Next Steps for Learnstream.page

# Defining the problem

As I said in the last long, rambling email, I think we need to change some of the ways we do things. Now I'm thinking we need to take another step back and look again at what problem we're trying to solve. 

Ok, how do we do _that_? We shouldn't just craft some problem that we happen to solve already. But then there are lots of problems in the world, and we have to narrow it down to something we can offer some insight on. I don't really know how to do this well, but here's a random one: 

PROBLEM! I find it very hard to learn X. X is not taught (well) at school. There are some scattered articles, videos, forum posts, etc. about X, but they are not assembled in any way I want to deal with them. The best way to do stuff with X seems to be forging ahead, googling particular problems that I encounter as they come and hoping there are answers for me.  (Substitute X = programming language, math topic, how to be good at a video game, cooking, etc.)

Also we should keep in mind the one we are (probably) going to be paid to solve. It's not the wide-ranging abstract stuff I like to think about, but it's certainly worthy:

PROBLEM! I'm not learning physics well in high school, but acquiring a stronger background would help me succeed in college. 

Or, we can look at what other people are doing, get inspired, and remix or expand. There's a new Sal Khan TED video that is really great. It covers both the video and exercise aims of Khan Academy: http://www.ted.com/talks/salman_khan_let_s_use_video_to_reinvent_education.html

Here are some especially cool things:

- Flipped classroom. This idea seems really promising to me, and apparently it's been a thing for a few years. The flipped classroom is letting students follow video lectures at their own pace at home and then work on problems at school, where they can get instructor & peer interaction See http://www.youtube.com/learning4mastery

PROBLEM! The flipped classroom is great, but I don't even have a classroom. I'd like to have the peer and instructor support online.

- The attached graph. That graph doesn't exactly show what he says it does, but it's an inspiring idea. He says that a lot of people are slow on a certain concept (lower slope -- they complete few exercises in a given amount of time on the site) but eventually they get it and race ahead. With a normal one-size-fits-all lecture, however, they'd just be stuck struggling to keep up. Just relating to the idea of being left behind rather than too stupid to get it is really empowering (see: STEMs, analysis 2)

PROBLEM! As Khan revealed, I've been left behind in some subject, even though I really have a much greater potential. But I don't want to rely on Khan academy to make sweet modules and exercises for whatever subject I'm trying to learn -- what's a quicker way? Also it seems like something is miSRSing...? 

As you can see I'm biased towards this "educationifying a new domain" thing. There's definitely the advantage of not having much in the way of solutions (as far as I know). Luckily I do have some potential ones! But before they are revealed, it's worth spending more time to see if solid problems for us to solve. LSers, give me problems!!

# Possibilities

I'm sure you all read my last email with the problems. Here's the follow up with solutions that you've been eagerly awaiting, and it's a long one.  


Problem: I'm not learning physics well in high school, but acquiring a stronger background would help me succeed in college.

Creating a really good physics learning application would be a worthy goal of eight weeks. Also, half-assing the course content, even on a beautiful platform, won't really give us good user results that we can use to take things to the next step.

Two things about content: The Eric Mazur peer instruction ideas seem quite good and there is exercise content we can take. See http://www.youtube.com/watch?v=lBYrKPoVFwg Also Saeta mentioned that the big thing that students need to learn better is being able to apply mathematical techniques to physical problems. Kinda in contrast to the peer instruction stuff, but we can try to think of good ways to do that in addition.

We can still take this to a new platform though. I was trying to imagine something that would work well on a mobile device. Basically, it's question-centered but you can pull up video or scroll though a document and play the connected video clip. Not too different but only one mode at a time, which is kinda what we want anyway. I imagine that we'd provide all of the questions and answers, but perhaps we can have a way to take notes and pull up comments too.

Also I was thinking of making it more game-like, e.g. have some kind of story to bring things into context. One I thought of is where you are trying to save the Fukushima nuclear plant. You have to learn stuff on the spot but you are equipped with your Learnstream Device, which you can use at any time to check on lectures and stuff. Various problems arise and you have to answer a question about what to do, but you can use the Device to pull up lectures, notes, etc. Not sure yet whether that is too politically incorrect. Another difficulty is using the SRS but still making it seem reasonable for things to repeat. 

It'd be great to have an illustrator as our 4th person if we do that...


Problem: The flipped classroom is great, but I don't even have a classroom. I'd like to have the peer and instructor support online.

A pleasing and natural way to get questions answered or just discuss a concept is through chat. In other words, instant and free-form communication. As I've discussed a bit with Neal, we can incorporate all of our existing functionality into a single chat interface. There can be multiple modes of sending text in order to save things as questions or answers, text and videos can be linked or embedded, and a bot/AI-instructor can handle spewing questions for review or questions that need answers. Attached is one example of how this might work. There is also precedent for a chatroom where one can do special stuff with each line of text: StackExchange chat! See http://chat.stackoverflow.com/

An offshoot from this is the much simpler idea of just combining chat and video -- just let one of the members of the room set the video and then load it to the same time for everyone. Apparently YouTube actually had this for a while:

"They already had such a feature essentially like a chat room they were called Streams. Where users chatted in "real time" and anyone inside the Stream could add videos. Creators of Streams had customizing options such as custom backgrounds and such. The owner of the Stream was even able to assign administrative options to other users of their choice. Was quite popular but only a certain percentage of users, they claim. I liked them. I never created my own, but I did venture in and out a few. It never fully made it out of Test Tube and eventually they were discontinued." Streams! 


Problem: I find it very hard to learn X. X is not taught (well) at school. There are some scattered articles, videos, forum posts, etc. about X, but they are not assembled in any way I want to deal with them. The best way to do stuff with X seems to be forging ahead, googling particular problems that I encounter as they come and hoping there are answers for me.  (Substitute X = programming language, math topic, how to be good at a video game, cooking, etc.)

First, we shouldn't completely ignore the possibility that there is hope in what we have. But we have to start from square one with the design process. Today I was thinking about some ideas for going from a good real life experience to a cool application, which I can discuss later. But I'd also like to step back even more and observe what good students do and get people to try out different live and online experiences to try to figure out a basic framework for how people are able to use their focus and what works naturally. 

Ignoring Learnstream for now, there are many interesting partial solutions in existence. Blogs, StackExchange, Quora, screencasts, presentation sharing, etc. One simple idea is just to have a hub that links to things in a more orderly manner: "Look there first, then go there if you want to do that, ..." But of course we are more ambitious. 

To give the best answer to this problem requires first understanding what the best way to learn is. The influence of AJATT has been that immersion is a really good method. This is also the Suzuki/mother-tongue/Talent Education method, which is applied to music. It'd be interesting to think more deeply about how this could be translated to something like physics or math and some interesting application ideas would probably emerge. I have some more ideas on learning theory stuff, but I won't make this email any longer/more boring with that (OK I will -- basically, learning = (observation, mimicry, feedback) cycle with various inputs and side processes -- you'll need this in a bit).

Without further ado, a bunch of ideas that may or may not relate to the problem, some just partial inspirations:
- As mentioned, set up chatrooms, possibly with some intelligent agents, and see what emerges.
- Somehow add video and text documents to StackExchange sites and combine it all into some super online classroom thing.
- A LiveMocha-type setup for exercises or other small tasks. (See the community features of LiveMocha here: http://www.youtube.com/watch?v=HOzXI7FYp-8&t=1m11s)
- A game-like environment where students try to stump each other with good questions and/or answer existing questions faster than the other people. 
- Instead of just a simple go-here-read-this model for linking content, do something like Scvngr but with "checkins" at web addresses and challenges that will help you learn. See the how to play video here: http://scvngr.com/
- Doing is learning. Let's look at that in the context of programming. One really good way to learn, I would guess, is to work on open source projects. You get to 1) observe the code that others have done so far 2) mimic this with various modifications 3) get feedback from code reviewers, etc (I'm not super clear on this stage; it may need improvements). Not only do you get to learn but you create something in the process. We can try to exploit some of the ideas and tools used in collaborative software development to do collaborative learning in various domains. Note real analysis minus Learnstream already follows this to some extent: observe Prof. Su do proofs, do slightly different proofs, get feedback from graders. See also http://www.xamuel.com/how-to-learn-difficult-journal-articles/


If you were paying attention, one other problem:

Problem: As Khan revealed, I've been left behind in some subject, even though I really have a much greater potential. But I don't want to rely on Khan academy to make sweet modules and exercises for whatever subject I'm trying to learn -- what's a quicker way? Also it seems like something is miSRSing...? 

Probably something to think about in grad school... 

Neal:

> Some notes and suggestions from my family/sarah (over spring break):
> 
> UI Features:
> 
> 	Be able to change the font size in documents (currently too small!)
> 
> 
> Problems currently with learning stuff online:
> 
> 	Hard to find free, high-quality material
> 	Hard to find material that is organized (ex: LessWrong sequences are an attempt at organization, 
> 	but i still have a really hard time).
> 	Hard to validate the content -- reliability issues
> 	Need someone to ask if you have questions
> 	
> Some general learning problems:
> 
> 	Hard to focus and stick with something
> 	Need visuals or audio -- text by itself is too dry
> 		On related note, people get really excited in classes here at mudd when there is a short video or mathematica demo
> 	Need context
> 	
> 	
> Some solutions (pair with whatever problems above):
> 
> 	Look for material on society pages (ex: American College of Rheumatology; has vetted info)
> 	Pair audio, visual, and text
> 	Lots of simulations, etc.
> 	"Dense citations" -- everything is hyperlinked to primary source, and hyperlinked within platform
> 		Related note: I get really frustrated with news sites that link to "topics" pages instead of primary sources. No excuse
> 		in modern day to *not* have links to primary sources. Would also provide good extensional material.
> 		Problems: Not all primary sources are available for free online. Solution: host server off-shore and steal all content!
> 	Input methods: easy audio recording; people can speak to the computer, then tag audio clips. Could also function as
> 		"self-lecture recall" as well (see Study Hacks?)
> 	Handwritten notes: Snap picture with phone, email in and the notes are ocr'd and tagged to doc based on section headers
> 		(or similar ideas... think evernote)
> 
> 
> Thoughts? Also, more coherent thoughts coming on Ryan's email soonish. FML school is too much work.

Ryan:

> What's a society page?
> 
> W/r/t input methods:
> 
> One of the ideas I forgot to mention -- it may be similar to the LiveMocha model or the game-like thing -- is a very streamlined way to do micro-lectures (e.g. answering a question) by video, text, audio, whatever and then a lot of focus on voting and ranking things. Kinda like Sophia in spirit but not just recycled material from old schoolteachers. We can have one button video recording like Facebook (I think? don't think I've actually tried it)
> 
> Also, there is a lot of potential with the Kno (http://www.kno.com/the-kno) and potentially more tablets like it for more natural input. I can elaborate on this in relation to the real-life-inspired Learnstream redesign.