Latest / DiveIn: Diving into Special Education's Most Complex and Pressing Debates / Beyond the Hype: AI, Disability, and Equity in Education

Transcript

Beyond the Hype: AI, Disability, and Equity in Education · DiveIn: Diving into Special Education's Most Complex and Pressing Debates

0:08
From the Division of Research of the Council from Exceptional Children,
0:11
this is Dive In. I am your host,
0:14
Federico Vitaler, professor at the University of Illinois
0:17
at Chicago.
0:22
Welcome, welcome, welcome, welcome to our eighth episode of our second
0:25
season of D Dive in. Today we
0:29
have another timely topic,
0:32
a topic that creates both excitement and
0:35
hope, while also anxiety and fear.
0:39
Today we're going to talk about artificial intelligence, AI,
0:43
and the application of these tools for students with disabilities and education
0:47
at large. We'll take a critical look, not only examine
0:51
its benefits, but also many of its potential risks.
0:55
Can AI remediate historical inequities for students with disabilities
0:59
and other marginalized students or has the potential
1:03
to increase such inequities? We also talk about
1:06
issues of transparency today and standardization, as well as the
1:09
relationship of AI with teachers and students.
1:13
And for all that, I do not have one guest,
1:16
but I have three of them. First,
1:20
I have Richard Allen Carter,
1:22
Jr. He is the holder of the Dr.
1:25
Ted Hasselbring Chair in Special Education Technology
1:29
at Indiana University, Bloomington. Dr.
1:32
Carter researches the education of students with disabilities
1:36
in modern learning environments with a focus on technology enabled
1:40
personalized learning, virtual learning, and blended learning environments.
1:44
He's the co author, with Lin Sang and Yu Ting Yu and
1:48
Peng Peng, of a recent article published in Review of Educational
1:52
Research titled Let's Chat about Artificial Intelligence for
1:55
Students with a Systematic Literature Review and Meta Analysis.
2:00
Also today we have Mary Rice. She's an associate professor at the
2:04
University of New Mexico. She used to be a junior high school teacher and
2:08
taught English language arts and TESOL classes. Her research focus
2:12
is at the intersections of literacies, identities and agencies
2:16
in digitalized settings. She's also the author of
2:19
another literature review recently published in the Journal of Computers in
2:23
Schools. The title of the review is the Use of
2:27
Artificial Intelligence with Students with Identified a
2:31
Systematic Review. I will post all these
2:34
links of these both reviews in the description of the podcast.
2:39
Finally, we have Joaquin Arguello de Jesus. He's a
2:42
Dominicano, requeno, decolonial, anti racist, bilingual,
2:46
community, clinical, and school social worker.
2:49
He was raised on the Manito Ax trail. He's also a
2:53
PhD student in the Department of Language,
2:56
Literacy and Sociocultural Studies at the University of New Mexico.
3:01
Joaquin is currently engaged in various research projects through indigenous
3:05
pedagogies, land and water, traditional practices,
3:08
and online and AI technologies.
3:11
So are you ready? So let's don't waste any more time and dive
3:15
in into the conversation about AI.
3:21
Well, thank you so much Marie, Joaquin, and Richard for being here with
3:24
us. You know we're going to get started. This is a topic that creates
3:28
also a lot of, like, excitement, but a lot of, a lot of concern as
3:31
well. So I'm glad we're supporting this. And I want to start with
3:35
some broad question about, you know, what are the some ways that AI
3:39
has shown promises for students with disabilities?
3:43
I'll jump in. Sorry. So I think what we've
3:46
seen, thinking about AI as a tool, as a
3:50
process has been around for a long time. I think we've seen
3:54
AI be instrumental in some aspects
3:58
of accessibility with students. For instance, if we
4:01
think back to a very broad general generation.
4:05
I'm sorry, definition of AI, we're thinking about speech
4:08
to text, you know, text to speech, those type items which we
4:12
really are integral in how we, how we've served students with
4:16
disabilities for years. And, and thinking about
4:20
how, you know, how, how that good and,
4:23
and how it's ubiquitous in schools really has been the
4:27
excitement and the challenges you spoke of, you know, earlier really has,
4:30
has shown up when it comes to generative AI. So I think that'
4:35
that space is really interesting to play in
4:38
right now. And I'm glad we're going to talk with some great minds today about
4:41
how, how that might, how that might shape out well.
4:45
And a lot of things right now that people are excited about in
4:48
my realm have to do with things involving
4:52
writing. And because people are historically
4:56
we haven't done writing instruction in school, so we assign writing rather
4:59
than teach it. And so then this tool comes along
5:03
or this people call it a tool or sometimes people call it an
5:06
agen and we say,
5:09
oh, this thing can give us some sort of feedback
5:14
on writing that is more than what a teacher would
5:18
do. And everyone is
5:21
saying how happy that is. Like, oh, this thing will tell
5:25
you where to put your headings in or where you need to make more
5:29
descriptive words or something like that.
5:33
And because for the most part people haven't had
5:36
the access to that kind of stuff then and people are
5:40
very excited about that and we
5:44
can problematize that because
5:48
that's actually what I'm apt to do.
5:53
But at least the initial grasp. There's something about
5:58
kids who usually just get red marks all over their paper or
6:02
told that they didn't write enough or who maybe don't get any
6:05
attention at all as writers or there's a
6:09
long standing notion that students who have
6:13
been identified with disabilities can't think of ideas or
6:17
can't take an outline and make it into a draft, that this
6:21
is going to be, you know, some kind of saving thing.
6:26
And it's either it could be that, but it also,
6:30
is it that, or is it that that body annoys
6:33
you or it's like,
6:36
frustrated you in some way and that this is a way to,
6:39
like, ease the tension around
6:44
that body that frustrates you? Yeah. And, and I
6:47
would say thinking about it from a, from a
6:51
student standpoint, myself being a
6:54
dyslexic learner, even an adult learner. Right.
6:58
I, I also like to think about what are healthy ways we can even
7:02
frame what learning differences are. Right. Because the,
7:06
the, the categorization and the viewpoint of somebody with a disability is
7:10
a deficit discourse which automatically sets students in a place of
7:14
them having to prove their worth, to even be with their peers,
7:16
be in a classroom, and then if there's extra tools that are
7:20
being offered to supposedly make them equal to others,
7:23
that that's a different place of learning than if they were seen as
7:27
learning like everybody else and maybe having an alternative way
7:31
of processing cognitively or even physically with any
7:35
differences that they have. And so for me, I think,
7:39
you know, one of the things that I'm hoping this technology can do
7:43
is help get over some of the current limitations. For example,
7:47
I know Professor Rice has asked me to support her in some publications around
7:52
the speech to text programs and how they don't even work for students with accents.
7:56
Right. And so you have a population that's marginalized by
8:00
a school system that's English dominant, that doesn't recognize how that pushes
8:04
kids out, and then further dominate, marginalized by
8:08
the domination of. Well, now this software doesn't even help you
8:11
because it can't recognize your accent. So maybe, and I'm not a tech expert,
8:15
but maybe AI can help address some of that. However,
8:18
I'm still worried that AI will be a tool that is
8:22
used in a way where students feel they have another task to learn on
8:26
top of everything being categorized as disabled. Right.
8:30
So not only they're not seen as normal, they're a population
8:34
within a population that has an extra task to learn, an extra tool
8:37
to supposedly make them feel normal, normal to peer learning, when that's
8:41
not even clearly defined. So I'm thinking about what are the ways we're setting
8:45
students up for success or if we're giving them a longer road with more obstacles
8:49
to feel like they're learning at levels that aren't quite clearly
8:53
expressed by schools or teachers. That's very
8:56
interesting. Yeah. What I'm hearing you all saying is,
8:59
like, they have the potential to provide
9:03
the kinds of levels of support so that students can feel successful
9:08
or achieve some sort of proficiency in writing or reading.
9:12
But at the same time, as you say, Joaquin, it could be
9:16
just one more thing that they need to do to get there,
9:20
and maybe they don't. It may be even difficult for them to learn the
9:23
technology. Right. So now you have, like, two different tasks
9:26
in front of them that are. Positioned them as the vision.
9:30
Right. So that's interesting to think about. But I know both
9:34
of you wrote very interesting literature reviews on the topic,
9:38
and I have read both of them, and I will put them on the
9:41
description of the podcast for this week. So can you.
9:44
Well, while you're reading this literature review, can you share some
9:48
concrete examples that the studies were
9:52
saying these are tools that are showing some success
9:56
with X or Y students with learning differences?
10:00
Yeah. So I'll go first. Yeah.
10:03
So what we looked at in our Live review with Dr.
10:07
Ling Zhang, what we saw, we went back to look at using that
10:10
broad definition of AI to include. To include text
10:14
to speech, for instance. And Joaquin, I think, made a wonderful point,
10:18
because having a southern accent, as I do,
10:22
it doesn't always pick up me as well, you know, and it's just not something,
10:27
you know, I. So, you know, I totally
10:30
understand that point. But we looked
10:33
at that. We also looked at. We looked at robots that were being used
10:37
to support individuals being served with
10:41
autism. We looked at many of those kind of elements
10:44
to just step back and say, you know, all right, how.
10:48
How has. How have we. Using a
10:52
broad definition, like, what are the tools that we
10:55
have used that we can, as educators,
10:58
researchers latch onto to say this was AI this whole
11:02
time? Because AI, again, really didn't
11:05
pick up steam, concern, challenges.
11:09
You know, as you mentioned earlier, up until, what,
11:12
2022 with the advent of ChatGPT.
11:16
So what we did was kind of step back to show that, you know,
11:19
special education has been working on this for a long time and through
11:23
intelligent tutoring systems as well. You know, those elements
11:27
to just say, hey, we've been a leader in this space and, you know,
11:30
maybe we should also be a leader moving forward instead
11:33
of being something that is taken and moving. You know,
11:37
someone else takes this and takes the lead and starts running, which I feel that's
11:41
kind of where we are right now. So our piece was about
11:45
what has kind of worked in the past as well as to show,
11:48
hey, we're. We're actually the people that can move this forward as well.
11:52
So let's. Let's take the reins. Yeah,
11:54
well, a lot of those studies that I
11:58
was looking at back before that
12:02
were done before there was this massive explosion, were still
12:07
sort of in the positive discourse,
12:10
positive realm, oh,
12:14
AI can solve this. And they were really around a number of
12:17
issues around identification of
12:20
students, so saying,
12:24
here's how we're going to use AI to sort
12:27
children. And that was an issue that I actually brought up as potentially being
12:31
problematic. So it works, but like, at what
12:35
cost and to whom? And do
12:39
we really want to think about this? Sorry. To interview
12:42
Marie. How they work, how they were using the tool to identify students.
12:46
Well, with speech. So they were saying, how we are
12:50
going to. So this is what constitutes. We're going to put,
12:53
get. We're going to get a big database of speech and then we're going to
12:58
decide, we're going to let the AI decide what constitutes acceptable
13:03
speech. And then who doesn't meet that threshold,
13:06
then that's who we're going to identify as
13:11
having a speech disability. And you
13:15
can see how then that opens
13:19
up really a lot of potential issues.
13:22
And then of course, then they're going to find, oh, the AI worked for that.
13:27
And I think whenever we look at studies about AI because
13:31
people want to run out and do this right now.
13:35
I was the managing editor for Online learning for a bit,
13:39
and I'm the current editor in chief of The Journal of
13:42
K12 Online Learning Research. And people
13:49
do studies and they do intervention work and they say,
13:53
we didn't do anything for these people and we did
13:56
something for these people. And we found when we did something,
14:00
then it helped.
14:02
And, and the, and,
14:06
and of course.
14:10
And so then we have to think about, well,
14:13
when we change something, something changed what that
14:17
means. And I actually was
14:21
looking at a lot of stuff with a former grad student about language support
14:25
too. And these people doing ESL
14:28
EFL work right now saying, well, we used something
14:32
like duolingo to help people learn language
14:37
and it helped people learn language. Well, yeah, I would imagine
14:40
that if you gave somebody nothing and gave
14:43
somebody something, that the something would help.
14:47
And so then we should
14:51
be interpreting research with
14:56
caution and be really careful about, you know, when we say,
14:59
oh, these were successful, what we really mean by
15:03
that and you know, speak with some, some clarity.
15:07
And then also, like, I did this review,
15:10
you know, back in the day, now,
15:13
like at 2023, seems like eons ago,
15:16
like, but to say,
15:20
let's think about the ins and outs of
15:24
what this really means in terms of if we're going to put
15:28
a big bunch of stuff into a database and then let the database
15:31
tell us where the cut score is and then use that to
15:36
put children in boxes and put labels on them and
15:40
what the implications are for that for the.
15:43
Maybe the duration of their lives, but at least for their schooling.
15:47
Yeah, so we're doing that with or without the machine too. But I
15:51
understand your point. Yeah. I was just thinking
15:55
as a full time school social worker at
15:58
school specific schools, but also through the district wide programs I've developed,
16:03
there's an interesting reality and I
16:06
found difficult even in doing lit reviews around what is the
16:09
emotional process of learning? Right. What are students feeling
16:14
even before the lesson plan starts? What are the dynamics in
16:17
the environment either with peers or teachers or related to a subject matter
16:21
that has students either being open or closed down or reactive and defensive.
16:26
And I would say most related to this would be in a program
16:30
we started where we got grant funding to use manipulatives, you know,
16:34
fidgets, things that students can play with to help them self soothe.
16:38
Right. Which supposedly supposed to help them with another sensory listen
16:42
and pay attention and participate more. And I'm
16:46
not sure if AI can be
16:49
programmed right or if there's even an interest because we're not really
16:53
talking about the emotions even today, right? In schoolings
16:56
or in this podcast and in our research
17:00
and in our field, emotions are everything in
17:03
human interaction. While we're sleeping, while we're awake, whatever we're
17:07
doing, the emotions lead what we're able to or
17:10
what's being asked of us. Yet in education it's
17:14
only justified when there's emotions in the way, right?
17:17
When it's problematized, Right. Unless we're talking about gifted classes and
17:21
students who are just seen as are
17:25
labeled as so smart we have to give them
17:28
extra resources to keep up with their brains. But it means everybody else,
17:32
right? Everybody else in special education or everybody else with any learning difference
17:36
is seen as the opposite. So I'm wondering how literature
17:40
reviews can also be more holistic and say, well let's
17:43
not just deal with the emotions of the 10% limited
17:47
to 10%, even though there's way more students in education
17:51
that aren't passive learners. But what are the emotions of teachings
17:55
of teachers, of teaching staff? Because I think we're also assuming every
17:59
classroom will have a highly certified special ed teacher in it.
18:02
When in reality most classes function through
18:06
underpaid under trade educational assistance and
18:09
volunteers are long term subs that are contracted. What are
18:13
the emotions they're bringing to the fear of teaching populations
18:17
they may not feel comfortable with? And how does that cause students to react with
18:21
emotions that AI can't even be programmed to understand
18:25
or compliment. Well, AI is supposed to actually like
18:29
when you're talking about some things like autism Spectrum disorder,
18:32
that's they're supposed to train you, quote unquote, to have emotions
18:36
ostensibly because you don't have a person in the classroom who can
18:40
help you with that. That. Take me to one of the questions, right, about AI
18:43
doing the work of teachers. They've been saying that in a year we're going to
18:46
have this super AI. You know what it's called, the generative
18:50
AI, that it's going to be as
18:53
powerful as the best scientist in any area.
18:57
Right. Do you see in the future that there's a
19:00
risk that AI may take the roles of teachers, whether general
19:04
or special educators? A couple of things with that. First, what I
19:08
think what Joaquin just mentioned, which I really like, and that's the idea
19:12
that teachers are in theory we
19:15
have, and by the law we should have, you know, certified teachers.
19:19
And that of course is not the case all the time.
19:23
Mainly what I see right now in both the
19:27
literature as well as kind of funding
19:30
opportunities is that it is these are
19:33
AI that are being delivered for teachers to access
19:37
to deliver instruction to students. So I think that's
19:41
where we're kind of comfortable in our development right now.
19:45
And that means a couple of things. One, to just kind of take advantage
19:49
of this database of evidence based practices and
19:53
however you want to, you know, what, whatever you want to put in there
19:57
to say teacher, we were teacher
20:00
ask questions of the AI,
20:03
gets recommendations, then teacher applies recommendations
20:07
to the students. I think that's
20:11
most of what I see. And an example would be something like using
20:15
AI to write strong
20:19
IEP goals, using AI to determine
20:25
what evidence based practice works
20:28
for students. There's some challenges
20:32
there for a couple of reasons. One is that
20:36
for instance, if you just put in academic data to ask
20:39
for recommendations as to what student, as to
20:42
what intervention might support student, one you don't
20:46
like. Academic data is but 1/100th of what it
20:50
means to be a human and a student, right? So I mean,
20:54
it's really difficult to make really strong decisions
20:58
on what intervention will support that student just based off
21:01
results of data. The second problem is that whatever
21:05
that recommendation is, it assumes a couple things.
21:08
It assumes one, that the teacher has access to the resources,
21:12
that they have the knowledge and that they have the support
21:15
from, they have support from administration,
21:20
whomever to actually deliver that
21:23
intervention. So you know, it would be like for
21:27
instance saying, you know, student performed this, you know,
21:30
performed at this level. And the AI says,
21:34
well, all you need is this reading
21:37
program. Well, this reading program is a, is a,
21:40
you know, is a, is more of a school wide initiative. I think
21:44
that's kind of where I'm seeing this right now.
21:47
I think I know what you're talking about. The 2026 debt kind
21:51
of mark where he says, hey, we, where we've got the, we've got this new
21:55
giant, we've got access to this new tool that
21:59
is going to revolutionize. I think we'll probably
22:03
push back on that a bit because where we
22:07
started with this were AI that were developed for students
22:11
to use. Then where we are in this space right
22:15
now is AI developed for teachers
22:19
to access to, then deliver instruction
22:22
to students. And then I think we will revert back in
22:26
some time. It might not be 2026, it might be later
22:29
to then using AI to develop for students to use,
22:34
you know, own their own. So it's kind of like,
22:38
it's kind of like we're seeing, it's kind of like we're seeing waves of AI.
22:41
Wave one, which we talked about earlier, wave two, where we are right now,
22:45
these databases. But that is the human in the
22:48
loop idea, right? You know, it's like we've got AI, but we're
22:52
going to have someone access that AI to then apply
22:56
to the student. And therefore we do have some control
23:00
over the AI because we can, again,
23:03
assuming that we have, we have the skills necessary,
23:07
we can make the distinction. And is the AI producing
23:10
something that we believe in that is void of the bias
23:14
that we're going to talk about, no doubt here a little bit later?
23:18
But I also think that there's some design in making
23:22
sure that like AI
23:26
is supposed to be doing this work and even if teachers are
23:31
here, then their work can be replaced.
23:34
So I mean, if you think about right now, there is
23:38
a deliberate effort to take people who are
23:42
in positions and remove them from those positions so
23:46
that AI can do work in government sectors
23:50
right now. And so, and I think that like
23:53
when we talk about AI discourse, one thing that's frustrating to me
23:57
is that we, we kind of are like, oh, AI took
24:01
your job or AI placed you in special education,
24:04
or AI took your services away, but AI
24:08
didn't do those things, a person did. And we can say that AI did all
24:12
we want and that it was the result of a database making these decisions,
24:16
but. And it sort of helps us like
24:19
remove responsibility from whoever the powers
24:23
are that be. But. And I think it's designed to do that
24:26
it's designed to be a responsibility removing tool.
24:30
But there is a deliberate
24:34
effort to funnel public
24:39
money into private hands and AI happens to be a tool that
24:43
rapidly is going to assist in that. And so then,
24:47
and I and Richard for a long time has said,
24:51
you know, I've known him since probably
24:54
2014 has said for a long
24:58
time it's a tool and people are going to use this tool to, you know,
25:02
do. People are going to look at the tool and they're going to try and
25:04
decide what, how to, how to do a thing that they've wanted
25:07
to do for a while and then try and find new problem problems.
25:11
And Neil Selwyn talks about that too.
25:14
And technology people look around for problems that
25:19
weren't problems until tech could come in and do surveillance and
25:23
different things. But you know, but we have to like,
25:26
you know, make sure that we're clear eyed about that too. That yeah,
25:30
there's a, definitely a special education teacher shortage
25:33
and it's more acute in
25:36
some areas of the country than others. So there's some nuance to it in
25:40
that where there's only a little bit of a shortage. It's actually, it's a
25:43
different market than where there's very bad shortages, which I think
25:47
is interesting. And then the shortages in New Mexico
25:50
are different because of the linguistic landscape of New
25:54
Mexico versus other landscapes.
25:58
But the, you know,
26:01
people aren't going to care when they just come through and say I
26:05
got, I got the, I got the solution. So yeah,
26:09
I mean I think one of the things that you say, Mary, is one
26:13
of the problems also the dangers
26:18
of this tool right now is the particular context is where it's going to be
26:21
implemented in particular context where it's going to be expanded.
26:24
Because as you say, we're in a very particular
26:27
in the U.S. cultural and historical content in where you
26:31
say there were dismantling civil rights protections for students with disabilities.
26:35
We just have a show, last show was about the Becerra vs
26:38
Texas case about 16 states asking
26:42
for declaring Section 544 as an unconstitutional.
26:46
It's happening in a context where
26:49
we're trying to expand vouchers for students with disabilities when we know that to
26:53
go to a private school they're pretty much giving up all the idea
26:56
rights. So I think one of the interesting things that are
27:00
coming from our conversation is not so much about the tool
27:04
but also about when the tool is going to be expanded, exploded and
27:08
implemented in this very tumultuous time
27:12
where the tool may ended up
27:15
using for some of the uses that you were mentioning Maris, that are
27:19
quite problematic. I know Joaquin
27:22
wanted to say something, but I just wanted to jump out and clarify
27:26
that. Yeah. And that's really how I was thinking along those
27:30
lines. We're having a wonderful conversation about the ethics,
27:34
not just the ethics of school or the ethics of teachers in a teacher
27:38
role. But, but I want to be, you know, I want
27:41
us to be clear. A lot of learning isn't
27:45
isolated to a dynamic between one teaching person
27:49
and one student. It's not a didactic,
27:52
one directional process. Right. And a lot of
27:55
teaching improvements, I would say, have to do
27:59
with the ethics and that dynamic. Because more than
28:03
anything, I would say let's take an IEP meeting, for example.
28:07
A lot of times those are dynamics where somebody
28:11
who has a license to assure things are
28:15
ethical and healthy on behalf of a student is faced with
28:19
a team of people who only want to talk
28:22
about a product or a score or a number.
28:26
However, there's more learning for these individuals
28:30
in hallway conversations around. Is is what you did and
28:33
said in that meeting or in the classroom,
28:37
was that really ethical? And how do we overcome the detriment it had on a
28:40
student? Right. So my question then becomes,
28:44
can AI have an ethical component or recognize when
28:47
it's being used unethically in place of a teaching certified
28:51
staff or being used as a tool harnessing
28:55
power and dynamics? Right? Because classrooms are very,
28:59
are a very territorial, they're a very power dynamic
29:03
space. Because again, schooling historically and
29:07
today has always been seen as a place of control for passive
29:10
learning. And anything different than that is usually met
29:13
with eventual discipline,
29:16
public external discipline. But initially energy,
29:21
personality, charisma and language
29:25
dynamics of power and control to either cohorse all
29:29
the way to demanding students passively
29:32
agreeing with how. Right. How lesson plans are being
29:36
implemented? Because let's be clear, right?
29:39
Curriculum and lesson plans really tend to operationalize
29:43
any type of existing conscious or unconscious bias
29:47
of teachers. Now if we're saying now AI is going to be
29:50
involved and AI is an instrument that will exponentially
29:56
carry forth whatever the teacher is doing. And the teacher doesn't
29:59
even include ethics or holistic learning needs of a
30:02
student, then what, what are we saying the product is going to be
30:05
that ultimately a student will be held responsible for,
30:08
right? The student's grade. The student's product,
30:12
now with some unknown algorithm software
30:15
is going to be seen of whether or not they're a good student or not
30:18
based on if the teacher even considers ethics
30:23
or holistic needs of a student and then now has an algorithm
30:27
replacing Some of what they're doing. Right.
30:33
Richard? Yeah. And just
30:37
building off what, what Kane just mentioned, Mary and I
30:40
met in 2014. We were working in the center on Online Learning,
30:44
Students with Disabilities, the universe of Kansas.
30:48
And it's, it's really interesting how
30:52
these two kind of parallel show.
30:57
Online learning had been happening for a long time, exciting until Covid hit and
31:01
then it was not exciting as exciting anymore. However, it is something
31:05
to think about because if I think
31:09
there is some traction for this national voucher program
31:13
that will potentially move forward,
31:16
and when that happens, that traditional classroom really
31:20
is going to be disrupted across the country,
31:24
offering lots of different ways that students do access.
31:28
And what I loved about what Keane said, because it's something that Mary and I
31:31
would talk just ad nauseam about, was really that,
31:34
like those moments that, like, how can AI,
31:38
if AI is actually your, your a
31:42
major source of your instruction, how can it replace
31:46
seeing your teacher in a grocery store? Like,
31:49
what do you lose? Like, what do you lose in a hallway
31:53
conversation with your friends? Like those type things?
31:56
And Mary and I, I mean, again, here's the thing about. It's like,
32:00
you know, there really is no expert in AI. AI right now.
32:03
If you meet someone that says they're the expert, I would just go the other
32:07
direction. It's all conjecture. We're all learning, we're all
32:11
at the same. But mainly what it is is to
32:15
think about that element right there. What Joaquin mentioned,
32:18
because what can happen is,
32:23
you know, families make decisions about where they want their students to
32:26
learn. There's going to be, if, if this goes through, there will just be
32:31
multiple options for them too, but you're going to
32:34
lose that piece. There's AI is the,
32:37
is the P that cannot be replicated.
32:40
So ensuring that thinking about moving forward,
32:44
that is going to be the part that we, that,
32:47
that we really need to be mindful of. AI can
32:51
do a great job, an amazing job of delivering instruction right
32:55
where you need it, answering any question you may have.
32:59
But when it comes to that fabric of society,
33:03
that's. That's not something that AIs is
33:06
going to be able to do. And I'll be quiet right after this because we,
33:10
we, Mary and I will, I'll do our best. Our best line
33:13
from all of online learning was, and now
33:17
using AI is that, you know, John Dewey said that school
33:21
was to produce the best, you know,
33:26
to produce democratic, you know, individuals ready for the democracy.
33:29
Right? And so our thing was, so the, the joke was,
33:33
okay, so the best way to do that is to remove them from the democratic
33:36
society. Right. The best way to prepare an individual
33:41
to be a citizen is to remove
33:44
them from the only mandated.
33:48
I think that is super challenging and something that could
33:55
continue to happen.
33:58
It seems like that relational aspect and
34:02
so far unreplaceable. Now I want to touch a related
34:06
comment a related question about the transparency of
34:09
these tools. They seem something that may be like a black box and
34:13
the question is, you know, who controls the black box? What's in
34:17
that black box? Making decisions and how this transparency
34:20
can affect the usefulness for educators and
34:24
students and parents. And we
34:28
mentioned already these databases that are. That are behind these
34:32
machines are. Can be biases. There is, there is some lot
34:36
of algorithms and data in where
34:39
those algorithms are operated that
34:43
may not be completely known to teachers,
34:46
may need completely known to administrators or parents and they
34:50
may be having. Making a lot of decisions about students
34:54
life. Yeah, I would say that transparency is actually pretty low,
34:57
but that's part of the power structure of it. So as
35:01
if you tell people a lot about. There's actually been some
35:04
studies of young people of Gen Z ers and below
35:09
where when you tell them more about how what AI is and
35:13
what it does and they're less interested in using it. So then,
35:20
so then you may not have folks that'll do that. So when I the
35:24
article that I just published when
35:27
I was talking to these kids about you know,
35:30
just very basic things about. Because I'm not a.
35:33
I'm a. I do have linguistic training but I'm not an
35:37
engineer about what AI is
35:40
from a language standpoint and very basic things about
35:46
databases and helping them understand very large numbers
35:50
and things like that. Then they made different decisions about what they wanted to
35:53
put into the chatgpt and
35:57
what kinds of feedback they were asking about it. But. And I found
36:01
myself a lot of times saying to them well, you're the writer.
36:05
And when I was in. When I was a writing teacher and I'm still
36:09
a writing teacher I guess in some ways like my students are just
36:13
quite a bit bigger. So I would give
36:17
them feedback sometimes and I would think well I'm the teacher so they should,
36:21
you know, take my feedback. And then it was down the line right
36:25
As I progressed emotionally and ethically that where
36:29
I would tell them more like well you don't have to do
36:32
this. This is just my response rather than feedback
36:37
and then they're the writer. But a
36:40
lot of times as teachers we think you should just do this because I marked
36:44
it and helping them understand
36:49
that AI is Not
36:52
something that is universally correct or right.
36:55
But that's not the. When people
36:59
make these things, that's not how they want you to think about it. They want
37:02
you to think about it as indispensable and is
37:06
the only way to consider the problem. And they
37:09
want you to make it a partner in all of your decision making.
37:13
And they actually want you to make you think
37:16
that it is somehow at least partially sentient and that it can come
37:20
comfort you. And you know, because like, I have
37:23
students write things to me all the time about different
37:28
things that have happened to them. And even kids used to write
37:31
stuff about, well, there. This one story
37:35
I got is a grand.
37:38
They were going to miss school because their grandmother was
37:42
repatriating to Mexico to marry the man that she
37:45
couldn't marry 60 years ago because
37:49
even though she was pregnant because the family didn't approve. And so she went
37:53
to. Came to the United States and married somebody else. And then the secret all
37:56
unraveled and they found each other again and she repatriated
38:00
and married the. Married the father of her actual oldest.
38:04
And I can't understand how you'd read a story like that
38:08
from a child and be like, well, you got to put some headings in here
38:12
and maybe it needs some more details. So,
38:16
or, you know, those sorts of things. And, or even,
38:20
and even if ChatGPT or one of those things can tell you,
38:24
oh, that's a very compelling story or, oh, that sounds like a good plot for
38:27
a novella, it's not the same as if you really engage with a
38:30
child. And I'm also, I need people to understand I'm not nostalgic
38:34
or not naive about some of the fact
38:38
that adults on the landscape are
38:41
not always great to kids because I go to schools all the
38:44
time and I see a very wide range of adults
38:48
in their ability and interest
38:52
in engaging with them. And so the,
38:55
so there is this transparency piece about what
38:59
this is and what this does and who we're going to share it with.
39:03
And again, states have taken this up differently. Kentucky is
39:06
very much like, oh my gosh, we need to tell the parents all
39:09
this stuff. But a lot of other states have,
39:14
you know, have, have not come out with that strong of language. I think that's
39:18
a wonderful transition. Right. And, and I think the nostalgia of
39:22
us, his story, right,
39:26
of always making reference to Manifest Destiny, the wild, Wild west,
39:29
where we're exploring new lands. I mean,
39:33
let's be real, you know, the, the genocide of indigenous
39:36
people wasn't complete. Right. And the land grant
39:40
Universities and the lands public schools are on are still
39:43
considered stolen land by people who still live in the area. Right. So there's this
39:47
colonial reality still playing out in schools that I think is not
39:51
even being considered. So if we want to talk about transparency,
39:54
I think transparency between a learner, what could be identified as a learner, and a
39:58
teacher, which again is a very didactical one directional
40:02
reality that I don't really think is the truth.
40:06
But transparency on a bigger level, I would say how can
40:10
AI or does the school or does
40:13
administration or policy even want AI to address if families even
40:17
understand schools and schooling? Right. And let's
40:20
think about families who show up to parent teacher conferences or
40:25
IEPs, or just receiving communication of status of
40:29
how their students doing throughout a semester. Right. Do schools
40:33
even ethically take on the responsibility of making sure
40:36
families understand what learning curriculums are? Right. The reality of
40:40
students having to not just go to different classrooms.
40:43
Right. In early childhood education, which was my last job at the district,
40:47
of course, it's mainly one teacher with some supplemental adults, but the rest
40:51
of education is students being in several locations and
40:55
having to learn to manage several different adults,
40:59
personalities, temperaments, learning styles.
41:02
And is that part of the transparency we're going to ask AI to translate for
41:06
families and students or is AI just going to replicate and compound that?
41:10
So now there's not just several adult spaces, but now several versions
41:14
of AI that students will have to address. But the teachers are
41:17
just gung ho by saying it's the wild wild West. Let's see what this
41:21
new contraption is going to do on top of students.
41:25
Like we've done it on top of indigenous land and on top of indigenous people
41:28
that still exist. So I think the transparency piece,
41:32
I would love to know an algorithm or that it's being programmed to really help
41:36
a full transparency for communities,
41:39
families, students to better understand learning.
41:43
But that's going to get into some very delicate spaces
41:46
of if teachers think it's being used to monitor them.
41:49
Right. Their performance, their abilities to do
41:53
and carry out policies that come from federal or state, whether or not
41:56
they agree with it. Yeah, I mean, one of
42:00
the things I think you're mentioning is like if
42:03
the issues of, you know, transparency and understanding schooling
42:07
and the context of schooling, in the case you were mentioning, the example
42:10
of talking about the colonization and erosion of indigenous
42:14
people from certain lines, if we haven't done until now,
42:17
why do we expect AI is going to solve that problem? Because it's being
42:21
implemented on that particular context. So it's just A tool.
42:24
I don't, I don't see that's going to come as a redemption to that because
42:28
we haven't done it for hundreds of years. So why
42:32
AI was going to come and suddenly fix this.
42:36
That's a wonderful reality, right, that we have to say AI then is
42:40
now becoming a part of colonial schooling. Right. Because schooling in the
42:43
US was made to deculturize,
42:47
delanguage, assimilate and attack
42:50
the spirit of not just indigenous people, but also Mexicano and
42:54
Latino people through mission schools. And the policy and
42:57
the values of colonial oppression are within the
43:01
policies, within the power dynamic of
43:04
what schooling is. Unless, unless a student
43:08
happens to be in some very different small learning
43:12
environment with all of who they are, is encouraged to flourish and however that learning
43:16
needs to take place. Right. And it requires the natural
43:20
resources that are. That further denude the
43:24
land. So which I
43:28
have written about. So,
43:31
you know, I have a few, a few questions remaining. So we talked a little
43:35
bit about some of the potentials and some of the limitations
43:39
of AI. I want us to spend a lot of time dreaming,
43:42
thinking about what are the possibilities?
43:46
I mean, how would you would like to
43:49
this movement or AI be
43:53
implemented or used in the future? Right. So let's,
43:56
in very, maybe transformative and redemptive ways
44:00
to change some of the problems, the current problems that someone mentioned.
44:04
So I'm going to start with Richard and we'll move around. Let's dream
44:09
about how this can become a tool for actually transformation
44:13
rather than just reproducing the same inequities that we've been dealing
44:16
for the last decades and decades. Thank you.
44:20
So I think a bit is to demystify
44:24
AI. These tools,
44:28
these products are increasingly
44:32
easier to create and therefore would
44:37
be that in your local context,
44:41
individuals create their own, support their
44:45
own, you know, those type, those type things, whatever your needs
44:49
might be. Like right now we feel like we need venture capital
44:53
to come in to, you know, you know,
44:56
invest $2 million to create something that we're going to use
45:00
across the country. And I think that's really the wild, wild west space
45:04
that, that I'm seeing is really individuals realizing just
45:08
how much revenue can be generated from
45:12
education, which I think is absolutely the wrong way to look
45:15
at this, but that's really where that
45:20
space is. And ultimately here is an opportunity to
45:23
leverage AI to
45:28
be able to empower the students within that context
45:32
to one, address their own challenges.
45:35
Two, really, you know, empower them to think
45:39
about what their lives may be like moving
45:42
forward, maybe opportunities for them and for the teachers and
45:46
for the community to collaborate around. All right, what tools are
45:49
important to us? How can we, how can we do this? And again,
45:54
it is, you know, it's kind of like you've
45:57
got some entrepreneurs that are trying to,
46:00
that are trying to really go in and generate
46:04
revenue out of education. You've got some entrepreneurs who are really saying
46:08
AI is powerful. So let's create canvases
46:12
that anyone can use. Then that does
46:16
still. And that does put a layer of, you know,
46:20
of where families, students, teachers,
46:24
whatever could really address those outputs that we've talked about
46:27
over and over today as we should be talking about. But that's
46:31
that kind of layer of critique that really does have
46:34
the potential to support students. So I would say maybe
46:39
taking that under kind of what we've known about the maker space,
46:42
empowering students to, and families and community members to
46:46
work together to create the tools that they need. There's,
46:50
there's lots to think about in terms of dreaming and also
46:54
in terms of making sure that people have information
46:57
that they can use to make the decisions that they
47:00
want to use. Which in some cases means that the
47:04
use of AI in education would be very minimal.
47:08
And in some cases it may not mean that.
47:11
So AI, if you
47:15
go talk to some people who have been
47:19
doing engineering work, they've had AI for a long time,
47:24
so many decades. And it was unimaginable
47:28
that you would just take it and bring it into an educational space
47:31
because it was supposed to be for things like my
47:35
brother in law's an astrodynamicist and he went down and worked for NASA for
47:39
a little bit and his job was to recalibrate
47:42
like this, this satellite
47:46
with a bunch of different mirrors on it to make sure all the mirrors were
47:50
the exact right angle to take the picture. Now that's what AI was for,
47:53
or let AI find your cancer.
47:58
So there's, I mean there's not like there wasn't like good uses
48:01
of AI and there wasn't supposed to be. But we have to think about
48:05
the, what are the really good uses of
48:10
AI for people? And then, you know,
48:13
if you. There's some, There's a thing. There was a question about book recommendations
48:18
and Brian Merchant wrote this great book that we
48:22
talked about one of the classes that Joaquin and I did together.
48:26
And it is about the Luddite movement and the Luddites
48:30
aren't people who hate technology. It's about making
48:34
sure that the people who.
48:37
The benefits are distributed and the
48:41
harms are reduced, especially to the people who have
48:44
Always been paying the price.
48:47
And so I went down to a conference in Chile
48:51
summer ago, and people just kept saying,
48:55
well, this is the price of progress. This is the price of progress.
48:58
And I'm not the person who normally pays
49:02
the price of progress, but I have to go in schools and work,
49:07
or I get to go in schools and work among people who
49:11
have been paying the price of progress for 500 years.
49:14
And that it's insulting that the same people
49:18
keep paying the price of progress. And so
49:21
my dreaming is all about figuring out how
49:26
we've got to figure out the right ways to use the stuff that is going
49:29
to reduce the harms and
49:33
expand the benefits and make sure that people
49:37
have the most information possible so that they can
49:41
do the things with it that they are wanting to do.
49:46
When I went and looked at the 50 largest school districts in the United States
49:50
to see how they were using AI, then even within states,
49:55
you have one school district that is like, oh, we had an existing
49:59
computer science initiative, and so all of our students get computer science
50:03
and we just folded it into there. And the neighboring school district said, we're going
50:06
to use AI to body scan students when they come in the door.
50:11
And so that's not okay.
50:17
And my fear was that there are some groups of students who are going to
50:20
get to use AI to send rockets to the moon,
50:25
do fancy stuff, right? So I
50:29
go work with students who are doing all kinds of things with looking at
50:32
space, and they're doing it right now, we're designing weather
50:36
balloons and things like that. And then other students.
50:40
It's going to be a tool of, you know, doing the
50:44
same thing in education that we usually do, but more
50:48
of it and harsher so
50:52
and so. And who's that going to land on?
50:56
Yeah, I really like the examples you're giving. Right. The contrary
51:00
examples. And I remember at one point I was in a really small charter school
51:05
and they barely had resources for Chromebooks
51:09
or something, I believe it was. And we ended up finding a sixth grader
51:12
that was hacking into government websites at lunch. I mean, that was his
51:16
brilliance, right? He could code in ways I didn't
51:19
even understand. And so how
51:23
do we try to find when we dream? How do we find healthy ways that
51:27
technology can unleash solutions and realities
51:31
to benefit of everybody? Now, I didn't say that as the benefit.
51:34
What I'm saying is once we had a robotics program and that student
51:38
was able to harness and channel that coding,
51:41
well, then they went to state and regionals and they won in ways that nobody
51:44
Imagined right. The first year they were there. And so for me,
51:47
you know, if we want to dream about technology and AI,
51:50
I think, well, can we get
51:54
over our own fear as adults of saying, well, maybe the students should be the
51:57
ones in a healthy way programming AI and we
52:01
might actually see a level of scaffolding, a level of varied
52:05
teaching and meeting students needs. Not because AI will
52:08
do it, because maybe they need to teach AI to
52:12
teach the teachers what they need holistically,
52:15
right? Not in an algorithm way that nobody understands, but to lay out
52:18
in very clear ways to teaching staff. This student from all
52:21
of you will need these various things. Things, right.
52:25
And make that concrete. Because let's be honest,
52:29
student voices in the history of schooling have, have never been a formal part
52:32
of learning. Right? It's never been a multi directional process
52:37
focused on relationality. And if we need technology to
52:40
highlight the relationship that should exist and the negotiation between teacher
52:44
and learner and that the teacher probably should be learning
52:48
more than students on how to be a better teacher all the time through
52:52
their own cognitive algorithm. I could dream
52:55
maybe AI will help point that out and then sound an alarm when the teacher
52:58
gets lazy and is just spitting out stuff to the same class
53:01
every time. Right? Or over disciplining, AI will sound an alarm
53:05
or AI will raise its hand, you know, like, you know, the 1980s movies of
53:08
robots and say, you know, little Johnny needs X, Y and Z when
53:12
the teacher's purposefully or maybe unconsciously overseeing
53:16
those things. Right. Wow, those are
53:19
great dreams. Well, thank you, the three of you for delighting
53:23
me and reach our audience understanding It's
53:26
a complex topic and there's so many. I mean, I have more
53:30
questions and many things to say and
53:35
we'll keep people thinking about it and debating between them.
53:38
Thank you so much, Marie, Richard and Joaquin for being here with me
53:42
today. Thank you, thank you, thank you.
53:50
Thank you for listening to Dive In. I hope
53:53
you learned from this episode as much as I did.
53:56
Please help us to spread the word about the show. See you next
54:00
time.