This was a panel. Pat Yongpradit and BT Twarek co-moderated a panel consisting of Julie York, Maya Israel, Allen Antoine, and Jacob Shulman.

It was a good mix. Julie York is a High School teacher, Maya Israel, a researcher, Allen Antoine is the Director of CS Ed Strategy at UT Austin's Texas Advanced Computing Scenter, and Jacob Shulman is a tech entrepreneur as the founder of Jippity and long time private tutor.

Pat and BT managed the panel well - good questions and didn't make the mistake of having every panelist answer every question.

Some answers missed the mark for me like the question of how to fit AI into a crowded curriculum. There could be multiple full sessions on this one question so it was no surprise that the answer had to be pretty generi - basically view it as a tool and something to use across everything. I would have liked an answer that differentiated learning CS vs using CS and learning CS in a CS class vs a non CS class but time couldn't allow for that. The answer was probably the best that could be given under the constraints but I would have rather a question that was more focused which would have allowed a more directly actionable answer.

On the other hand, I felt the question and answer to how AI can make CS more accessible was more helpful. The idea of using AI based tools that could provide very specific immediate feedback and also how those tools can be used for self learning. While I liked this answer, provided by Maya Israel, it also left me concerned but we'll talk about that in a bit.

Allen Antoine made a good point on how AI still can instill a sense of wonder in kids which can be used to drive interest and broaden participation but I'd caution that the shine eventually wears off. He also expressed concerns about communities getting left behind which is a very real concern. Every technology that has been touted as a great equalizer has never been one. The haves get the best of the tech and the have nots get the cost cutting budget saving tech foisted upon them. AI in the best sense can uplift communities by implementing tools like what Maya mentioned and I wrote on above or they can leave them further behind if those communities get bad AI instructors to replace good teachers.

Jacob Shulman made a couple of interesting points that I didn't agree with but that's okay, as I said, this was a nicely formed panel - that means differences in backgrounds and differences of opinions.

Jacob's product is jippity.pro- it's an online larning IDE that integrates AI. It's basically a vibe coding platform. Looks interesting but currently I'm not yet sold on vibe coding as either a practical or educational tool.

Jacob talked about the value of code reading and how you can learn that way - the implication that you can get some vibe coded solution and figure it out from there. Now, I do agreee that code reading is important and we should do more of it with our classes. You can indeed learn from reading other peoples code but it's not a replacement for all the other ways we learn to code.

I mean, think about it, how often does a kid not get something in math and then they read a solution or better, step through it as instructed by the teacher, TA, or professor and they leave thinking that they get it only to realize on the next exam that they really don't.

Reading or just being dumped into existing code isn't going to do it except for a small subset of kids. I've seen this type of "just dump them into the code" work for tech pros starting at a company even when they're working on a platform and in a language new to them but I've just as often seen it fail unless the new hire got other supports and these are people who while being new hires have extensive CS knowledge.

Actually, if I'm right on this point, that's a potential danger for those self-learning tools.

Jacob also talked about the explosion in students verbal skills and that this is what he's heard from English teachers, presumably due to students using LLMs which are raising their level of discourse.

I've seen the exact opposite. English and other teachers lamenting that the kids aren't learning how to either write or communicate at all but rather they're just parroting what they get from chatGPT.

Now, I could be wrong, Jacob could be wrong or the answer could be somewhere in between. Wish we had time for a discussion but in any event, allowing for these disparate views make for a good panel.

Next up were questions. I wanted to ask one but didn't get a chance. Don't know if I would have been called on regardless but the question answer period ended when after a couple of good questions, one participant took the opportunity to share their individual views on teaching AI rather than asking a question.

So, after that, Pat and BT wrapped up the panel by asking each member to give one piece of advice on a risk with respect to AI.

Julie commented on copyright issues, Allen, ethical concerns and Jacob on the dangers of taking extreme positions on either side - always AI or never AI - all great points. Pat mentioned the concern that kids will only learn AI through prompt engineering and not really about AI - a concern I share as well.

Now, back to my question - I wanted to ask about putting up guardrails to make sure we don't lose student - teacher interaction and student - student interaction. A huge part of teaching is all about connection and relationships with one's classes and a huge part of learning is interacting with one's classmates and one's teachers. If a kid is always going to a chatbot what will happen to their human to human skills. If a teacher just uses AI assignments and autograders how well can they get to know their kids?

I was disappointed not getting to ask about this. Not because I expected a full and useful answer - the question's too big for a quick response but rather, my feeling that very few, if any attendees were thinking about this.

So, I was psyched when both BT and Maya's answers on risk spoke directly to this question even though it was unasked. BT commented on how we're teaching kids to be citizens and that has to be taught even when all the tech is stripped away and Maya hammered in a concern for mental health and an over reliance on chatbots. After that, everyone was thinking about it.

Really great thought provoking panel so thanks to all the panelists and moderators.

Last up was a discussion of the 2024 CSTA teacher landascape survey. The facilitators were BT Twarek so I had the pleasure of attending a BT run session back to back, Lisa Novohatski, Sonia Koshy, and Shaina Glass.

Great presentation and discussion. I won't go into any details on the survey as it will be published soon enough but a couple of points I found interesting. One was that it's really challenging to create a meaningful survey like this. The CSTA wants to get as many CS teachers as possible but they only have contact info for CSTA members. The facilitators also pointed out that even with the efforts they took they ended up withe HUGE number of false surveys (bots I'm presuming). It's also hard getting insightful answers. They can ask how much a teacher might emphasize a topic or issue but what does that mean for an individual teacher? I really don't envy the survey writers at all.

It turns out it's also really hard interpreting the answers. For example, if one is at a school with a single CS class that happens to be a requirement, diversity might not be an issue for you at all, or maybe if you're school isn't diverse it's a huge issue but you can't do anything about it.

Challenges about.

In any event, another interesting and well run session and keep a lookout for the survey once CSTA releases it.

So, that's it for the sessions I attended. Still have to talk about the exhibit hall, receptions, hallway track and Ceveland. That's all up next.