that a good teacher can induce forma student and use to prompt learning. Software can be a part of increasing how often this process happens.

Teachers won't be replaced by software in environments where human discourse is necessary or valuable.

But research professors who either can't lecture or don't try to lecture well? It's almost inevitable. Lectures are a commodity that can be recorded and broadcast at a massive scale. Individual or small-group instruction isn't a commodity.

>Teachers won't be replaced by software in environments where human discourse is necessary or valuable.

that's true ... until discourse with AI becomes more valuable and efficient investment of time.

In the meantime, discourse with a good teacher in China for example may be more valuable then one with a bad teacher here.

If discourse with AIs is more valuable than discourse with human beings, humanity is on its way out anyway. Why would anyone teach a human being anything in that case? It would be AIs teaching each other.

Relevant article: https://news.ycombinator.com/item?id=7030778

I think the author is spot on. I think the idea of firing all teachers to get the results you want undermines the profession as a whole and I really appreciate his analogy to firing the "bottom 10%" of programmers. There is just so much more a teacher has to do to reach "success", especially when dealing with low-income communities, students with disabilities, or even different levels of aptitude/learning styles. I don't see teachers replacing software with the current solutions.

Credentials: I'm a former TFA teacher turned software engineer and I tutor web development in 1-to-1 sessions on nights/weekends.

Author here. Thanks for the kind words. Udacity's pivot is a shot across the bow for Startups that can't figure out the difference between teaching and media. Their coaches are a good start- step in the right direction. If they had perfected that with academics they might not have had to pivot.

It doesn't matter whether robots or humans are teaching your kids, if the whole approach to education is wrong:

Current model of education: Multidisciplinary, value-free, with an emphasis on standardized tests.

Compare this to:

Alfred North Whitehead's reform model: Transdisciplinary, values-laden, with an emphasis on the importance of imagination. See http://en.wikipedia.org/wiki/Alfred_North_Whitehead#Views_on...

Note that multidisciplinary means teaching small parts of a large number of disconnected subjects, whereas transdisciplinarity means teaching a few important ideas that organically link many different subjects and that apply to real life.

I'm the founder of an edTech where we are using software to _IMPROVE_ teacher training. Teachers, especially new teachers, need a lot of support, not just from in school peers, but outside of school. They also need constructive feedback on their practice (from peer teachers, parents, and students). Current training is one size fits all and not personalized. Current evaluation only focuses on test scores.

Yes they will.

Same incorrect argument people always pull out about with automation. Prove you need some => then you need all.

Software will make teachers 10%+ better then they'll fire the bottom 10% (or just 10%)

As it is class size mattering is a myth. A good teacher can teacher quite a large class. Studies don't show a small class makes a lot of difference.

[Citation needed.]

The teaching of a good teacher can't be replaced by software.

Unfortunately, a whole series of public policy decisions have forced teachers of multiple levels to teach badly.

And there's no reason not to replace someone lecturing to a 200 person class or someone merely prepping for standardized tests with software.

Teachers won't be replaced entirely by software, they will be augmented by it.

However their influence and once revered role in society as educators, proliferators of knowledge, will be dethroned by a growing increase of digital information and autodidacticism.

I don't know where you work, but at every company I've worked at they do fire bad coders (as well as good coders who are hurting the team for some other reason).

Every job will be replaced with software (and hardware for jobs with a physical component).

The only question for each profession is "when?".

Not every job... somebody has to make the money from automation otherwise what's the point?

Until they are.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact