Algorithmic Education vs Indie EdTech
February 12, 2016 1:33 PM Subscribe
An algorithmic education, despite all the promises made by ed-tech entrepreneurs for “revolution” and “disruption,” is likely to re-inscribe the power relations that are already in place in school and in society.
Whether it's annotating the scholarly web, creating connected copies through wikity or domain of one's own, alternatives to algorithmic education are offered by the indie edtech movement.
I mean, it's not even universally applicable to things that *don't* involve human nature.
posted by smidgen at 1:52 PM on February 12, 2016
posted by smidgen at 1:52 PM on February 12, 2016
Hi. I work in this field. AMA.
I can say an awful lot about this, but I would add one huge caveat:
Just like self-driving cars, "algorithmic education" is not going to happen overnight, and is going to require a lot of human input at first.
I think that it's reasonable to say that OER (Open Educational Resources) and free/decentralized content are going to become more prevalent over the next few years.
However, the curation of that content is a big question that remains to be answered. To work effectively, any sort of algorithmic/adaptive learning system needs to be "trained" on a large body of content that has been carefully screened for quality, and has been accurately categorized, tagged, and mapped to curricula and competencies. That's going to require a huge amount of (highly-skilled) human effort.
While the production and availability of content becomes more decentralized (a la the web), we may end up seeing the collection and distribution of that content become incredibly centralized (ie. like how Google and YouTube effectively became gatekeepers and monopolies).
Unless we develop Strong AI in the next few years, we're still going to need librarians, instructors, and curriculum designers to keep doing the jobs that they are currently doing.
The tools and standards for content creation/distribution are also sadly lacking (we're working on that!). There are a few factors at play that will make this difficult to overcome -- "raw HTML" and the Open Web are dying a tortured death; we don't have good tools for sharing/embedding foreign content; nobody has the budget or willpower to rewrite existing content into [some new standard that might not take off].
Also, there is a tendency to overcomplicate educational metadata and standards -- there are so many educational paradigms that any schema that can represent them all ends up being overwhelmingly complex and ambiguous (ie. impossible to parse in any meaningful way or share across disparate environments -- LOM is one example, and horrifyingly, I most often see it embedded inside of other, even larger schemas).
I sit on a few standards committees, and it's incredible to see the number of totally legitimate use-cases that the other representatives have.
Disclaimer: These opinions are very much my own.
posted by schmod at 2:32 PM on February 12, 2016 [3 favorites]
I can say an awful lot about this, but I would add one huge caveat:
Just like self-driving cars, "algorithmic education" is not going to happen overnight, and is going to require a lot of human input at first.
I think that it's reasonable to say that OER (Open Educational Resources) and free/decentralized content are going to become more prevalent over the next few years.
However, the curation of that content is a big question that remains to be answered. To work effectively, any sort of algorithmic/adaptive learning system needs to be "trained" on a large body of content that has been carefully screened for quality, and has been accurately categorized, tagged, and mapped to curricula and competencies. That's going to require a huge amount of (highly-skilled) human effort.
While the production and availability of content becomes more decentralized (a la the web), we may end up seeing the collection and distribution of that content become incredibly centralized (ie. like how Google and YouTube effectively became gatekeepers and monopolies).
Unless we develop Strong AI in the next few years, we're still going to need librarians, instructors, and curriculum designers to keep doing the jobs that they are currently doing.
The tools and standards for content creation/distribution are also sadly lacking (we're working on that!). There are a few factors at play that will make this difficult to overcome -- "raw HTML" and the Open Web are dying a tortured death; we don't have good tools for sharing/embedding foreign content; nobody has the budget or willpower to rewrite existing content into [some new standard that might not take off].
Also, there is a tendency to overcomplicate educational metadata and standards -- there are so many educational paradigms that any schema that can represent them all ends up being overwhelmingly complex and ambiguous (ie. impossible to parse in any meaningful way or share across disparate environments -- LOM is one example, and horrifyingly, I most often see it embedded inside of other, even larger schemas).
I sit on a few standards committees, and it's incredible to see the number of totally legitimate use-cases that the other representatives have.
Disclaimer: These opinions are very much my own.
posted by schmod at 2:32 PM on February 12, 2016 [3 favorites]
Yeah, this is one of the those things that make me shake my head at people who believe machine learning is magic that is universally applicable to everything, including nebulous societal problems.
If Moses had waited until 2016 to make his fateful hike up Mt. Sinai, he'd come down carrying The Ten Algorithms.
posted by jamjam at 3:23 PM on February 12, 2016
If Moses had waited until 2016 to make his fateful hike up Mt. Sinai, he'd come down carrying The Ten Algorithms.
posted by jamjam at 3:23 PM on February 12, 2016
No one should dispute the effectiveness of a well-chosen statistical model to... model the input features given some fitness function. Nor that gathering data is hard. The issue is what the features (tags) and loss function are in the first place.
Feel free to correct me here, but unlike self-driving cars, which has very clear criteria, I feel like who you chose to do the tagging and who you chose to describe what fitness actually *is* in education is going to have inordinate impact on people who are generally ignored by the system now because they will end up living at the margins of the statistics you're going to come up with. That's kind of the point of the article.
posted by smidgen at 3:29 PM on February 12, 2016
Feel free to correct me here, but unlike self-driving cars, which has very clear criteria, I feel like who you chose to do the tagging and who you chose to describe what fitness actually *is* in education is going to have inordinate impact on people who are generally ignored by the system now because they will end up living at the margins of the statistics you're going to come up with. That's kind of the point of the article.
posted by smidgen at 3:29 PM on February 12, 2016
Huh. My university uses what I assume is algorithmic education to teach remedial math. It's not for the students in the very lowest remedial math class, but for the ones who need a review of algebra before they can move on to pre-calc. As I understand it, the way it works is that it does a pre-test to figure out which concepts each student doesn't understand. Then it starts on the first concept that the test suggested the student should study. The student does an interactive module that explains the concept, and then they do some problems. If they get the problems right, they're done and they move on. If they don't get them right, they go back and do another interactive module, and then they try more problems. Then they do it again, until they're getting the problems right, at which point they move on. There's a teacher available to explain things if the student needs it.
The fascinating thing is that some students really, really love this and feel like they learn much better this way, and some students really, really hate it and claim that they've spent two months on the first topic and are never going to be able to move on. I would love to look into which students do well with it and which don't, but I don't think we have that data.
In general, I have mixed feelings about it. On the one hand, I hate the fact that we are outsourcing pedagogy to some educational technology company that is selling us a product. I also just instinctively hate the idea that students are interacting with a machine rather than a person. On the other hand, I think that universities are not very good at teaching remedial algebra. I don't think that teaching remedial algebra is really in most mathematicians' skill set, and mathematicians are the people who do the teaching at universities. People act like the alternative to the program is an excellent, engaged professor, but it's often some bored, irritated grad student who has no idea how to teach this stuff and isn't doing a very good job hiding their contempt for anyone who needs to be taught it.
posted by ArbitraryAndCapricious at 4:38 PM on February 12, 2016 [3 favorites]
The fascinating thing is that some students really, really love this and feel like they learn much better this way, and some students really, really hate it and claim that they've spent two months on the first topic and are never going to be able to move on. I would love to look into which students do well with it and which don't, but I don't think we have that data.
In general, I have mixed feelings about it. On the one hand, I hate the fact that we are outsourcing pedagogy to some educational technology company that is selling us a product. I also just instinctively hate the idea that students are interacting with a machine rather than a person. On the other hand, I think that universities are not very good at teaching remedial algebra. I don't think that teaching remedial algebra is really in most mathematicians' skill set, and mathematicians are the people who do the teaching at universities. People act like the alternative to the program is an excellent, engaged professor, but it's often some bored, irritated grad student who has no idea how to teach this stuff and isn't doing a very good job hiding their contempt for anyone who needs to be taught it.
posted by ArbitraryAndCapricious at 4:38 PM on February 12, 2016 [3 favorites]
Indie edtech?!? What happened to edupunk?
posted by humanfont at 5:14 PM on February 12, 2016 [5 favorites]
posted by humanfont at 5:14 PM on February 12, 2016 [5 favorites]
I am... let's say "highly skeptical" about anything designated as "indie" in tech these days. If you're out there doing it and making money - or, let's face it, even not making money - you're pretty much part of the establishment. Sorry.
posted by koeselitz at 5:18 PM on February 12, 2016
posted by koeselitz at 5:18 PM on February 12, 2016
What? How is "indie edtech" any less likely to "re-inscribe the power relations that are already in place in school and in society."
The article lists a bunch of advantages of indie edtech: decentralized control over the platform and content, better protection of privacy, less of a factory model, etc. I don't see how any of those accomplish the stated goal.
It reminds me of early-2000s arguments for open source software as a liberating tool. Which, yeah, it is in some ways, but actually a classroom isn't any more liberated just because its computer runs OpenOffice on Linux. There are good reasons to prefer open source, even from a social justice standpoint. But empowering the median end user is rarely one of them, especially with respect to power relations.
As a more privileged person, open source and otherwise "indie" stuff has been great for me, even emancipatory. It has let me learn more, start projects on a shoestring budget, feel a sense of agency over my stuff. But that sense of emancipation is premised on my position of access.
As long as we have the educational system as it exists—a coercive government bureaucracy in the grips of a sclerotic corporate oligopoly—it is unlikely any technology, indie or otherwise, will emancipate it. Meanwhile, we've had Montessori and Waldorf education for a century, and it continues to be awesome and emancipatory. The problem is that only more privileged people get access to it, while less privileged people get stuck in prison-like public schools.
The "indie" thing seems like false conscience. Technology is a tool. Maybe indie ed tech is a better tool than the "algorithmic" variety. But what matters, I think, is who wields the tool, and what they seek to accomplish.
posted by andrewpcone at 5:43 PM on February 12, 2016
The article lists a bunch of advantages of indie edtech: decentralized control over the platform and content, better protection of privacy, less of a factory model, etc. I don't see how any of those accomplish the stated goal.
It reminds me of early-2000s arguments for open source software as a liberating tool. Which, yeah, it is in some ways, but actually a classroom isn't any more liberated just because its computer runs OpenOffice on Linux. There are good reasons to prefer open source, even from a social justice standpoint. But empowering the median end user is rarely one of them, especially with respect to power relations.
As a more privileged person, open source and otherwise "indie" stuff has been great for me, even emancipatory. It has let me learn more, start projects on a shoestring budget, feel a sense of agency over my stuff. But that sense of emancipation is premised on my position of access.
As long as we have the educational system as it exists—a coercive government bureaucracy in the grips of a sclerotic corporate oligopoly—it is unlikely any technology, indie or otherwise, will emancipate it. Meanwhile, we've had Montessori and Waldorf education for a century, and it continues to be awesome and emancipatory. The problem is that only more privileged people get access to it, while less privileged people get stuck in prison-like public schools.
The "indie" thing seems like false conscience. Technology is a tool. Maybe indie ed tech is a better tool than the "algorithmic" variety. But what matters, I think, is who wields the tool, and what they seek to accomplish.
posted by andrewpcone at 5:43 PM on February 12, 2016
from article: "There are some competing and contradictory arguments made by algorithms and education technologies when it comes to identifying deviance or excellence."
Yes, it's very hard to trust the arguments that algorithms like to make. Sneaky, sneaky algorithms.
posted by koeselitz at 11:35 PM on February 12, 2016
Yes, it's very hard to trust the arguments that algorithms like to make. Sneaky, sneaky algorithms.
posted by koeselitz at 11:35 PM on February 12, 2016
« Older Heliciculture | The opposite of rape culture is nurturance culture Newer »
This thread has been archived and is closed to new comments
Garbage in, garbage out.
posted by smidgen at 1:51 PM on February 12, 2016 [1 favorite]