Hello to everyone. This is my first post on the list and I'd like to jump in here with the question of how much programming understanding is possible at the k12 level:

I taught an AI programming class at my high school last year (and recently spoke about it at ISTE).

I used Jupyter Networks based on Andrew Ng's Deep Learning Coursera course to lead the students subfunction by subfunction toward a neural network (finalized and tested in Spyder) they wrote from scratch with the exception of the backprop calculus. I taught them about back propagation of course, but the math is north of what you learn in an AP Calculus class so I chose not to try to summit that particular mountain and just gave them the code there.

I'll tell you that technically all 15 students were successful however in terms of the students actually understanding what they did in a way that carried over to another assignment, the results were more spotty. Most students were in a "Wow, I have to think harder about this" type of struggle but got there. A few students had an easy time of it, and one in particular zoomed way off ahead. There was also a small group where I just had to concede that they weren't close, and that I would have to build in slower, sturdier scaffolding in for the next go 'round, which will be this upcoming year.

(I don't want to give the impression that I've figured out the way to do this right, I'm going to keep experimenting and looking to see what tools become available that will help. I'm also not yet in a place where I have much I feel comfortable sharing my assignments. As I've said, you can look to the Coursera Jupyter notebook design to see how I started last year, and I'm trying to make changes now so that more students can succeed. If I feel like an assignment really goes well though this upcoming year, I will look to share, and if others are aware of resources along these lines, I'd appreciate being told about them.)

But I am quite confident that there's no fundamental cognitive development threshold we're hitting here. I'm seeing some 17 year olds (and to be clear, they weren't new to coding) who get this stuff without much struggle, and I think could have actually learned it when they were even younger.

I think if these kids were better taught - with the pedagogy to be built in the years to come - the results would be considerably stronger than what I was able to achieve with them. And so yes, I think we should be looking into teaching AI algorithms at the k12 level. It amounts to potentially a several year speed-up of the AI/ML engineer development, which could be a huge benefit and probably makes people skeptical, but the key here is that AI algorithms are classically reserved for upper division CS coursework as much because of their minor importance in the grand scheme of CS as it is about raw difficulty. If we decide they should be given a much more prominent role - because industry now demands it - I think we should expect to be able to fast forward the process and the ultimate threshold is based on cognitive development.

Are 12th graders too mentally immature to build these cathedrals in their head? Some aren't ready.. Some definitely are though.

A couple other pieces of information that are probably best mentioned:

First, my school is one of those that gets attention for on "Best Schools" lists because we get really strong students. Success in my school - which I don't claim to have achieved yet in terms of teaching efficacy, but some of my students definitely succeeded - does not mean we should open the floodgates to all schools. Rather, as we have improved methods that allow more clear cut success at a school like mine, it will make sense to try at more typical schools.

Second, I did have an AI background before I became a teacher so it's still very much an open question how easily we can get CS teachers who can teach this stuff. Some current ones can be trained to do it, be we know that it was already tough to find teachers who could teach more basic programming. I hope we can get to a point where high school CS teachers do understand this stuff and are proactive participants in the learning, but I think another goal should be to equip students to learn "on their own", both with independent learning skills and quality online platforms that let precocious students keep zooming ahead along approved roadmaps without remaining dependent on the teacher in front of them.

Sincerely,

Matt

On Sun, Jul 28, 2019 at 12:57 AM Ken Kahn <xxxxxx@gmail.com> wrote:

Frank Zeng wrote:

... you really need to understand the algorithms in AI. If you are doing a simple AI project with students which is just “cat recognition”, if you don’t understand the algorithm, you will end up with just being able to say “that is what AI can do for you!”.

I would like to see a discussion of how much technical detail and algorithm understanding is possible and appropriate for different grade levels Understanding things like backpropagation and stochastic gradient descent in depth is probably inappropriate with very few exceptions for K-12. A "light" understanding may have some value. Even with no understanding of the algorithms one can learn how providing more and varied training data improves accuracy, what kinds of mistakes a "cat recognition" system makes, etc. One can learn from playing with a "cat recognition" system how similar systems could help diagnose x-rays or help farmers eliminate weeds.

Is the idea that school students should know AI algorithms something like the (controversial) idea that students should know how a computer represents and multiplies numbers? I think those few who want to learn about binary representations and floating point circuitry should have support but I don't think it is for the mainstream K-12 students.

AI professionals can (and should) help teachers understand AI but we also need to be careful. New Math came about because professional mathematicians knew mathematics but not learning sciences and pedagogy.

Best,

-ken kahn

On Sun, 28 Jul 2019 at 03:12, Frank Zeng <xxxxxx@aafie.org> wrote:
I’m an AI professional, so I can share some of my opinion from another perspective here.

I did take a look at the AI courses being offered to educators like ISTE's AI courses. Like many similar courses, their objective is more on “AI popularization”, instead of understanding AI technologically. The important point is that, if you want to feel comfortable to teach AI to students (no matter what grades the students are), you really need to understand the algorithms in AI. If you are doing a simple AI project with students which is just “cat recognition”, if you don’t understand the algorithm, you will end up with just being able to say “that is what AI can do for you!”. And, figuring out how it works is not something you and your students can brainstorm and find the solutions.

To understand AI algorithms, the educators really need to learn AI technology. Learning how to teach AI in classrooms can only happen if you understand how AI works. If we talk about offering AI certificates to educators from universities, the certificates must be from the Computer Science Departments of the universities, not from School of Education. Also, it is hard to believe that one without math background can understand AI algorithms, unless he/she learns math prior to learning AI.

It is said that in other countries, like China, the central governments have mandated that all K-12 schools will open AI courses to students. So far, China has produced more than 15 different sets of AI textbooks for K12 schools. Their goal is to open AI courses in most K12 schools before 2021. We are lag behind and need to act quickly. Our educators should learn AI directly from AI professionals so that they can learn quickly and feel comfortable in teaching AI in classrooms (Learning from online videos is not easy for educators without much background). The best way probably is to ask AI professionals to give educators free lessons. My team is willing to give K12 educators free AI lessons through live online classes using Adobe Connect. I think a 32 sessions of interactive classes would give K12 educators a very clear picture in terms of how AI works (even without math background). With Adobe Connect, we can have all the related educators from a single county in a state join the AI class by Internet from home or office at the same time. We could cover all K12 schools in a short period of time if we also have other AI professional teams join us for doing this.

Best,

Frank Zeng, Director
AAFIE Artificial Intelligence Labs

From: xxxxxx@lists.aaai.org <xxxxxx@lists.aaai.org> On Behalf Of Smith, Rachel
Sent: Saturday, July 27, 2019 1:16 PM
To: xxxxxx@lists.aaai.org
Subject: Re: [AI4K12] Simple considerations

Hi Brenda,

Thanks for sharing your insight. I think your idea about creating a website targeting teens may work really well-- especially if we could offer resources for guidance counselors. The idea of a competition is also neat. Today, there's one in Pittsburgh, I believe. Check out the WAICY website for more info.

It was really interesting for me to learn what's going on in TX. At ISTE this year, I had the chance to meet with affiliates from across the country, and I learned that many states are struggling to implement CS. CT recently adopted CSTA standards and a CS requirement. I've been working with our state's Computer Science Council to develop a plan to implement CS instruction-- from teacher preparation/certification to curriculum. It's a lot of work with a lot of moving parts. Just like Texas, we have amazing K12 educators, and we have amazing districts. Yet each district is approaching the new requirements differently. Some offer CS as electives at the high-school level, others integrate it into core subjects from preK-12. There are many approaches, and a one-size-fits all doesn't work. I often look to Rhode Island, which has done a tremendous job of integrating CS into K12, for inspiration. Right now, there are a few universities offering a CS certification program in CT, and the state is working to create pathways for educators like me, who have a Master's in Literature but are teaching CS. I'm on the board of CECA, our state's ISTE affiliate, and we are always striving to help educators feel more comfortable teaching CS.

I don't know how many of us are familiar with the design-thinking process, but I've recently used it to apply for Google's EI program. I'd like to make it easier for teachers to understand AI so that they feel comfortable enough to teach it. ISTE's AI course is a great resource. Yet even after taking it, trying to interpret the college-level AI courses on YouTube, and reading so many articles/studies, I still feel I need to learn more to truly teach it to my middle-school students. I'm trying my very best, though, and know that while I have some introduction lesson plans crafted, it will take me a while to feel truly comfortable with this subject. But here's the thing-- admitting that I don't know the answer to a question and working with the student to find a answer is pretty cool. Not only does it make me more human, it empowers my students. At middle school, I can be goofy, make mistakes, and admit that I don't know and kids will respond to that.

I want to work on a solution that meets K12 teachers at their level. I have so many ideas, but have been afraid to put them into action because I don't have the background knowledge. But the application process to the EI program taught me that I still can contribute in some meaningful way.

Anyway- I've veered from the topic. Thanks for sharing, and I look forward to learning from and connecting with everyone in the future.

Best,
Rachel Smith

On Sat, Jul 27, 2019 at 12:14 PM Brenda Flinn <xxxxxx@gmail.com> wrote:
Hello all,

I work in support services at a high school and I get to somewhat observe daily activity. I would like to pass on some simple observations as you consider things.

First, Texas tried to add computer science to the curriculum, but recently dropped it. It was a difficult fit which I will offer two reasons in my opinion:

(1) how to use a TI calculator instructions are in the math books, computing instructions are absent, and

(2) few teachers are prepared to teach computer science.

Now, like it or not, Texas heavily influences textbook content for all states. Changing the content is a 20 year political endeavor even digitally. And on top of that, state boards of education must approve books used in the classroom. For example, English teachers can only teach literature from a choice of books on an approved list.

So, I would offer that you make a better way for certifying educators to teach computer science. Computer science college graduates are locked out of teaching in public schools because they lack the teacher education requirement hours mandated by states. Encourage universities to offer computer science and educator courses in one degree much like that of music educators.

Incentivize AI challenges attached to high school mathematics classes as an extension of basic, current lesson plans. School participation in interscholastic competitions is active and a source of pride (like debate or social studies). You have to work with those judging graded interscholastic competitions to get one going in AI that matches what is on the plate in the classroom.

Finally, an easy to navigate, responsive website is needed with the excitement of AI news, clear pathways on how to become an educator of computer science in each state ... all geared toward an age 13 to 18 audience plus their teachers and parents.

Okay, I hope you can use these observations as you grow. I can offer you latex development for printing, website creation and maintenance, and content writing.

Please let me know if I can be of assistance.

Sincerely,
Brenda Flinn
To unsubscribe from this list please go to http://archives.simplelists.com

--
"The best teachers are the ones who show you where to look but don't show you what to see." - A. K. Trenfor

This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. Please notify the sender immediately if you have received this email by mistake and delete this email from your system.
Note that any views or opinions presented in this email are solely those of the author and do not necessarily represent those of the Newtown Public Schools.
Finally, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.

</pre>This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. Please notify the sender immediately if you have received this email by mistake and delete this email from your system.
Note that any views or opinions presented in this email are solely those of the author and do not necessarily represent those of the Newtown Public Schools.
Finally, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.

To unsubscribe from this list please go to http://archives.simplelists.com
To unsubscribe from this list please go to http://archives.simplelists.com

To unsubscribe from this list please go to http://www.simplelists.com/confirm.php?u=7iFgzRrSxotHsd8X0NtA75NWAiTrRREJ