What Students Do (And Don't Do) in Khan Academy
blog.mrmeyer.com
blog.mrmeyer.com
The fact is KA is a great tool for learning many subjects, and will only continue to improve. The biggest take aways should be:
- spaced repetition of subject matter is extremely useful for long term memorization - bite sized, repeatable lessons allow students to digest more and "rewind" over parts of lectures they may have missed - "gameification" and small rewards are useful for getting students to keep pursuing a subject - students shouldn't move on to more advanced subjects until they've mastered a particular foundational skill
No teaching tool or method is going to be perfect, but I think KA gives parents, educators and most importantly students a really useful tool for self-paced learning that traditional methods can't easily offer.
The reason I say state-funded schools wouldn't like this is because they work on a false premise. Supposedly that education is not free/cheap enough, and that government funding is required to make it work for the masses. You take away that, and public/state schools become glorified daycare so that the parents can stay productive instead of having to babysit their offspring.
One addition: I've just setup a recurring donation to the Khan academy. One small step at proving that society can function without state-coercion.
KhanAcademy is a powerful force for good in education, especially in instances where a parent wants to help his or her child with homework, but does not possess the educational foundation necessary to do so.
Additionally, when I was a middle school teacher, my students would always have questions about topics we wouldn't have time to get to in class. KhanAcademy and TED were invaluable resources to point them towards to encourage their curiosity.
One thing that can't be said enough — which we internalize deeply at KA but is often confused outside our office — is that KA intends to be a resource for education, not a complete solution.
It's a resource for students who wouldn't otherwise have access to this content and for those having trouble with whatever education they're currently getting. And it's a resource for teachers who want a tool to help make sure students cover core skills at their own pace — so they can spend their class time getting students to work together, using those skills in all sorts of creative ways, and generally adding the highest value teachers can add.
I mostly find Dan's advice (and the article's comments) constructive. But taking a step back, it makes me proud that an educational resource covering this amount of content is free, open to this sort of critique, and constantly improving itself. See the comment from Justin Helps, one of our content creators, on Dan's post.
I can understand the political incentives not to openly state such an ambitious goal - the educational establishment is massively powerful, and doesn't want an upstart replacing them. But is there a reason why, in principle, KA wouldn't want to be a complete solution to all of education?
It sounds like your argument is that KA doesn't actually claim these things and that it is not reasonable to evaluate it on those grounds. I don't think that finding that Kahn Academy doesn't do what it claims in these regards qualifies as attacking it, unless critique = attack. Further, finding KA lacking in the types of questions it asks students doesn't require offering a better alternative in order for the critique to be valid.
The rest of your comment is probably not that controversial, but these aspects of KA are not what Meyer was evaluating. Special pleading doesn't give Kahn Academy credit for the claims they make. I'm sure they will continue to improve--as the 800lb gorilla in the room, they have plenty of attention from people who can offer suggestions. The Common Core requires students to think more deeply about the mathematics than most of the state standards that came before it, and the types of questions that KA asks its students are too superficial.
The author writes: "If one of Khan Academy’s goals is to prepare students for success in Common Core mathematics, they’re emphasizing the wrong set of skills."
I would like nothing better than the academy to publicly state this is NOT a goal of theirs. Being good at math and passing math tests are orthogonal* concepts. In my experience teaching to the test is the fastest way to kill the love of any subject because, at the end of the day, the only answer you can give to the question "Why is this important? or "Why should we do it this way?" is "Because the test says so." It's death to independent thought.
I don't know who his advisor is but please kill this dissertation before it gets out into the world and he uses it for self promotion on Good Morning America.
*intentional hyperbole
We have no interest in preparing students for tests just for the sake of it, but if you look at the Common Core standards you'll see that they are well thought out (really!) and designing a curriculum around them is a very sensible thing to do. Of course, we won't restrict ourselves to the material covered by the Common Core if/when we feel there are gaps.
We're interested in helping students learn to solve problems, not in teaching them how to take tests.
That means until that ideal is reached KA could say "Mastery at X depth of knowledge but note, for this standard Y is required" or allow teachers to write free response questions and have them tag the questions by standard, type, depth of knowledge etc. and grade them by hand in the KA system. Including hand graded questions (by a student's teacher)with an adequate tagging system KA mastery statistics might be significantly more accurate.
Now, if you believe the curriculum is poorly designed or the test is poorly correlated with it, I encourage you to make that argument. But explaining why "Know relative sizes of measurement units within one system of units including km, m, cm..." should not be part of the curriculum is a lot harder than making emotional pleas about "love of the subject".
(I've skimmed parts of the common core on a few occasions. It looks pretty good to me. http://www.corestandards.org/Math/Content/4/MD/ )
However, look at it as an optimization problem. CC testing standards and format may be ideal testing tools for 49 minutes 5th period after lunch 5 days a week in a 40 person classroom, performed in perfect lock step by all 40 students, only failure mode is repeating the entire class or developing a hatred of the topic, aka standard public school experience.
However, what does "alignment" WRT KA mean when the educational experience has completely changed? Its analog, you can watch a video 50 times will you get it, jump back and forth, all that. I'd propose the "best" testing technique for that environment probably looks a lot like KA and not so much CC because you need instant feedback and simple questions. Its not quite as good as personal 1 on 1 tutoring, but its better than sitting in class, more importantly the experience is inherently different than sitting in class...
The proposal in the linked article seems to be that KA epic fails, solely in "alignment", because it doesn't test like CC, which is trivial to test and prove and analyze statistically. Factually looks like a good study. Maybe it fails the relevance test because "alignment" is fuzzy to me (maybe it has a special jargon meaning among CC teachers?) I think MAYBE OP and I are concerned that KA aligns pretty well to a decent curriculum (CC) WRT topics and order of topics and presentation standards, and I hope KA doesn't ruin the KA assessment portion to better match CC assessments solely to get a checkbox somewhere.
Maybe we can abstract out KA from the discussion. I'm tolerably good at math and have tutored people. If you define alignment with CC as slavishly following their assessment technique at all times even though I'm 1 on 1 in an ongoing human conversation, such that everything I speak must statistically match the CC assessment percentages in style, tone, and technique, that's ridiculous. If you defined alignment as I'm talking about sets of linear equations with infinite solution set the same general time, order, and way the CC talks about it, thats OK, even if I assess a tutored student the "wrong way" by asking the wrong type of questions (which may very well be the right question for that kid at that moment while also being on average the wrong way to teach all American Kids in a 40 person classroom)
Its seems reasonably non controversial that CC curriculum and assessment are local maxima on society wide averages for teaching math. It seems reasonably non controversial that KAs assessment tools seem to be a local maxima for computerized delivery of help to kids that need extra help. I (and possibly OP) don't think it reasonable to trash talk KA because the local maxima of effectiveness for society as an average whole in a classroom setting doesn't necessarily match KAs local maxima of effectiveness for "kids who need help at home". The danger is screwing up KA enough to make it useless for the kids who need it after school, just to achieve a marketing checkmark.
If KA doesn't align with CC, it's because there is some particular topic that KA forgot about. E.g. maybe common core requires students to "Solve problems involving addition and subtraction of fractions by using information presented in line plots", but Khan forgot about that one.
I also don't think it would be harmful if KA made sure it covered the basic topics of AIEEE or the IIT-JEE either (important Indian standardized exams). A lot of people want to pass those exams and it would be a marginal amount of extra work for KA to help them.
I see two. One is called PARC, one is called Smarter Balanced. And they want to use them to close schools and fire teachers, so look for everyone to teach to whatever sample tests they can get their hands on.
I am afraid your healthy view of CCSS is not widely shared, and the discussion would be much better if CC was offered purely as PD. Teachers love anything that works.
One thing that is a big change from most states' previous standard, is that the Common Core includes 8 Mathematical Practice standards--ways that we want students to think about and do mathematics. These are things like: don't give up when solving problems, know and use the right mathematical tools for the right job, be precise, make and critique arguments, and pay attention to patterns. As with everything we learn, students learn to engage in these behaviors by practicing them, and they get practice by engaging in tasks and interactions that require them to draw conclusions, cite evidence, and apply concepts to solve non-routine problems. In a traditional classroom, these standards of mathematical practices should be woven in with all of the lessons that we teach, and we should assess students on their learning on these standards as well as the content standards.
While the Common Core might facilitate an expansion of standardized testing by giving a common set of expectations to assess, many people erroneously conflate the standards with the assessment of the standards. This is understandable --Meyer does not seem to spend any time on drawing the distinction. The CCSS doesn't mandate or recommend any particular test. And while it makes sense to speak of assessment of the Common Core, it doesn't make sense to speak of the Common Core assessment.
Having said that, any tool that claims to assess the Common Core should also assess these mathematical practices. One such tool is the Smarter Balance Assessment Consortium's test. Meyer claims that KA doesn't prepare students to be successful on the SBAC test, because students spend most of their interactions with Kahn Academy doing exercises and problems that do not require the depth that the Common Core calls for. This assertion seems supported by his data.
You may be right that KA represents the current local maximum for computerized delivery of mathematics, but Meyer is evaluating KA as a curriculum, not as a resource for kids who need extra help. In this role, Meyer's data indicates that KA's interactions with students falls short of depth of knowledge required by other assessments of the Common Core. His hypothesis that KA focuses on these types of questions because they are easier to be scored by a computer seems pretty reasonable to me. I don't think it is trash talking KA to evaluate the product it delivers against the claims they make for that product.
Finally, suppose KA found a great way to have students engage in higher order thinking skills: maybe require students to explain their thinking or make conjectures or solve multi-step problems that require them to apply concepts in non-standard ways. I don't think this would make it useless for kids who use it for extra help, do you?
Common Core is not a set of tests; it's just a curriculum that sets some basic standards for what students ought to learn. It's reasonable to be concerned about standardized testing as a means for assessing whether those standards are being met, but it doesn't sound like you actually have any objection to the Common Core curriculum.
CCSS is untested (unless you count its imposition on 40+ states a test) so we'll have to wait and see if it moves the US up the meaningless PISA test rankings.
What good is being good at math yet being unable to answer any questions about it? Would you get on board an airliner with a pilot who flunked the flight certification tests but assures you that he's really good at flying anyway?
That said I basically agree with you that it's not reasonable to call someone "good" at math if they're unable to do simple calculations (with the caveat that lots of great mathematicians screw up simple calculations on a regular basis -- but they're still at least usually better than the average person). But there are certainly people who memorize enough formulas to pass the test, yet never acquire or don't retain any conceptual understanding. In that sense, passing a test is a necessary but definitely not sufficient condition for being good at math.
The good news is that these catastrophes are dumped even faster than they come into ascendance and CCSS is confirming that trend. We just need to remember this the next time academic geniuses decide we are doing it all wrong and -- instead of offering a little professional development -- decide to offer all new standards, textbooks, and standardized exams.
I too have a soft spot in my heart for bold strokes, I just want them to come from folks who know what they are doing. The CCSS authors did not.
ps. I do not think thesis says what you think it says.
Please don't assume that people who think the CCSS are on balance a good thing are people who are inexperienced in the classroom. There is room to disagree about aspects of the CCSS, but the discussion will be more productive if our assertions are supported by evidence, rather than name-calling.
I am unfamiliar with any large-scale failures of the CCSS with struggling students, but I have seen struggler excel with an approach that emphasizes understanding. Of course, anecdotes are not data, and I'm open to hearing how the Common Core fails our weakest students. I don't think it is perfect, but I do think it is better than any of the alternatives I've taught with.
If we're being honest, in many places in the US, the emphasis on misusing standardized test scores to shut down schools and fire teachers was well under way long before the CCSS. And to be clear, I don't think this makes any sense. But I don't think it's reasonable to pin this on the Common Core, it's just that it provides an enemy that people from around the country can hate on.
Designing a curriculum around a set of standards does not have to be "teaching to the test" and a cursory reading of the CCSS makes it clear that the answer to questions like "Why is this important?" or "Why should we do it this way?" should never be "Because the test says so." Students should "construct viable arguments and critique the reasoning of others." How does this kill independent thought?
The professional educator dialect of English is not very approachable for the layman. To many people it sounds like somebody throwing out $10 words to make the arguments sound more intelligent.
I think Dan's blog is largely pitched at other professional educators. All professions develop their own technical vocabulary.
As an analogy, I recently enjoyed an article about the implementation of the v8 garbage collector by Jay Conrod: http://jayconrod.com/posts/55/a-tour-of-v8-garbage-collectio...
Here's a sample paragraph:
"Distinguishing pointers and data on the heap is the first problem any garbage collector needs to solve. The GC needs to follow pointers in order to discover live objects. Most garbage collection algorithms can migrate objects from one part of memory to another (to reduce fragmentation and increase locality), so we also need to be able to rewrite pointers without disturbing plain old data."
I think the article is an excellent piece of technical writing, but it would also be reasonable to say "the professional compiler engineer dialect of English is not very approachable to the layman."
That said, good communicators need to be able to change their presentation based on the audience, and it's too bad that the people at the meeting you mention couldn't find a common vocabulary.
If you want to see something by the author of this post that is pitched at the layman, he made a very nice TedX presentation a few years ago:
http://www.ted.com/talks/dan_meyer_math_curriculum_makeover?...
We compiler guys do that on purpose in order to keep our salaries high!
On the other hand, only maybe 1% of the population has to suffer thru garbage collector theory in CS classes, and of that 1%, maybe only 0.0001% ever implement a GC "in the real world" and can understand that post. There just aren't that many GC programmers and there are a lot of humans in the general public. In that case jargon is perfectly acceptable, even expected. Also the core of a programmers job is rarely making public presentations.
That's a profound question for me.
The author also writes, "I find it more likely that Khan Academy’s exercise set draws an accurate map of the strengths and weaknesses of education technology in 2014. Khan Academy asks students to solve and calculate so frequently, not because those are the mathematical actions mathematicians and math teachers value most, but because those problems are easy to assign with a computer in 2014." I think this statement is correct. The Khan Academy developers have devoted a lot of thought to how to computerize problems rather than exercises, and they are still working on building new problem types for the mathematics lessons. I'm actually quite impressed with some of the problem types that show up in more advanced mathematics lessons on Khan Academy that go beyond the eighth grade level implied by current United States curriculum standards. But the author's point is well taken that the Common Core, and good mathematics curricula around the world, expect reasoning and explanation from pupils in mathematics lessons already at elementary school age. That currently is indeed difficult to automate.
The author's comments on development of Smarter Balanced Assessment (SBC) as a new platform of online mathematics instruction is of course especially interesting. I will be curious to see how this program continues to develop, and I would be glad to hear from developers of that program in this thread too.
(Disclosure: I am a teacher of voluntary-participation prealgebra mathematics classes to advanced pupils of third-, fourth-, and fifth-grade age, using the Art of Problem Solving prealgebra textbook by Rusczyk, Patrick, and Boppana as a core text with many supplementary materials from around the world. I do not view Khan Academy as a competitor to my classes, but rather as a helpful supplement for many mathematics learners around the world. My daughter, who is of a younger age than eighth-grade age, is currently doing many of the same Khan Academy lessons as those just reviewed by the blog post author. I recommend Khan Academy to my friends, and it will probably stay in the mix for the mathematics studies of my three minor children for quite a few years yet.)
> The Khan Academy developers have devoted a lot of thought to hope to computerize problems rather than exercises, and they are still working on building new problem types for the mathematics lessons. I'm actually quite impressed with some of the problem types that show up in more advanced mathematics lessons on Khan Academy that go beyond the eighth grade level implied by current United States curriculum standards.
Thank you.
> Common Core, and good mathematics curricular around the world, expect reasoning and explanation from pupils in mathematics lessons already at elementary school age. That currently is indeed difficult to automate.
This is definitely true. By no means is Khan Academy a complete solution for mathematics instruction, nor is it likely to become one in the next few years. Our site should ideally be used alongside other materials that test reasoning and explanation in a non-automated way, as well as projects that allow students to apply their knowledge. We have early experiments that take a much more interactive approach to teaching certain concepts and encourage more experimentation, but we're probably years away from developing them into a shipping product.
> The author's comments on development of Smarter Balanced Assessment (SBC) as a new platform of online mathematics instruction is of course especially interesting. I will be curious to see how this program continues to develop, and I would be glad to hear from developers of that program in this thread too.
If I'm not mistaken, many of the SBAC problems (especially free-response ones) are hand-graded, which can work if you charge for each test taken, but can't work for us when we have millions of problems done every day, all graded automatically. That said, we're exploring strategies like peer grading in other subject areas (particularly our programming curriculum) and would be excited to bring them to math as well.
> early experiments that take a much more interactive approach
> we're exploring strategies like peer grading
Are you guys working on any genetic / learning algorithm implementations of AI sufficient enough to evaluate responses to arbitrary prompts? It sounds both like something that should be feasible with modern tech, especially when time to response can be large (a student can wait hours for an answer to a complex essay response) and the scale can be arbitrary (as many server clusters as "its worth" yet also feasible (we can see similar scale literary processing in other spaces).
I can imagine such an algorithmic implementation where you take some common query given to students across hundreds of schools, with all the teacher graded responses, and maybe after a few years it can get within an accuracy confidence threshold to start grading the prompt, and after another few years providing critique on that grade. That would be the most revolutionary educational tech since the Internet itself, even if it were pigeonholed in several years studying the same single problem for now. Sounds awesome either way.
You can even use that peer grading tech to build up a library of problem / solutions to build your causal relations table from. I just hope you guys have someone in a back room working on this, because to me it seems as real world usable computational linguistics happen more often, there arise a ton of useful applications even in the minimum case.
However just now it required me to signup/login. Signing up asked for PII, which is intrusive and strictly unnecessary for letting me see educational videos. I gave a fake email, along with the other fields. It then complained it couldn't send an email to it. When I pressed their button to fix the email address, it then popped up a window for me to enter it, however it was auto-forcing it down into an obviously stupid position such that it was impossible to hit its Submit button. You can enter a new address but not hit Submit because too close to the screen bottom's edge.
I love the orig vision, and the core content (Sal's lectures and his style) but looks like they're now falling down at the simple stuff along the edges of the problem space. Failing over inside the unnecessarily added complexity (GUI/JS shenanigans, control fetish, privacy intrusion, etc), rather than in the essential inescapable parts (text, images, embedded canned video, links, curricula design, effective communication, etc.)