Latest / Azadeh and Guy on AI / AI in Computer Science Education from CSTA 2026

Transcript

AI in Computer Science Education from CSTA 2026 · Azadeh and Guy on AI

• 0:00
  SPEAKER_00: Hi, and welcome to...
• 0:03
  SPEAKER_01: Episode 8 of Azadeh...
• 0:06
  SPEAKER_00: And I...
• 0:07
  SPEAKER_01: ...an AR.
• 0:08
  SPEAKER_00: All right, and I am in New Orleans at the Computer Science Teachers Association meeting.
• 0:15
  SPEAKER_01: Yeah, that's amazing.
• 0:17
  SPEAKER_01: Tell us everything, Guy.
• 0:18
  SPEAKER_01: What is happening out there?
• 0:20
  SPEAKER_00: Well, I'm eating a lot of vinyas, but it really is happening.
• 0:26
  SPEAKER_00: Everything else is secondary.
• 0:28
  SPEAKER_01: Yeah, enjoy it.
• 0:31
  SPEAKER_00: Yeah, and I think that this is a really interesting space
• 0:34
  SPEAKER_00: because computer science teachers have been teaching about machine learning and AI
• 0:40
  SPEAKER_00: for a long time, much longer than we've been talking about generative AI.
• 0:44
  SPEAKER_00: Yeah, yeah, that's true.
• 0:47
  SPEAKER_00: And so I've been wondering about what's happening in the space as AI is suddenly
• 0:53
  SPEAKER_00: in everybody's life and suddenly a piece of information
• 0:57
  SPEAKER_00: that was really highly specialized And computer science teachers felt like they were
• 1:03
  SPEAKER_00: in the know is suddenly out there in the wild,
• 1:06
  SPEAKER_00: being interpreted and misinterpreted and used in multiple ways.
• 1:10
  SPEAKER_00: And so that's really interesting to see how this specific teacher community is responding.
• 1:16
  SPEAKER_01: Yeah, yeah.
• 1:20
  SPEAKER_00: And so the first thing is that
• 1:23
  SPEAKER_00: the Computer Science Teachers Association has just released its new standards.
• 1:28
  SPEAKER_00: And these standards include AI in them.
• 1:34
  SPEAKER_00: So they were already writing the standards as AI was emerging.
• 1:39
  SPEAKER_00: And it was just the right time for them to take a little bit longer.
• 1:43
  SPEAKER_00: I think that they delayed the release a little bit to make sure
• 1:47
  SPEAKER_00: that they can include AI more comprehensively into the standards.
• 1:53
  SPEAKER_01: Did you have a chance to take a look at the new standards?
• 1:58
  SPEAKER_00: I just took a peek.
• 2:01
  SPEAKER_00: And so there are the obvious things about ethics and about how does it work.
• 2:08
  SPEAKER_00: And so there are multiple layers,
• 2:11
  SPEAKER_00: obviously from the perspective of computer science,
• 2:14
  SPEAKER_00: which is not,
• 2:16
  SPEAKER_00: you know,
• 2:16
  SPEAKER_00: can I use AI for writing my homework for me,
• 2:20
  SPEAKER_00: which is a totally different question, right?
• 2:24
  SPEAKER_01: Yeah, yeah.
• 2:24
  SPEAKER_01: Mm-hmm.
• 2:26
  SPEAKER_00: And so anybody who's interested,
• 2:28
  SPEAKER_00: I would argue that you want to go and check out these new standards
• 2:34
  SPEAKER_00: and really see how they looked at the work happening.
• 2:42
  SPEAKER_00: And just to remind anybody who's listening and may not know
• 2:46
  SPEAKER_00: about how standards like this work, these are Generally, a recommendation.
• 2:55
  SPEAKER_00: They are not mandatory in any way because this is an organization.
• 2:58
  SPEAKER_00: This is not a state agency that is dictating.
• 3:02
  SPEAKER_01: Mm-hmm.
• 3:03
  SPEAKER_01: Mm-hmm.
• 3:05
  SPEAKER_00: That being said,
• 3:05
  SPEAKER_00: what usually happens after the standards are developed by an organization is
• 3:10
  SPEAKER_00: when states write their own standards,
• 3:13
  SPEAKER_00: they tend to at least take a look at what's available for the national organization.
• 3:19
  SPEAKER_00: So I expect that in the next few years,
• 3:21
  SPEAKER_00: that would really start having an impact on what's happening in the states.
• 3:30
  SPEAKER_01: Mm-hmm.
• 3:30
  SPEAKER_01: Mm-hmm.
• 3:32
  SPEAKER_00: And what I do love is all of this is available online and you can see everything.
• 3:39
  SPEAKER_00: So I'm right now looking at Kindergarten and looking at all of the resources
• 3:47
  SPEAKER_00: that they have and all of the So they kept it to a reasonable number of standards,
• 3:55
  SPEAKER_00: so you're not drowning in information and it's hyper-specific.
• 3:59
  SPEAKER_00: Instead,
• 4:00
  SPEAKER_00: the standards are kept at a more general level,
• 4:04
  SPEAKER_00: and that allows you to focus more specifically about what's happening in...
• 4:13
  SPEAKER_00: In this case, in computer science, but obviously AI is all of that.
• 4:17
  SPEAKER_00: So, for example,
• 4:18
  SPEAKER_00: if I look,
• 4:18
  SPEAKER_00: I'm right now, just to give an example, I'm looking at the middle school standards.
• 4:23
  SPEAKER_00: And the middle school standard number five is use an AI tool to generate outputs
• 4:30
  SPEAKER_00: that assist in solving a computational problem.
• 4:35
  SPEAKER_00: So it really is like, here is how an AI tool can help me in my work.
• 4:40
  SPEAKER_00: But the next one is how a machine learning model generates classification or predictions.
• 4:50
  SPEAKER_00: So one is, how do I use AI?
• 4:51
  SPEAKER_00: And the other one is, how do I learn about AI?
• 4:56
  SPEAKER_00: Yeah.
• 4:59
  SPEAKER_00: Different aspects of all of this are happening, and this is in the middle school grades.
• 5:04
  SPEAKER_00: So they're really pushing the machine learning models in middle school with the idea
• 5:10
  SPEAKER_00: that if kids are going to be using AI,
• 5:13
  SPEAKER_00: and they're probably going to do that throughout their middle school,
• 5:16
  SPEAKER_00: definitely high school career,
• 5:17
  SPEAKER_00: we want them to have a better understanding of how these models are being generated,
• 5:22
  SPEAKER_00: and basic machine learning models are a great way to start.
• 5:28
  SPEAKER_01: Yep, yep.
• 5:29
  SPEAKER_01: And I know that you presented a poster, right?
• 5:33
  SPEAKER_00: Yes.
• 5:34
  SPEAKER_00: We presented a poster that wasn't about AI.
• 5:37
  SPEAKER_00: It was really about the work we're doing in Nebraska,
• 5:41
  SPEAKER_00: about solving the problem of having enough computer science teachers.
• 5:44
  SPEAKER_00: And so we have a project called CS Futures.
• 5:47
  SPEAKER_00: It's led by Lin Tia Tso.
• 5:51
  SPEAKER_00: I'm a co-PI, so is Brian Dorn and a few others.
• 5:55
  SPEAKER_00: Wendy Smith is another co-PI.
• 5:57
  SPEAKER_01: And the goal is
• 5:59
  SPEAKER_00: really to try and solve the problem of having enough teachers
• 6:03
  SPEAKER_00: that can teach computer science because we have a state requirement
• 6:07
  SPEAKER_00: that everybody needs a computer science class to graduate.
• 6:11
  SPEAKER_00: And we have a shortage of teachers that can actually teach.
• 6:14
  SPEAKER_00: Even if they're certified to teach, they may not be able to teach.
• 6:18
  SPEAKER_00: And our goal is to make sure that all of our pre-service teachers that can teach
• 6:24
  SPEAKER_00: about computer science and AI have a preparation in it and thus work across
• 6:31
  SPEAKER_00: the state,
• 6:32
  SPEAKER_00: which is something that a lot of other states are very, very interested in.
• 6:35
  SPEAKER_01: Mm-hmm, mm-hmm, mm-hmm.
• 6:40
  SPEAKER_00: And it's really interesting that, and that's my main observation about AI.
• 6:48
  SPEAKER_00: There are a couple of conversations about AI that are happening in this space.
• 6:53
  SPEAKER_00: And one of them is the general conversation about AI.
• 6:56
  SPEAKER_00: And that is,
• 6:58
  SPEAKER_00: how do we prevent cognitive offloading and reducing friction from learning exactly
• 7:04
  SPEAKER_00: where we want friction to happen?
• 7:07
  SPEAKER_00: And that's a concern we're all having.
• 7:09
  SPEAKER_00: Whether I'm teaching in graduate school or pre-service teachers or in
• 7:14
  SPEAKER_00: a first grade classroom, that is a question on everybody's mind.
• 7:18
  SPEAKER_00: And I think that that is a question that computer science teachers are dealing with
• 7:24
  SPEAKER_00: at a great level, especially if they're teaching how to code, because
• 7:30
  SPEAKER_00: All of the models are very good at writing code, at least at writing passable code.
• 7:36
  SPEAKER_01: Mm-hmm.
• 7:37
  SPEAKER_01: Mm-hmm.
• 7:37
  SPEAKER_01: Yeah.
• 7:38
  SPEAKER_01: Yeah.
• 7:40
  SPEAKER_00: And so that's on everybody's mind.
• 7:44
  SPEAKER_00: And there was a keynote yesterday about computer science,
• 7:49
  SPEAKER_00: but especially AI and the frameworks from MIT and MIT's Media Lab to think
• 7:57
  SPEAKER_00: about how do we want ethical and useful AI that doesn't replace thinking
• 8:02
  SPEAKER_00: and doesn't allow kids to offload the thinking into a large language model.
• 8:10
  SPEAKER_00: And so
• 8:14
  SPEAKER_00: that was really interesting to hear that from their perspective
• 8:18
  SPEAKER_00: with this crowd of computer science teachers.
• 8:21
  SPEAKER_00: There was definitely a conversation about data centers and
• 8:26
  SPEAKER_00: the potential environmental impact, which is a topic That is on everybody's mind.
• 8:32
  SPEAKER_00: I think that very,
• 8:35
  SPEAKER_00: very often,
• 8:36
  SPEAKER_00: and I'm still hearing that, that discussion stays at a very general level.
• 8:42
  SPEAKER_00: And so it's really hard to have that discussion without a deeper exploration
• 8:51
  SPEAKER_00: of what does it mean?
• 8:52
  SPEAKER_00: What does a data center really impact?
• 8:57
  SPEAKER_00: With data, instead of just, we think it's going to be really bad.
• 9:03
  SPEAKER_00: Or we think it's going to take this many liters of water.
• 9:06
  SPEAKER_00: Or gallons, I think, would be what we usually use in
• 9:12
  SPEAKER_01: the United
• 9:13
  SPEAKER_00: States.
• 9:14
  SPEAKER_00: But having a comparison of, you know, how does it compare to other things in our community?
• 9:21
  SPEAKER_00: What percentage of the water use is?
• 9:24
  SPEAKER_00: It is.
• 9:25
  SPEAKER_00: And starting to think about this as if we were really concerned about water,
• 9:30
  SPEAKER_00: we cannot really be concerned just with data centers.
• 9:33
  SPEAKER_00: We need to look throughout our community.
• 9:36
  SPEAKER_00: And that's something that is always on our mind.
• 9:38
  SPEAKER_00: It's like, are we comparing things?
• 9:44
  SPEAKER_00: And we've had this conversation before,
• 9:46
  SPEAKER_00: even on the podcast, and that's an important conversation to have.
• 9:51
  SPEAKER_01: Yep, yep, that's true.
• 9:53
  SPEAKER_01: Did you hear anything about the specific stage at which AI can be incorporated
• 10:00
  SPEAKER_01: into computer science education?
• 10:04
  SPEAKER_01: I mean,
• 10:08
  SPEAKER_01: when you teach kids how to code, at what stage AI should be incorporated?
• 10:14
  SPEAKER_00: So the standards partially answer that, right?
• 10:16
  SPEAKER_00: They want to start that in middle school to help you start thinking
• 10:23
  SPEAKER_00: while using AI very, very carefully.
• 10:27
  SPEAKER_00: But here and there,
• 10:30
  SPEAKER_00: there's a little bit of a discussion, but not as much as I would have hoped for.
• 10:36
  SPEAKER_00: So it's definitely part of the package.
• 10:42
  SPEAKER_00: I think that most of the presenters are still wondering how much and where and
• 10:48
  SPEAKER_00: where should I allow it.
• 10:50
  SPEAKER_00: And so there wasn't anything specific that I succeeded in going to.
• 10:54
  SPEAKER_01: Mm-hmm.
• 10:56
  SPEAKER_00: That really made a cogent argument about how they're doing it and how much of
• 11:01
  SPEAKER_00: that they're doing.
• 11:02
  SPEAKER_00: So I've seen more presentation about machine learning than I've ever seen in a conference.
• 11:10
  SPEAKER_00: And I've been going for a few years now.
• 11:12
  SPEAKER_00: And so for the first time, there were a lot of machine learning models.
• 11:16
  SPEAKER_00: Here you can try this.
• 11:20
  SPEAKER_00: And I think there's more of an emphasis on that because it helps people understand how AI works.
• 11:28
  SPEAKER_00: But from reactions,
• 11:32
  SPEAKER_00: I've started to wonder if that's becoming,
• 11:36
  SPEAKER_00: it's both really,
• 11:37
  SPEAKER_00: really helpful to look at machine learning models and really unhelpful.
• 11:43
  SPEAKER_00: And what I mean by that is
• 11:47
  SPEAKER_00: Machine learning models like the ones from Google and other ones,
• 11:55
  SPEAKER_00: we use them and we feed them just like AI for oceans from code.org.
• 12:01
  SPEAKER_00: We feed them very limited amount of data, right?
• 12:05
  SPEAKER_00: 50 pictures is a lot.
• 12:07
  SPEAKER_00: Usually people feed them 10 pictures to separate dogs and cats and
• 12:11
  SPEAKER_00: then see how accurate they are and usually say something like,
• 12:15
  SPEAKER_00: see, it's really stupid.
• 12:17
  SPEAKER_00: It really can't tell a dog from a cat.
• 12:21
  SPEAKER_00: And I think that there's a danger that we're teaching and learning the wrong lesson
• 12:27
  SPEAKER_00: from that because LLMs are Really interesting because they are so large.
• 12:38
  SPEAKER_00: And it's not just that they have more data,
• 12:41
  SPEAKER_00: it's that when you have this much data,
• 12:45
  SPEAKER_00: the relationships and the complexity,
• 12:48
  SPEAKER_00: you know,
• 12:48
  SPEAKER_00: when you have millions and millions of parameters and more,
• 12:55
  SPEAKER_00: It stops being like a model that separates fruits and vegetables,
• 13:01
  SPEAKER_00: which was an example from this afternoon,
• 13:04
  SPEAKER_00: that you can discuss and why does it look at me and think I'm an apple and all of
• 13:08
  SPEAKER_00: that and think about what happens at scale may be qualitatively different,
• 13:14
  SPEAKER_00: not just quantitatively different.
• 13:16
  SPEAKER_00: It isn't just becoming More accurate.
• 13:19
  SPEAKER_00: It's able to do a whole bunch of things that you can't do with
• 13:23
  SPEAKER_00: a simple machine learning process.
• 13:27
  SPEAKER_00: So scale matters a lot.
• 13:29
  SPEAKER_00: And I think that some of our computer science teachers,
• 13:32
  SPEAKER_00: because they've been familiar with machine learning for so long,
• 13:35
  SPEAKER_00: I still think about these models as just very large machine learning problems,
• 13:42
  SPEAKER_00: which in some ways they are, right?
• 13:44
  SPEAKER_00: That's how we got there.
• 13:46
  SPEAKER_00: But the algorithm itself is so much more complex,
• 13:51
  SPEAKER_00: especially because language data is almost self-tagging, right?
• 13:57
  SPEAKER_00: Because the words themselves, the tokens themselves carry meaning, not just relationships.
• 14:04
  SPEAKER_00: And so you don't need as much tagging as you do in other,
• 14:08
  SPEAKER_00: you know, when you do AI for Oceans, you're tagging.
• 14:13
  SPEAKER_00: And you're saying to the machine,
• 14:14
  SPEAKER_00: this is a fish,
• 14:15
  SPEAKER_00: this is not a fish, this is a fish, this is not a fish, that kind of thing, right?
• 14:19
  SPEAKER_00: And that's how it learns.
• 14:21
  SPEAKER_00: Here you've got algorithms that actually teach themselves to a certain degree,
• 14:25
  SPEAKER_00: and that is a very, very different thing.
• 14:30
  SPEAKER_00: So drawing conclusions from one to the other is a little bit trickier, I think.
• 14:36
  SPEAKER_00: There was a good episode on Hard Fork,
• 14:39
  SPEAKER_00: which is one of the podcasts I listened to, That really talked about this.
• 14:45
  SPEAKER_00: It's like there are people who are sure AI is going to be sentient and thinking in
• 14:50
  SPEAKER_00: the sense that is similar to how humans think, not exactly the same.
• 14:55
  SPEAKER_00: We know that.
• 14:56
  SPEAKER_00: And there are people who are sure it's not.
• 14:59
  SPEAKER_00: And there are people in the middle who are saying, I don't know, really.
• 15:04
  SPEAKER_01: Yeah.
• 15:05
  SPEAKER_00: And I'm not ready to make a judgment.
• 15:08
  SPEAKER_00: I think they are not right now,
• 15:09
  SPEAKER_00: but that doesn't mean that they will never be really thinking in a sense
• 15:14
  SPEAKER_00: that we would recognize as humans.
• 15:17
  SPEAKER_00: And so keeping that question open instead of making
• 15:20
  SPEAKER_00: a sweeping decision is something I think about a lot.
• 15:27
  SPEAKER_00: Anyway, what have you been doing with AI when I'm here at CFDA?
• 15:34
  SPEAKER_01: That's a good question.
• 15:37
  SPEAKER_01: Everyday stuff,
• 15:38
  SPEAKER_01: nothing very specific or different from what I use AI for on a daily basis,
• 15:46
  SPEAKER_01: if that makes sense.
• 15:48
  SPEAKER_00: Yeah.
• 15:49
  SPEAKER_00: So what was your favorite way to use AI this week?
• 15:52
  SPEAKER_01: Oh, that's a very good question.
• 15:55
  SPEAKER_01: Let me see.
• 15:57
  SPEAKER_01: I'm working on a document.
• 15:58
  SPEAKER_01: It's kind of personal, so I cannot talk about the details.
• 16:02
  SPEAKER_01: But I'm doing a lot of back and forth with AI to help me write good paragraphs,
• 16:11
  SPEAKER_01: coherent paragraphs.
• 16:13
  SPEAKER_01: I'm feeding AI with samples of paragraphs, and then I'm feeding AI with my own information.
• 16:19
  SPEAKER_01: I mean, my resume, my professional resume, nothing too personal.
• 16:22
  SPEAKER_00: Yeah, yeah, yeah.
• 16:23
  SPEAKER_01: And then I use AI to write paragraphs about me.
• 16:27
  SPEAKER_01: Then I go back and edit.
• 16:28
  SPEAKER_01: So there's a lot of back and forth, as we always talk about, our interactions with AI.
• 16:35
  SPEAKER_01: Yep.
• 16:36
  SPEAKER_01: So mostly that.
• 16:38
  SPEAKER_00: Yeah.
• 16:39
  SPEAKER_00: My favorite one this week is I fed AI the program for this conference.
• 16:44
  SPEAKER_00: And I said, I really want to.
• 16:48
  SPEAKER_00: Focus on elementary and to focus on AI,
• 16:54
  SPEAKER_00: obviously,
• 16:55
  SPEAKER_00: and to focus a little bit on cyber because I'm trying to understand
• 17:00
  SPEAKER_00: a little bit more what is happening in the cyberspace.
• 17:05
  SPEAKER_00: And it came up with suggestions.
• 17:10
  SPEAKER_00: And I went back and forth a couple of times.
• 17:12
  SPEAKER_00: I said, well, I want a little bit more of this and a little bit less that.
• 17:15
  SPEAKER_00: And I added that I wanted some rural meetings.
• 17:19
  SPEAKER_00: And it created a schedule for me with where to go and when to go.
• 17:23
  SPEAKER_00: And in some cases, it says you can have this or that.
• 17:26
  SPEAKER_00: And it tagged each one of them and color-coded them so I can quickly go through,
• 17:31
  SPEAKER_00: which is really nice because...
• 17:35
  SPEAKER_00: Often, I go to the conferences,
• 17:38
  SPEAKER_00: and there are so many sessions, and it's so overwhelming, and it's unbalanced.
• 17:43
  SPEAKER_00: And what that helps me do is it helps me also have a balance where I'm getting
• 17:48
  SPEAKER_00: a really good cyber session for elementary kids,
• 17:51
  SPEAKER_00: but I'm also getting a session about rule computing and other things.
• 17:57
  SPEAKER_00: So it was actually a really nice program.
• 18:00
  SPEAKER_00: So that was my favorite thing this week.
• 18:02
  SPEAKER_01: Yeah, and it could be very helpful,
• 18:04
  SPEAKER_01: especially if you're going to an ERA with, I don't know, hundreds of sessions.
• 18:09
  SPEAKER_01: Yes.
• 18:10
  SPEAKER_01: Yeah.
• 18:11
  SPEAKER_00: Yeah, that's what I did last ERA.
• 18:12
  SPEAKER_00: It was impossible to get through.
• 18:14
  SPEAKER_00: I mean, there's so many and it's so overwhelming and you eventually just make a random decision.
• 18:21
  SPEAKER_01: Yeah, yeah.
• 18:22
  SPEAKER_00: And so it's nice to... And again, I deviated from the schedule it created for me.
• 18:28
  SPEAKER_00: I don't feel obligated to it.
• 18:30
  SPEAKER_00: But it does provide a really solid starting point.
• 18:34
  SPEAKER_00: And if you want to go and see a specific person who you know will be presenting,
• 18:39
  SPEAKER_00: you can do that.
• 18:40
  SPEAKER_00: Obviously,
• 18:40
  SPEAKER_00: it recognized the fact that I had a poster for two hours and it says,
• 18:44
  SPEAKER_00: this is what anchors your whole program.
• 18:46
  SPEAKER_00: You have to be there and present.
• 18:49
  SPEAKER_00: I'm glad you told me.
• 18:51
  SPEAKER_00: But...
• 18:52
  SPEAKER_00: But it was helpful.
• 18:54
  SPEAKER_00: And so that for me was really invaluable today.
• 18:59
  SPEAKER_00: So that's it.
• 19:01
  SPEAKER_00: All right.
• 19:01
  SPEAKER_00: That's what I learned this week.
• 19:03
  SPEAKER_01: Perfect.
• 19:06
  SPEAKER_00: All right.
• 19:06
  SPEAKER_00: So with that, I think we'll log off.
• 19:11
  SPEAKER_00: Thank you for listening.
• 19:12
  SPEAKER_00: And we'll see you on our next episode of Azadeh and Guy on AI.
• 19:19
  SPEAKER_00: All right.