0:06

Welcome to Demystifying Instructional Design, a podcast where I interview

0:10

various instructional designers to figure out what instructional designers do.

0:14

I'm Rebecca Hogue, your podcast host. If you enjoyed this podcast, please consider subscribing or leaving

0:20

a comment on the Show Notes blog post and consider

0:23

helping to support the podcast with a donation to my

0:26

Patreon account. Welcome Autumm and Lance to Demystifying Instructional Design.

0:32

This is a bit of a different episode of a

0:34

podcast. We're going to talk today a little bit about ChatGPT

0:37

the AI phenomena that is lighting up higher ed anyways,

0:43

if not other spaces as well.

0:45

And so the first thing I'm going to ask is if you could do a quick introduction and give us

0:49

a little bit of context. And I'll start with my context is I teach instructional

0:55

designers. So I'm teaching at the master's level and my students

0:58

are instructional designers. And so for me, I'm looking at it largely as

1:04

is this tool useful for my students?

1:08

That is the context that I have for it.

1:11

And I'll pass it over to Autumm and then Lance.

1:14

Hi. I'm Autumm Caines.

1:16

I am an instructional designer in a faculty development office

1:21

at the University of Michigan, Dearborn.

1:24

I always think it's important to add what office you're

1:27

in when you're talking about instructional design, because as Rebecca's

1:30

audience will know, the context in which you work is

1:34

an instructional designer can have a huge impact in the

1:38

type of work that you do. So coming from the perspective of a faculty development office,

1:44

it's more than just instructional technology and it's more than

1:47

just instructional design. It's also faculty development.

1:50

We can get into that a little bit more as we go forward.

1:53

But just to set up the context, I'm also instructional

1:57

faculty at College Unbound where I know Lance from, and

2:00

I actually know Lance from this podcast as well.

2:03

That's one of the reasons that I reached out to

2:05

Lance was because I heard him on this podcast and

2:08

then we ended up becoming colleagues at College Unbound,

2:11

But I teach two classes at College Unbound.

2:14

One is Digital Citizenship and the other one is Web

2:17

and Digital Portfolio. So that's just a little bit about my context in

2:22

terms of where I'm coming atthings from, in terms

2:24

of ChatGPT. I have been looking into large language models since

2:31

probably 2021 with the upset with the firing of Timnit

2:37

Gebru at Google from when she was working, she was

2:41

the head of Ethics and was working on the Lambda

2:43

large language model and started paying attention to some of

2:47

the advances that were happening in that technology around that.

2:51

So I do have a tendency to come at it from a little bit more of a critical and ethical

2:56

perspective. But I don't want to go for too long.

2:58

I want to turn things over to Lance and let him introduce himself.

3:01

Sure. Hi, I'm Lance Eaton. I am Director of Digital Pedagogy at College Unbound.

3:06

I would say what department I'm from, but we don't

3:09

really have departments because we're still a new enough college

3:12

and have that new college smell where there's lots of

3:14

different hats and ways that we try to work around

3:18

action teams as opposed to traditional departments.

3:20

But if I did, it would probably be Academic Affairs,

3:22

which is I think where I mostly sit. And my role is that mixture of working with faculty,

3:28

doing faculty development and kind of helping to support them

3:32

in the development of teaching and learning courses and using

3:37

different tools. And sometimes that tool is a pen and pad and

3:40

sometimes it's an LMS and sometimes it's artificial intelligence.

3:45

For me, I've had a lot of interesting thinking around

3:50

or just looking at AI for a couple of years

3:54

now. I just started to develop a more critical view of

3:57

technology over the 2010s. It started to pop up on my radar and then

4:03

really for much of the from the start of the

4:06

pandemic until bout a year and a half ago, I

4:08

was I guess that was halfway through the pandemic or

4:11

whatever phase we're in now. I was working at the Berkman Klein Center for Internet

4:15

and Society and was both helping to run programs and

4:20

programing around Internet and society, and a lot of that

4:24

focused on AI. And so I got to see a lot of different

4:28

people in the industry coming with those critical lenses.

4:31

And so that stuck in my head a lot, especially

4:33

as ChatGPT became like the pet rock of the 2022-

4:38

2023.

4:42

And people really started to pay attention to a AI

4:46

generative tools in a way that they certainly hadn't previously.

4:50

Thank you both very much for your introductions and a

4:53

little bit of context there. I'd love to ask you a little bit about what

4:59

guidance you're giving students regarding

5:02

the use of Chat GPT And if you could tell me what that stands

5:07

for, that would be really helpful because I think the

5:10

audience would find that useful.

5:13

It's a generative pre-trained Transformer ChatGPT

5:18

And I do think that Lance is the perfect person

5:21

to answer this. I'm going to answer really briefly and just say that

5:26

I was a little taken aback when this was first

5:29

opened up, and I really wasn't sure what to do

5:31

as a teacher. As an instructional designer.

5:33

I had some ideas, but as a teacher I felt a little lost.

5:36

And working at College Unbound, Lance is the person that

5:39

I go to. When I have questions, right, And I knew people were

5:43

going to be coming to me. So I went to him and the school just responded

5:47

in an amazing way that Lance is going to tell

5:50

you more about now.

5:51

Thank you. Yeah. So this is again, being a fairly new school

5:56

there's things we don't have always to fall back on.

6:00

And also a lot of our practice and thinking is

6:03

student centered. And so I was playing with it, thinking about it

6:07

as well along in terms of instructional design and students and

6:09

whatnot. And I get an email from Autumm saying, I think

6:14

I have a student that has used it. And so that generated a discussion between me and her,

6:19

and it was at the end of the semester and like there's so much else going on, it gave us

6:24

an opportunity to really think about in this moment what

6:26

is what is most useful to do.

6:30

And this is where I think both of our collaborative

6:33

nature and the way that CU structures itself is to

6:36

go after the student wasn't the right approach.

6:39

There was a part of our reaction that was just like a hot dang! Like, go student for being that quick

6:45

and figuring it out. We had that moment and we celebrated that for a

6:48

moment. I want to give this person like some kudos for

6:50

creativity. We had the moment of just having that frustration and

6:55

a mixture of just not... Sometimes it's a little ego driven, but like not being

7:00

happy that the student did it. And then we just also were like, What's behind this?

7:05

And I think that's where we really got our momentum

7:08

and what has gotten our momentum for our school as a whole is really understanding the students and their uses of

7:12

this. I go back to I want to say it's folks

7:16

like James Lange and when people are doing this type

7:21

of thing, which is framed in all sorts of deficit

7:24

language of they're cheating, they're stealing, they're what have you,

7:28

they're often indicating things aren't working for them.

7:31

They're often indicating this is more of a sign for

7:34

help, a sign for a lack of trust, lots of

7:36

different things. And so very quickly, me and Autumm realized like, why

7:40

don't we try to find out? And so our first goal was to craft an email.

7:44

And I sent it out to students saying, Hey, this

7:46

tool exists. And I think some students may have used it and

7:49

we are interested in learning more about it in this

7:52

non-punitive way. We just want to understand what led you to it.

7:55

What might we be able to learn about why you

7:58

found your way to this tool? Nobody responded to that.

8:01

We hoped... nobody did. That was okay.

8:04

My partner recommended, well what if you did an anonymous survey?

8:07

And so we put that out to our students at the very end of the semester.

8:10

We probably got, I think, four or five students, and

8:14

this was at the end of the semester and during

8:16

break that before everything really exploded, before you were seeing

8:20

references to it in podcasts and in mainstream stuff, we

8:24

had three or four students who were saying like, yeah,

8:27

like we started to use this and we're using it in these ways.

8:30

And so we thought that was interesting. And that kind of jolted me to think about the

8:35

semester for us started January 9th.

8:38

This is before most schools start their fall semester.

8:42

And so we needed something in place. And so we developed a policy that we felt was

8:48

like... recognizing we don't really understand the fullest implications of

8:53

these tools. We don't want to just do a blanket ban and

8:57

be like, Nope, you can't use it under any condition.

8:59

And we wanted to create safe conditions that if students

9:02

use it, then it can they can identify that they

9:06

use it. So that can also invite questions and understanding and things

9:09

like that. I think the potential for it can be is as

9:13

vast as some of the challenges and the concerns around

9:16

it. Right. So there's a lot of knee jerk reactions.

9:18

There's a lot of valid reactions about the ways this

9:21

interferes with how we demonstrate learning by students.

9:25

But I think there's lots of possibilities for us to

9:28

leverage it. If we can find versions of it that aren't steeped

9:33

in all sorts of exploitative practices. But I'll pass to Autumm for her take.

9:37

Yeah. So when I saw some responses that I thought potentially

9:42

could be synthetic text, my first thought wasn't like cheating.

9:48

My first thought. like Lance said, it was curiosity.

9:51

It was like, wow, if they figured this out at

9:53

the same time, I did worry.

9:56

It's so new. It's such a new technology.

9:58

This is December of 2022.

10:01

It dropped on November 30th of 2022. The technology has

10:06

been around for a while now. We can go all the way back to the sixties

10:09

if we're just talking about chat bots.

10:12

Eliza. was in the sixties. Weizenbaum's ELIZA.

10:15

But in terms of the transformer technology, the idea of

10:19

using neural networks around large language models to

10:23

be able to create text so smooth and so clean.

10:27

And so it sounds so convincing, right?

10:30

That's been around for probably about two years now, but

10:33

you always had to pay for it. It's always been behind some kind of paywall or

10:37

part of some type of product. I mean just to open it up to the entire

10:41

world right at the beginning of finals for higher education.

10:45

That's not insignificant.

10:47

Whenever I talk about this, I always think it's really

10:50

important to put the context around it.

10:52

Yes, it's a huge jump in terms of technology, right?

10:56

In terms of the tech that is going on, the

10:59

interface, the way that you use it, the smoothness, all

11:02

of that. But a big part of the hype, a big part

11:05

of everything that's going on around this has to do

11:08

with the fact that it's free and the timing in

11:10

which it was released. Those are just two really big parts of it.

11:14

And especially in that moment, I was hearing all of

11:19

these news articles and things coming out about people being

11:23

really punitive with their students.

11:25

I read an article about a student who failed an

11:29

entire class when it was discovered that they had used

11:32

this tool. And it's my understanding also, there's really not a way

11:37

to prove without any doubt whatsoever that a student used

11:41

this tool. You can run the text through some of the detectors,

11:44

but those are flawed. They're flawed in my testing of this.

11:49

And I don't think anybody even tries to pretend that

11:52

they can return a 100% positive or 100% negative.

11:57

There's tons of false negatives and false positives.

12:00

So I really... I guess I say all of this to say I'm not surprised that the students, when we

12:04

sent them that email and just said, hey, we're just curious, did you use this?

12:08

Have you heard of this tool that nobody responded?

12:11

Because like I said, there were tons of articles and

12:15

news releases out there with people saying that they were

12:18

punishing students for using this tech.

12:21

So if I were a student and I used it out of curiosity

12:25

and I also think it's a little bit crazy to say that students wouldn't use it if I were a

12:30

student, I'd be curious. I'd want to try it out at least, and I

12:34

don't know if I would actually submit that work, but

12:36

I'd be tempted to, especially if it was the end

12:38

of the semester. And I was really busy and I had a lot

12:41

of pressure. I don't know of any academic integrity statement from any

12:47

university that mentions AI generated anything.

12:52

I don't know of any classroom policy that mentions any

12:55

of this kind of stuff. So I think it really does challenge us to think

13:01

about what we mean by cheating and to critically evaluate

13:06

and take, retake stock in what we mean, what's valuable

13:10

in an education. I felt really lucky that I was working for

13:14

College Unbound during this time so that I could think

13:17

about these kind of things with an amazing partner like

13:20

Lance and with a school that is student centered and

13:23

student focused. That's where I'm at with it.

13:25

With students right now, I guess I'll throw in at the end. My class policy right now is actually a little bit

13:29

more broad than the college's

13:32

policy I actually say it's fine if you want to use

13:35

it. Just tell me that you used it and describe how

13:38

you used it. I just think it's way too early right now to

13:41

be punishing students for it.

13:44

You mentioned something about a one credit course, and I

13:47

would love to hear more about that.

13:49

Yeah, absolutely. To just build off of Autumm's point about her policy,

13:54

like we created a policy that we put out as

13:56

this is our temporary policy, folks, individual folks are welcome

14:00

to adjust as makes sense for their classrooms.

14:03

And I think that was, again, like we want to be both student centered.

14:05

We want to empower faculty to make the right decisions

14:09

on behalf of their students. That was another piece of this as I went into

14:14

the winter break in conversations me and Autumm had, I just

14:18

had this brain blast. That's Jimmy Neutron reference for folks that are interested.

14:22

I just had this brain blast of like quintessential way I

14:26

could help figure out this challenge at College Unbound

14:30

would be to do it in a way that was

14:32

student centered. And so like literally got out, I had this

14:37

idea, I got on my phone, I texted the provost

14:39

and I was just like, What about this for an idea?

14:42

What if we do a one credit class that is

14:45

filled with students who are going to play with, learn

14:49

about and really think about ChatGPT and other AI

14:53

generative tools. And through that class we can create a recommended set

14:58

of policies for institutional usage.

15:01

Instantly got back, thumbs up, let's do this. Which also meant that oh I have to figure

15:06

out this course. So that was my winter break.

15:08

And then I realized there was another iteration or rather

15:11

again, my partner in conversation came up with this really great insight of what

15:15

if you could also connect it to writing courses. And so we're doing this one credit course in Session,

15:21

One of 17 week semesters and we

15:24

do eight week courses in session one, and eight week

15:27

courses in session two, and then sometimes sixteen week courses.

15:30

So in Session one, I'm doing a one credit course

15:32

where we're going to develop a rough draft of policies

15:35

around usage for faculty and students. In session two, I'm

15:39

going to try to connect with students that are taking

15:42

our writing course and have them sign up for this

15:45

class, and it'll really be an opportunity to kick the

15:47

tires on the policy. So they'll be taking a writing class.

15:50

they'll be using this policy to inform how they're going

15:53

to use the AI generative tools.

15:57

And that will be a bit of like really trying

15:59

to figure out like, what are the holes in it, what

16:01

are the ways that it really works. And in conversation with the faculty teaching those courses as

16:05

well, so that by the end of the semester we'll

16:07

have had students had a central role in developing it

16:10

and testing it and putting forward the recommendations for it.

16:14

That's where we are with it, we're about three and

16:16

a half weeks into the one credit course and it's

16:20

just been this there's about eight students in it and

16:22

it's been this rich conversation around like them getting to use

16:26

it and them getting to like really start to see

16:29

the answers it comes back with. And then also them delving into other content, other things

16:33

that are helping inform their opinions. The week by week things change because in the second

16:38

week we got the Time magazine report about how in

16:41

order to do content moderation of all of the horrible

16:45

stuff on the Internet that they scrawled in order to

16:47

like, make this open, AI, was paying Kenyan workers $2

16:51

an hour for content moderation, which is just another way

16:54

of saying they paid Kenyan workers $2 an hour to

16:56

be traumatized by like the worst of the Internet.

16:59

So every week there's these new things that help to

17:02

flesh out our thinking about it and conversations that we

17:05

have. They hear things like this and they're like, this needs

17:08

to be on like the UN's agenda. This is not correct.

17:10

Like the ways to get like really invested and start

17:14

to challenge their considerations. One of the earliest points that was just great

17:18

was I had students read some of Autumm's work and

17:23

raising some of those questions around what does it mean

17:26

to sign up and get an account with Open AI

17:29

where it asks for your name, it asks for your email and your cell number?

17:33

And we got into a discussion around like digital redlining.

17:36

And our students are predominantly students of color.

17:39

And so this generated part of when I created the

17:43

course or part of when I started to create the

17:46

assignments and the goal was for them to use or

17:49

engage with these tools, I recognized they would have

17:53

to create accounts or they would have to have access.

17:55

And so I've offered my credentials for them to log

17:58

in and to use. And as a result of that conversation I had, I

18:02

probably at this point, I've had half the students ask

18:04

to use the credentials to use it so that they

18:08

don't have to give up their own personal information to

18:11

an entity that has 10 billion dollars invested by Microsoft and

18:14

is like gathering up all sorts of data on the

18:17

users. So yeah, that that is all where we are now

18:21

is we're moving into the point in the course where

18:24

like, besides playing with it, we're really thinking about what

18:26

would we recommend for usage. That's the next discussion we're starting to have.

18:31

Getting back to your question, Rebecca, in terms of how

18:33

we're using it with students, I have been pretty vocal

18:37

and I've written a couple of blog posts that you

18:40

can link in your show notes, really being critical of

18:44

the idea of using it with students. I'm really hoping that those faculty who do teach in

18:52

a discipline where it makes sense to use it, take

18:56

a pause, take a beat, and think critically about how

19:00

they're going to ask students to use it.

19:02

And I suggested some techniques that I could employ to

19:08

make it so that they weren't forcing their students to

19:11

use it. They weren't forcing their students to sign up for accounts

19:14

at the bare minimum. But I guess I just want to say that I

19:18

do recognize that. I think that's discipline specific and especially, of course, I

19:23

just love the idea of a course that is specifically

19:26

designed to gather student voice and get student input about

19:32

university policy, about college policy.

19:35

I can totally see it. That's a situation where, yes,

19:38

the students should be informed to the point where

19:41

they actually have experience with the tool so that they

19:44

can give informed input.

19:46

But I love the fact, Lance, that you created like

19:50

a shared account. So that way nobody's putting your personal information at the

19:56

account level, but also it muddles up and it creates

19:59

noise in terms of the questions that they're asking, right?

20:03

Because it's not just the creation of that account, but also the inputs that you're putting into it.

20:07

And so by having everybody share one account.

20:10

I think that does a Greater good in terms of protecting students.

20:15

That's your influence at hand.

20:17

Makes my day. I thank you for that.

20:20

My next question is a little bit about what do

20:22

we do for instructors? Right.

20:24

What advice are we giving instructors?

20:27

How can they? How should they? How should they not?

20:31

What do we tell instructors about this new tool?

20:34

I personally don't think there's anything wrong with waiting before

20:39

you use it. So I guess there is. What do we tell instructors?

20:42

It depends on what the instructors coming to us with

20:44

needs for, right? So if they're coming to me and they're saying I'm

20:48

worried about cheating, that's a different conversation than I'm intrigued

20:52

and I want to use it. Right? So if it's I'm intrigued and I want to use

20:57

it, my first response might be, Do you really need

21:00

to use it? Do you really need to use it right now?

21:02

What are you doing with it? What are you teaching?

21:04

Is it directly related to what you're teaching?

21:07

But is there a way that you could use it and demo it for the students rather than making the

21:11

students have accounts? If you do want the students to have an account, could

21:15

you have a shared account or could you talk to

21:17

the students about their understanding of privacy and digital literacy?

21:23

Like I would say, if you're teaching like a digital

21:25

literacy, digital citizenship course, where are your students at?

21:30

Is this like a level two, three, four kind

21:33

of course, if this is intro, they might not have

21:35

a good understanding. Most people don't have a good understanding of digital privacy.

21:41

I just think that before you dive into using these

21:43

tools, you should have a good foundation of data sharing

21:47

and data collection and you should have examined some cases

21:51

of where things have gone wrong, data breaches and things

21:54

like that. You should have an idea of what kind of things

21:57

could go wrong. If they're coming to me and they're asking me about

22:00

cheating because they're worried about cheating, I usually try to

22:04

do some damage control around helping them to move away

22:09

from punitive approaches because I don't think they really do

22:12

any good. At the end of the day, I think they just

22:15

degrade our students' trust in us and our students' trust

22:19

in higher education. And I guess I try to talk to the instructor

22:23

and remind them how much is really built on that

22:26

trust, how much of education comes from that place, and

22:31

help them to realize that they're sacrificing so much

22:36

more. If they take a punitive approach, they're sacrificing so much

22:40

more than they would be if they took a more

22:43

open approach, trying to understand where the students are coming

22:46

from and trying to figure out how this really aligns

22:50

with their outcomes and the things that they're trying to

22:53

do in their course. So I usually bring it back to some type of

22:58

helping them to articulate some type of policy for their

23:01

syllabus, because it all comes back to expectations, right?

23:04

It all comes back to them really thinking about in

23:08

their heart and what's going on with them, their expectations

23:11

for the course, the affordances, the tools like ChatGPT and

23:17

DALL-E and all of these other generative tools can afford

23:21

students and helping them to articulate to students why it's

23:25

important and what they learn in the process.

23:27

In terms of ChatGPT, it's not a matter of we

23:32

want to create essays. That's not the point.

23:35

If that's the point, we're doing it wrong.

23:38

Right? The point is thinking through and being nuanced in your

23:42

thinking. So nuanced in your thinking that you're putting it down

23:45

on paper or on a screen and you're scrutinizing and

23:50

evaluating every word, every paragraph, every sentence to make sure

23:53

that things fit together and flow together.

23:56

It's that process of writing. It's not the product of the essay and helping faculty

24:01

to have that conversation with their students and helping faculty

24:04

to articulate that meaning to their students in a way

24:07

that helps the student to understand, I think is so

24:09

much more powerful than wagging your finger.

24:11

And if you do this, we're going to punish you.

24:14

I'm going to turn it over to Lance. Let him talk about it a little bit.

24:17

100%. Everything Autumm just said.

24:20

I think the only areas I would add is I

24:24

think there's some great value in them using it to

24:27

enhance some of their own work. And I think that's great with the caveats of the

24:31

potential concerns around privacy that we've already mentioned.

24:33

I think if they are going to do that, I

24:36

think it is important for them to also be citing

24:41

and identifying where their course, where their thinking is influenced.

24:44

And I come from this having worked with faculty

24:47

who have done the things that they don't like, that

24:49

they don't like their students to do. And so just really demonstrating transparency again in how they're

24:55

using it with their students. The other thing that comes to mind in using it

25:00

is and this is something, again, from conversations with

25:03

Autumm, is really emphasizing, no matter what they think of

25:08

it right now, I'm going to, going to

25:10

take the quote from you Autumm. This is the, quote unquote, dumbest that AI is

25:15

going to be. So I see a lot of dismissal and I see

25:18

a lot of it's fine, like it's fine. I'll catch them anyways.

25:21

And there's a whole other discourse around that approach or

25:24

around those concerns. But I think undermining it or not recognizing like now

25:31

with this research preview, it is getting better because we're

25:34

all training it to be better. And I think within that is really highlighting what I'm

25:39

starting to see. And I want to say, is Anna Mills

25:43

and a couple other folks have, Maha Bali I think,

25:46

has done this as well where they they're sharing

25:48

their dialogs and you're seeing through the questions in the

25:52

further development in iterations of their questions like the actual

25:56

dialog, you get some really interesting cool things and I

26:00

think that's a thing that is powerful and interesting and

26:03

valuable for for faculty and for their students to be

26:07

thinking about. I'm influenced by Warren Berger, who's written a couple of

26:11

different books about questions, and I think this is one

26:13

of those opportunities for us to really think about the

26:15

power of questioning and what you need to ask good

26:18

questions. And so there's, there's something within this that I think is a possibility for

26:24

no matter the discipline to really think about how like,

26:27

how do we ask good questions, how do we ask

26:29

meaningful questions and how do we refine questions as a

26:32

means of seeking knowledge. But in order to do that, we also have to

26:35

demonstrate some understanding in order to ask those deeper questions.

26:39

And I think there's a really rich opportunity there to

26:42

explore within all of this as well, from both faculty

26:45

and student side.

26:46

Yeah, that was part of what I was thinking with

26:50

instructional design and with my students is it is even

26:54

in order to use this effectively as a tool, you

26:57

need to know what questions to ask. And it's the same thing when you're doing analysis in

27:02

instructional design. When you start, you need to know what the important

27:05

questions are, right? Is this a training problem?

27:08

Who are the students? What are the characteristics?

27:11

Like all of these different analyses things are the questions

27:14

you need to ask, and if you don't know what those questions are, the tools are not useful.

27:19

And so I think there is an inherent skill in

27:22

learning how to ask the tool the right questions.

27:26

I think there is. And it's also different even though it's very, very

27:32

smooth. Right. And it sounds very human.

27:35

There is a big difference between asking a question of

27:38

a human and asking a question of a large language

27:41

model like ChatGPT.

27:44

If anybody is more interested in this, I was blown

27:47

away with the prompt engineering stuff that's out there, which

27:51

is all about how to ask it questions and understanding

27:54

the different ways that you can ask it questions, which

27:57

can be. This was really interesting to think about how you interface

28:01

with it and how it's different than maybe interfacing in

28:04

a human conversation.

28:06

This is probably of no value, but this whole conversation,

28:09

just like it just had me flashback to Hitchhiker's Guide

28:13

to the Galaxy when they're waiting for the answer to

28:17

life, the universe and everything. Like, I feel like there's like there's something there for

28:22

this conversation as well. It's the same dynamic.

28:27

Looking for all the right answers are not necessarily the

28:29

right questions or realizing what the machine is really built

28:32

for.

28:33

Lance you mentioned something interesting that I think is caused

28:36

an interesting conversation on Twitter as well, and that's around

28:39

citing. And I actually brought it up a little bit in

28:44

a from a blog post that Maha Bali had put

28:47

together, challenging not to anthropomorphize a computer system, but that's

28:53

worth citing. Are we not just doing that?

28:56

How does that make sense?

28:58

I guess I would say I don't know that

29:01

the citing itself is what anthropomorphizes the tool.

29:06

I think there's lots of other ways we're doing that.

29:09

And so citing might feel like one part of that

29:12

conglomeration, I would say citing it in the traditional sense

29:16

of citing, it feels like the fix for now as

29:20

we're still trying to figure things out.

29:23

I think all of it hearkens to trying to make

29:25

explicit where information comes from.

29:29

I think citing is a good start.

29:32

I think if we were to lessen or try to

29:36

detract that or deanthropomorphize it, it makes me think

29:41

about, what is it... Phipps, the article again, it is something

29:45

Autumm has shared with me in our conversations.

29:47

Phipps and Lanclos citation approach, which is like you

29:52

identify that you were using this, and that you understand

29:56

the repercussions of using a tool that works, that is

29:59

in part built off of like the illegal copying or

30:04

use of copyrighted works and also the various exploitated labor.

30:09

I feel like that's a way of threading.

30:11

It's like you're citing it. You're citing where this information came from.

30:15

And because you cannot tie it to individuals, you also

30:18

recognize that tool is an exploitative tool of sorts.

30:23

I think that's I think in my head that's one

30:25

way I've been thinking about or would think about in

30:28

this context. They offer that citation approach, I think slightly, and just

30:33

to be provocative and it's something like, nope, I think

30:36

that's what I will be using and will be encouraging

30:38

others to use. We can't just hide behind a citation because I think

30:43

maybe that's what it is. It doesn't anthropomorphize necessarily, but it hides.

30:51

It hides what really goes into that answer, both technically

30:57

and human cost.

31:00

Yeah. So I do have a blog post pulled up and

31:02

I can read the citation example that Phipps and Lanclos

31:07

propose. So this is what they're proposing as a

31:12

potential citation. And they say we offer the following text not because

31:16

we think it, but the relevant people will actually use

31:19

it, but because we think that they should.

31:21

And so it's this presentation paper work was prepared using

31:26

ChatGPT, an AI chat bot.

31:29

We acknowledge that ChatGPT does not respect the individual

31:33

rights of authors and artists and ignores concerns over copyright

31:37

and intellectual property in the training of the system.

31:41

Additionally, we acknowledge that the system was trained in part

31:45

through the exploitation of precarious workers in the global South.

31:49

In this work I specifically use ChatGPT to...

31:53

then they have some ellipses where you would fill in

31:55

the way that you actually used the work.

31:58

It's powerful. It's really powerful.

32:01

It really makes you stop and think, Should I actually

32:04

use it? If I'm going to acknowledge all of these horrible things

32:07

about it? I'm not sure that people would.

32:10

Lance is saying, Lance is embracing it.

32:12

And I think that's amazing, right?

32:14

But I think.

32:15

It also means I'm probably an irrelevant person rather than

32:18

a relevant person.

32:21

Oh! I don't know if you are, though.

32:24

I don't know if you are, though. I think what they mean by that, people who would

32:28

use it. And so if

32:31

you're willing to use something like that, I think that

32:34

definitely says something. I proposed in talking with them about the article, so

32:38

there is probably I need to put this disclaimer out

32:40

there. And if you go to the article, if you go to the blog post, you'll see at the top they

32:44

acknowledge that I had some input on this, just I

32:47

didn't write a word blog post, none of it.

32:51

But we had some conversations. We had some conversations about citation and about

32:56

different approaches to citation. And one of the things I put forward is you

33:01

could use this as activism if you really felt very

33:05

strongly the ethics of these systems and you wanted to

33:09

make a point, you could add a citation like this

33:11

to your presentation or your paper or whatever, and then

33:16

just have a use of ChatGPT that is so minimal,

33:22

right? Like I used it like a thesaurus and I changed

33:25

out this one word. So I used it just so that you could use

33:30

the citation to point out the abuses that can happen

33:33

through it. It's really powerful. It could be used in lots of different kind of

33:38

ways. It's also so powerful because it's using the very thing

33:42

that ChatGPT obscures. So ChatGPT obscures citation, it fabricates citation.

33:47

We don't know what is inside of these language models.

33:50

More than likely, there is copyrighted work in there.

33:54

We don't know that for sure. It's just we really don't know what's inside of the

33:57

language that it's trained on. And so this in a way, weaponizes the idea of

34:02

citation to really speak back to and bring some light

34:07

and bring some acknowledgment to some of the darker things

34:10

around this, around this particular tool.

34:15

I think that's fascinating. I like the power in that citation, that opening citation.

34:22

One of the ways that I've been using it, which

34:25

is totally not using it as a data lookup, I

34:30

have text that's in past tense that I need to

34:33

move to present tense and it's making it a whole

34:36

lot faster for me to write it because I can

34:39

take the stuff that I wrote in past tense, plug

34:42

it in and say, give me this in present tense,

34:44

and then I can work with the present tense version

34:47

that it gave me. And because I wrote the past tense, I own the

34:51

present tense. It's just saving me time.

34:54

So again, I'm using it strictly as a tool.

34:57

Now I'm questioning how it is now taking that data.

35:02

And I found it did do that to me because

35:04

I had asked it a question about a town that

35:07

we visited called Harrington Harbor.

35:10

And originally it said that this place didn't exist and

35:14

there was no knowledge of it. And then later, when I asked it about Harrington Harbor,

35:18

it came back and quoted what I had written about

35:21

it. Right. So it's like that.

35:24

Oh, now your database thinks Harrington Harbor is the information

35:29

that I just fed you about Harrington Harbor, which actually

35:33

shows you where it's getting the information from and it

35:38

isn't even fact checking it.

35:41

You can't argue like Wikipedia has systems in place where

35:45

things can be at least somewhat quality controlled in some

35:49

way, or at least say whether it's been quality controlled

35:52

or not. And fair because ChatGPT is a black box

35:57

there's no looking inside to see where the information came

36:01

from. I think that's fascinating.

36:03

Just on going back to the point that it's changing

36:08

all of the time, right?

36:11

And like it seems like every single day there's a

36:13

new thing coming out, speaking to what you're talking about

36:16

Rebecca Amazon is now warning their employees to stop using

36:23

it because they're seeing some of their, like trade secrets

36:26

in terms of coding, showing up in answers from ChatGPT

36:32

because they're using ChatGPT to help them code.

36:36

And so the ChatGPT is not keeping those secrets and

36:41

is then absorbing that and then sharing that with other

36:45

users who are out there. It just put a link in the chat that you

36:50

can maybe link in the show notes, but yeah, yeah,

36:54

because it's a learning machine.

36:56

We've been talking about learning technologies and it's coming back

36:59

to ID stuff and instructional design and technologies.

37:02

For years we've been talking about learning technologies.

37:05

We haven't been talking so much about technologies that learn,

37:07

which is a little bit different.

37:10

This makes me so happy in some ways. I'm not going to like there's a lot there that

37:15

there is some Schadenfreude going on right now.

37:18

I'm not going to lie to the point about the

37:21

responses. This is it.

37:24

If it gets edited out, I completely understand.

37:26

But I dropped this line this week when we were having

37:30

a discussion about it. This just came to my head like, like fluid

37:34

is that ChatCPT responds with all of the

37:40

self-confidence of a mediocre white dude.

37:43

It just does and I just ever since that came

37:47

to mind, that's all. Okay. I can only hear ChatGPT responses and the voice

37:52

of somebody named Brad.

37:57

Yeah.

37:58

Sorry to all of the Bradleys out there, right?

38:00

That's right. That's right. Disclaimer...

38:02

I think it's Chad, actually.

38:03

I'm sorry.

38:03

ChadGPT.

38:05

Because we were talking about it in the context

38:08

of its errors and things like that.

38:10

And that came to me.

38:13

And it just it sticks with me now. And that's all I can think about now is.

38:17

I love it.

38:21

I'll ask one more sort of question in this area

38:24

and then I think we'll close off. And that is we've talked about guidance for students and

38:30

guidance for instructors. What about instructional designers?

38:33

What do you think instructional designers should be doing with

38:36

this technology?

38:37

Well, the first blog post that I wrote on this

38:40

topic was actually me wondering if it could be an

38:43

instructional designer. There's going to be labor implications of this technology, I

38:48

think everybody. Everybody is clear on that, right?

38:52

OpenAI is doing research right now to try to better

38:57

understand the labor implications because their mission, they say, is

39:01

to make sure that AI is basically a net positive

39:05

in the world rather than a negative in the world.

39:08

And so they recognize and realize this is going to

39:10

have labor implications and trying to foresee and understand that

39:14

is really important if we're going to be responsible or

39:16

something like this. So I was just curious.

39:18

I was just curious if it could, like, do my

39:21

job. So I went in and pretended that I was a

39:25

physics professor, which I know nothing about physics.

39:28

I couldn't ask any discipline specific kind of questions, but

39:31

I just asked really general questions about my department chair

39:35

wants me to take my learning online and I'm skeptical

39:38

about it. I don't know if this is really for me.

39:41

And just tried to see what it was, what kind

39:45

of answers that it would give me. So going back to the idea of prompt engineering, one

39:49

of the fun things you can do is role play with it. You can ask it to role play with you.

39:53

And so I started off by saying, You are an instructional designer and I'm an instructor and you, if anything,

40:00

I think that's an interesting approach because I do think

40:03

that there is a lot of very standard answers in

40:07

instructional design. There's a lot of best practice that it's fine, I

40:13

guess, but I've always pushed back against it because I

40:16

think I'm not convinced that it is best practice.

40:19

I don't know who really makes that decision.

40:21

There's just a lot of it out there. And so I think it's kind of interesting just to

40:26

see what kind of vanilla answers it gives you to

40:29

instructional design problems and then say, okay, well can we

40:34

be more creative than that? Can we go beyond this knowing that these are maybe

40:40

some of the most generic answers that are out there?

40:44

So for ID folks, I mean, I think it's one

40:48

of those like really having to spend some time playing

40:53

with it, whether it's ChatGPT or really any of

40:56

these tools, both to understand their limitations and possibilities and

41:01

evolution and to be prepared for the questions that they

41:05

will get from faculty where I see it impacting potentially

41:09

ID the most is probably in a lot of the

41:14

OPMs. You're thinking about OPMs.

41:16

OPMs I can see the large scale online institutions.

41:22

Your SNUs, your ASUs, your Western Governor's looking to leverage

41:27

this in a way that reduces the amount of instructional

41:32

design folks that they rely upon or that they contract

41:36

out and using this as more systematic means of updating

41:40

their courses. I can see the pathways to that because in some

41:44

ways it feels very much like textbooks, textbook publishers in

41:50

their methods. We've got to sell more books, and so we've got

41:52

to come up with a new edition every two years

41:55

and we'll just switch the chapters around. There's some ways I can see some of the larger

41:59

scale ones being like, We've got to refresh our courses.

42:01

So hey, ChatGPT update it with this kind of flavor.

42:07

Like I could... I can see a series of APIs and plug-ins

42:11

being used to like do a lot of that.

42:14

I think that's and so I would say anybody that

42:17

is in that space just understanding it and seeing where

42:21

that starts to pop out because I just have trouble

42:25

not believing the ways in which those larger entities work

42:29

on this like assembly line.

42:31

And efficiency is always the better answer than anything else

42:36

that they aren't going to start thinking about it and

42:39

using it.

42:39

I think another big role for instructional designers right now

42:42

too, and I say this on January 30th of 2023

42:48

because it's changing all the time, like tomorrow, like it's

42:53

going to be different. I think your administrators need guidance on top of your

42:57

instructors. Your instructors are going to be coming to you and

43:00

asking you for help with this. Just all the different roles of the instructional designer.

43:05

There's the role of you as a content creator and

43:08

as a designer who's creating stuff for people.

43:10

There's the role of you as somebody who gives advice

43:12

to faculty, but then most of us have either interactions

43:18

with some type of administration or we have a director

43:21

who has direct access to administrators.

43:24

And I think all of us are very busy.

43:27

We have a lot going on, right.

43:29

And we don't have necessarily all the time in the

43:32

world just to concentrate on generative AI.

43:36

So I think they're going to be coming and asking

43:38

for advice and asking for guidance in terms of what

43:43

to do with all of the social change that's going

43:46

to be happening around what the impact of these tools

43:49

is going to mean, so just educating yourself and keeping on top

43:53

of that stuff that's changing, even if you can't stay

43:55

on top of it for every single day because it

43:57

is changing every single day. If you can stay on top of it once a

44:01

week, once every two weeks, whatever, to kind of keep

44:04

your eye on what is happening and how it's changing,

44:07

I think that it's going to be really important for

44:10

those who will be coming to you and asking you

44:12

for advice, which will be faculty, but more than likely

44:15

it will be administrators, too.

44:16

And I think there's an interesting point about whether or

44:19

not it could replace instructional design.

44:23

I don't think it can, but I do think it

44:26

does some things very interestingly, and that's I've been asking

44:29

it all of the questions, like I've generated some chapters

44:32

from my textbooks, but I'm using it saying, okay, here

44:37

are the questions I want this chapter to answer.

44:40

What do you say? And then it gives me an answer. And I'm like, actually pretty good.

44:43

I'll take that.

44:45

Hey, I will say, and should I say this, and

44:49

let's say this, just say it.

44:51

I will say I wrote that first blog post trying

44:55

to figure out if it would, if it could replace

44:58

me with a little bit of a little bit of

45:01

tongue in cheek, little bit of haha thing.

45:03

But then I realized the larger discourse of people who

45:07

really are looking at the potential impacts are not thinking

45:12

about that. They are thinking about whether it could replace teachers.

45:17

So instructional designers, without teachers, that's a really scary prospect

45:23

if you ask me. And that's the bigger thing to be concerned with.

45:28

Yeah, 100%.

45:31

Do you think that's where it's going? I think the thing I see is and this is

45:36

the it's the thing that with whether it's AI or

45:40

it's robots, that that seems to be sometimes missed or

45:44

under considered is like I hear folks being like there's

45:48

some jobs, that robots or AI won't ever replace that's not

45:52

they'll replace them en masse.

45:54

What it will mean is that you as an

45:57

individual instructional designer or plumber or whatever your profession, you

46:03

will be able to do a lot more, more quickly.

46:08

And so if previously you were only able to work

46:11

with ten faculty, you can now work with 20 or

46:14

you can work with 100. That's the thing that I'm seeing.

46:18

So it's not that you replace all instructional designers, but

46:21

it means one like the expectation of one instructional designer

46:26

will be much more expansive or be much more expected

46:29

than what it was previously.

46:32

And like, we see that within our technologies, every tech

46:35

like we have over the last 50 years, almost all

46:38

workers have become increasingly way more productive.

46:42

But we keep asking more and more of them and

46:44

of course, paying less and less of them. And so I think that's the thing, is what this

46:49

shows, or what this creates the opportunity to, is like now instead of and I've

46:54

seen and was playing around with this in December as

46:56

well when I was doing some advising, I was like,

46:59

What if I use the machine to help me get started? So rather than spending 5 hours of like hemming and

47:04

hawing, I used it as a like, let's see what

47:06

I can get here and start running more quickly.

47:08

And I think that is where the challenge is.

47:12

It's not going to be replacing all. It's just going to make the need for this fewer and

47:16

fewer. And I think when we talk about replacing teachers, like

47:19

to me in many ways, the big scale instructional

47:23

places that have 130,000 students, they pretty much replace teachers

47:28

there. Right. If you teach at those schools, you're not doing curriculum.

47:32

You are grading and you are doing discussions.

47:35

And I think they're doing that because they need to

47:37

at least justify that you actually have an interaction with

47:40

a human. Well, yeah, they will replace those, but meanwhile they can

47:43

also much of their staff is not faculty, but is

47:46

actual staff like instructional designers.

47:48

And now it's like, Oh, we can maybe even do

47:51

less. And so I think I don't want to be like,

47:53

Oh, we're all doomed because of robots and AI

47:55

But I think that is the threat that they pose.

48:00

And I think in industries we can start to see

48:02

that happening.

48:04

That's right. And yeah, I don't know if it I also don't

48:10

want to jump on the robots are coming for our

48:13

jobs pedestal like we know that doesn't end well.

48:17

And it's never 100% true. But what you just talked about, Lance, is job loss,

48:22

right? It's job shifting. If we have 50 instructional designers right now and each

48:28

one of them are working with ten instructors, if each

48:32

one of them can work with 50 instructors.

48:34

Yeah, that's a different situation.

48:38

It will be interesting to see how it plays out.

48:41

I guess the big takeaway is nobody really knows.

48:44

But one thing that we do know is that there will be some kind of labor implications.

48:49

There could also be job creation around this as well.

48:52

Right. Knowing people who know prompt engineering or who know how

48:55

to train a language model, those kind of things could

48:58

end up being something that are marketable skills that we

49:01

might need going forward. So I don't want to paint it as all gloom

49:05

and doom, but there will probably be upset and there

49:08

will probably be some reskilling that is needed and there

49:11

will probably be big changes.

49:13

I want to say thank you to both of you

49:16

for coming in and joining this little thought experiment around

49:20

ChatGPT and what it might be doing this semester.

49:24

I am really curious to see how all of this

49:26

is going to play out and where we're going to

49:30

be at come September, right when we go into another

49:35

semester after having had the summer as well and a

49:39

little bit more time for things to play out.

49:42

But thank you very much for your willingness to come

49:44

on and chat with me about Chat.

49:47

Thanks so much for having us, Rebecca.

49:49

Thank you.

49:51

You've been listening to Demystifying Instructional Design, a podcast where

49:54

I interview instructional designers about what they do.

49:57

I'm Rebecca Hogue, your podcast host. Show notes are posted

50:01

as a blog post on Demystifying Instructional Design dot com.

50:05

If you enjoyed this podcast, please subscribe or leave a

50:07

comment in the show notes Blog post.