0:00

All day, every day, we make

0:03

decisions. Some are so

0:05

small we barely think about them. Such

0:07

as what to have for breakfast, or which

0:10

route will avoid the most traffic on the way to

0:12

work. Others are more consequential

0:15

whether to take a new job or

0:17

how to spend or save our money.

0:19

And some decisions can have life or death

0:21

consequences. Such as weather to get

0:23

a vaccine or evacuate ahead

0:25

of a hurricane. Over the past

0:27

several decades, psychologists and other

0:30

behavioral science researchers have become

0:32

increasingly interested in understanding how

0:35

people make decisions like these. Why

0:37

we so often make bad decisions? And

0:39

how even seemingly small changes

0:41

in the way that choices are explained and

0:43

presented can make a big difference

0:45

in the decisions that people make. So

0:48

what have researchers learned about decision

0:51

making? Why do people

0:53

make bad decisions? Do bad

0:55

decisions happen when people don't have enough

0:57

information or when they're overloaded

0:59

with too much? How do behavioral

1:02

scientists define a bad decision

1:04

anyway. And how can decision

1:06

researchers findings best be deployed

1:08

in the real world to make a positive

1:10

difference in people's lives? Welcome

1:15

to speaking of Psychology. The flagship

1:17

podcast of the American Educational Association.

1:20

That examines the links between psychological science

1:22

and everyday life. I'm Kim Mills.

1:27

have two guests today. First is doctor

1:30

Lace Padilla, an assistant professor

1:32

of cognitive and information sciences at

1:34

the University of California Merset. Doctor

1:36

Padilla studies how people use data visualizations

1:39

to make real world decisions with

1:41

life or death consequences in

1:43

areas including hurricane evacuations

1:46

and vaccine uptake. Her

1:48

lab's mission is to help people make the best

1:50

possible judgments about their health and safety

1:52

by developing and testing new ways to

1:54

visualize and communicate complex

1:57

data. Our second guest today

1:59

is doctor Perfecto, Doctor

2:01

Perfecto is an assistant professor of marketing

2:03

at the Olin School of Business at Washington

2:06

University in St. Louis where she studies

2:08

how consumers make decisions. Much

2:10

of her work also focuses on improving

2:12

research methods, including designing

2:15

decision making studies that are

2:17

more likely to replicate in

2:19

real world settings. Thank

2:21

you both for joining me. Thank you for having

2:23

us. Thanks so much. So I just mentioned

2:25

about how behavioral science research can

2:27

help people make better decisions, but

2:29

the term better can be

2:31

a subjective term. What makes

2:33

something a bad

2:34

decision? How do you define good versus

2:36

bad in your research? Dr.

2:38

Padilla, let's start with

2:40

you. I think different groups of people

2:42

define good and bad decisions differently.

2:45

And, you know, there's some people

2:47

who might think that

2:49

a good decision would be the best

2:51

computational decision. If we had

2:53

all of the information available, we could

2:55

you know, calculate the most optimal decision.

2:58

But the truth is, is that humans don't function

3:00

like computers. And from

3:02

my standpoint and my work, I

3:04

think of a good decision as

3:07

someone making the, you know, using

3:09

all the information available to them to the

3:11

best of their ability. And

3:14

in contrast, we can compare that to a

3:16

bad decision. And those are a little bit easier

3:18

to identify. And I often see bad

3:20

decisions as clearly someone

3:22

misinterpreting or misunderstanding

3:25

information that was presented to them that

3:27

would ultimately lead to some type

3:29

of error in their reasoning. You

3:31

too come to the topic of decision making from

3:33

very different angles. So let's talk a

3:35

little bit about each of your research backgrounds.

3:39

So Perfecto, in your research,

3:41

what's the main thing you're trying to understand about

3:43

how people make

3:44

decisions? That is a great question.

3:46

So I would say the work that

3:48

I do tackles this

3:50

issue from a broader lens

3:52

for better or for worse than than

3:55

Dr. Pidez. I

3:58

spend a lot of my time making

4:00

sure that the

4:03

work that we're all doing in this

4:05

field of decision making research

4:08

is able to actually have

4:11

a useful impact out in the world. So

4:13

the work that we're doing, we're not learning

4:16

about how people make decisions just for the

4:18

fun of it, the hope, is that in the

4:20

end, we're actually able to

4:23

improve people's lives down the road.

4:25

One area that I'm focusing

4:27

on right now and some ongoing

4:29

work is realizing

4:32

that a lot of the ways that we've previously

4:35

studied decision making is

4:37

pretty abstract. So

4:40

we're using very hypothetical

4:44

design, sometimes a little bit strange,

4:46

a little bit artificial. And

4:48

that's done so that we can have a

4:50

really clear understanding of

4:52

what is contributing to somebody's

4:55

decision. But out

4:57

in the world, the world is not clean, the

4:59

world is not so tightly

5:01

controlled. And so something

5:03

that I'm looking at is that

5:05

in designing those studies, we've used

5:07

a lot of positive stimuli.

5:09

So we've asked people when we're trying to learn about

5:11

risk preferences are dealing

5:13

with uncertainty. We say, do

5:15

you want to have this

5:18

gamble with only upsides or

5:20

this gamble with smaller upsides.

5:23

And so we can certainly see if someone's

5:25

willing to go for a

5:27

bigger, riskier outcome versus play it

5:29

safe with a smaller positive

5:31

outcome. But both are still great. You

5:33

still get something. You still have a chance

5:35

at something. In the end. And

5:38

I have found my work and others

5:40

as well that those types of more

5:42

positive decisions we're more likely

5:44

to make errors in. When things are

5:46

going well and things are feeling great, you're

5:48

more likely to just go with

5:50

your gut. And I'd go with

5:52

that initial response. Sometimes

5:54

that's fine, but a lot of times

5:56

that then leads us to as Dr.

5:58

Bidea said misinterpret the information

6:00

we're dealing with makes errors

6:02

in evaluating those outcomes

6:05

and and lead us to an outcome

6:07

that is suboptimal for our situation.

6:10

So I'm sort of assessing the

6:12

magnitude of that issue going

6:14

back looking at some things we thought were super

6:16

duper robust and really

6:18

big, different really big effects

6:21

that may not actually be we're robust down the

6:23

world so that when we're actually trying to implement

6:26

our findings, we

6:28

will see results. Howard Bauchner: And Dr.

6:30

Padilla, what's main thing that

6:32

you're trying to understand how is your research different

6:34

from Dr. Perfecto?

6:36

My research is maybe different from

6:38

most people who study decision

6:40

making. Honestly, in in the fact

6:42

that I like to examine how

6:44

we can change the

6:46

information that people are presented with

6:49

to help them make better decisions. And

6:51

I specifically study data visual stations.

6:53

So I'm looking at creating

6:55

forecasts with uncertainty and visualizing

6:58

that data to try to help

7:00

people understand the risk

7:02

that they might be hunter. Said

7:05

that's a very, you know, holy

7:07

different angle that I bring to it.

7:10

So

7:10

what are some of the main reasons

7:12

that people make bad decisions?

7:15

THAT IS SUCH A GOOD QUESTION. IT IS WHAT

7:17

HUNDREDS OF PEOPLE HAVE SPENT DECADES TRYING

7:19

TO GET A GOOD GRIP ON

7:21

AND HUNDREDS MORE WILL IN THE DECADES TO

7:23

COME. I think both of us can speak

7:25

to one or two particular

7:27

reasons where people

7:29

might go wrong. I touched on one

7:31

already where folks might

7:33

go with their gut a little bit too readily. I

7:35

think that aspect of the positivity

7:38

or negativity of what you're dealing

7:40

with, people might not realize the extent

7:42

to which. That can influence

7:44

how likely you are to make

7:46

a snap decision versus sit with it

7:48

for a while. So I

7:50

would say one main group

7:52

of the problem is that people

7:54

just sometimes go too fast. You

7:57

know, feels right. They go with what

7:59

they see and then

8:01

they they end up somewhere they don't want to be

8:03

whereas if they had taken a little bit more time and

8:05

sat with it, they wouldn't have made those errors. And when

8:07

I say more time, I don't mean like

8:09

sleep on it, come back to it

8:11

in a few hours. I mean like thirty

8:13

seconds. Ten seconds. Even

8:16

that amount of time, we

8:18

readily find that folks can

8:20

and by we, I mean, not just me, but decision

8:23

making researchers more generally find that

8:25

people can recognize that

8:27

they might have misinterpreted something

8:29

or they're coming at the problem

8:31

from the wrong perspective or

8:34

even just that they made a mistake. They

8:37

and so by

8:39

just taking a

8:40

second, checking yourself, people

8:43

could be much better off.

8:45

Yeah. I I really agree with that. And I

8:47

think that that ties

8:49

nicely into the way that I think about it, which

8:51

is that, you know, throughout

8:53

human evolution, we have been

8:55

able to make some pretty fast snap

8:57

decisions that worked out fairly well for us. We

8:59

are now in a very complicated world with

9:01

lots of things that there's no reason that we

9:03

would have evolved the ability to

9:06

know, calculate long term

9:09

financial stock projections or,

9:11

you know, any of the very complicated

9:14

long term things that we have to make decisions

9:16

with. So for that reason,

9:18

our intuitions sometimes

9:20

fail us. So if we

9:22

can slow down, and try

9:24

to kind of activate a more

9:26

analytical approach that can sometimes

9:28

be very useful. And I

9:30

think in addition to

9:32

that say

9:35

by general refrain of

9:37

the way that information has

9:39

been presented to us isn't easy to

9:41

understand either. It is

9:43

hard for the average person to

9:45

look at probabilities and

9:48

forecasts and understand what they mean and make

9:50

effective decisions. And it's not

9:52

their fault. And I think that there's a

9:54

lot more that can be done to make that

9:56

information easier for all people to

9:58

understand, and a lot of people are working on

10:00

that particular problem. So I think it's a

10:02

dual issue of, you know,

10:04

we have some strategies that might not

10:06

be working for us, including making these

10:08

gut decisions. And also the

10:10

information that we're working

10:11

with. It's very complicated, more probably

10:13

more complicated than it necessarily

10:15

needs to be. But it's some of the

10:17

problem that we as a society

10:19

are not as math literate

10:21

or probability literate as

10:23

we should be. For example,

10:25

when people hear that a vaccine

10:27

is ninety percent effective, then

10:29

there are people out there who say, well, but

10:31

that's not good enough. What about the other

10:33

ten percent? And then don't get the vaccine. I

10:35

mean, what is wrong with us that we can't

10:38

understand that ninety percent is a pretty

10:40

good number.

10:41

Yeah. I'm People always

10:44

mention that as, you know, wouldn't it be

10:46

great if we all could be more mathler and

10:48

I agree? However, I do

10:50

think that that's not the only solution

10:52

because the American

10:54

population is fairly well

10:56

educated and We're not the

10:58

only group of people who need to

11:00

interpret this information. If

11:02

you can think about the people who

11:04

are most in need of decision

11:06

support, they likely have the least

11:08

amount of education. So I think in

11:10

some ways blaming

11:12

people for not being more literate

11:15

isn't fair. I think that

11:17

people want to make their best

11:19

decision and they might be in a

11:21

circumstance where they didn't have access to certain types of

11:23

education and that shouldn't mean

11:25

that they have poor

11:27

health outcomes. I think it's

11:29

our responsibility scientists and good

11:31

citizens of the world to ensure that

11:33

were not negatively impacting people simply

11:35

because they didn't have access to high

11:38

quality, higher education. Well,

11:40

let me ask, what are some of the effective

11:43

strategies for helping people to

11:45

make good decisions. And what are some of the

11:47

strategies that don't work and that might

11:49

even backfire by driving people to do

11:51

exactly the instead of what trying to

11:53

convince them to do.

11:55

And maybe this is the place where I

11:57

ask you both to explain some

11:59

terms that are used in in your

12:00

research, choice architecture, and

12:03

libertarian paternalism, for

12:05

example. Yeah. I

12:07

mean, I think choice architecture, one

12:11

could describe it. I don't know if

12:13

that Bidia sees

12:15

sees herself this way, but you could

12:17

imagine framing the

12:20

information, visualizing the information

12:22

differently, as a form of choice

12:24

architecture. So when you think of the choice architect,

12:26

it is the person who is

12:28

structuring how the

12:30

decision is phrased. So

12:32

it is not, it doesn't fall out of the sky.

12:34

Someone has to decide, are we gonna

12:37

frame the decision positively?

12:40

Like, do you want

12:42

to increase your contributions

12:44

to retirement? Or more

12:46

neutrally, what do you want to do with

12:48

the with your upcoming

12:51

elections? Do you want to increase or

12:53

decrease? Should increase come first? Should increase.

12:55

Come second, when should

12:57

you mention how much it's going to increase? All of these

12:59

decisions have to be made by somebody, that is

13:01

the choice architect. And

13:03

the way in which all of those

13:07

approaches, I don't want to keep saying

13:09

decisions, but the way in which all of these little

13:11

decisions that the person framing

13:13

the actual decision needs to make.

13:15

How that impacts what people

13:17

actually do in the end would

13:19

be would be choice architecture. Think,

13:22

libertarian paternalism comes from

13:24

a particular way of

13:28

creating of being that choice architect, of

13:30

creating those decisions. Thing. I'm

13:32

gonna I know as the person

13:34

as that's the choice

13:35

architect. I know what answer you should pick.

13:38

I don't wanna force you to

13:40

pick it. But I'm gonna make it

13:42

really easy for you to pick it and hard for you

13:44

to not pick it. And so are

13:46

are those methods then that that help people

13:48

make better

13:48

decisions? Well, that's to that. Now that

13:50

that's a thorny question. It

13:53

depends on the on the benevolence of the

13:55

the choice architect. Certainly,

13:57

AAA controversial question. These

14:00

days of who has the best

14:02

idea, what is best for people. So this

14:04

development of this term choice architecture

14:06

and libertarian paternalism is

14:09

a big celebration and getting

14:11

more people to save more for

14:13

retirement. That's great. If

14:15

you have the money to save for retirement, And

14:17

so some people who might not

14:19

be in such a good financial position,

14:21

they might be able to do better for

14:24

themselves with that money now instead

14:26

of being funneled

14:28

through defaults and

14:30

what options are preselected for you

14:32

and you need to do work to

14:34

do something else. They

14:36

might not understand, like, what

14:39

Dr. Badillo is talking about.

14:41

They might not be able to

14:43

figure out the how the decision

14:45

has already been made in some way

14:47

for them. And so whether

14:50

that's the right move or

14:53

maybe we've gone too far is is

14:55

certainly a hot topic these

14:57

days. Dr. Pete, I want to ask you

14:59

because you study some very

15:01

consequential decisions that people have to

15:03

make such as how to decide whether to

15:05

evacuate during -- when you're faced

15:07

by the third of a hurricane. And I know

15:09

you've come up with some data

15:11

visualizations that help people, but I mean,

15:13

what are weather forecasters doing

15:15

wrong that drives people to make bad

15:17

decisions when they are perhaps in

15:19

the path of of a

15:20

hurricane? This kind of ties

15:23

into what Dr. Perfecto was just

15:25

mentioning about who is the decider of

15:27

what the right decision is. And

15:29

the approach that I take is I

15:31

don't know if everyone should evacuate. That's, you

15:33

know, I'm I'm a cognitive

15:35

scientist. But I

15:38

do know if someone misinterpret

15:41

something, misunderstands information

15:43

being presented to them. So I can at least

15:45

identify when there

15:47

is a, you know, a massive

15:49

failing in how people are interpreting what

15:52

what they're presented with. So

15:54

I kind of take the standpoint of if I

15:56

can at least reduce some of

15:58

the errors, then,

16:00

you know, the the lab won't be in the process. I'm

16:03

not sure if I'm making the decision better, but I'm

16:05

certainly removing some of theirs. And we

16:07

can do that with data visualizations. And

16:10

part of what has happened historically is

16:12

people didn't appreciate the importance

16:14

of how that data visualizations

16:17

were. Displayed to people. And there's some

16:19

type of visualizations like the cone of

16:21

uncertainty, which kind of starts

16:23

at a point and grows, the width of

16:25

the cone grows, and

16:27

it is intended to show

16:29

the kind of the mean

16:31

path of the storm. And it is a

16:33

sixty six percent confidence

16:35

interval around the mean predicted

16:37

path. Well, most people don't know what a

16:39

sixty six percent confidence interval

16:41

is. And so instead they see it

16:43

as the storm growing in size over

16:45

time because they see a small point and it

16:47

gets bigger. So reasonably, they think the

16:49

storm is growing in size. That

16:52

is I

16:54

think visualization that hadn't been

16:56

tested for the last thirty

16:58

years that it's been in use. And so

17:00

no one really appreciated that it was

17:02

causing massive confusion

17:04

and misinterpretation of what the

17:06

storm was going to do. And

17:08

more recently, we've been studying these in a

17:10

careful way and studying alternatives

17:13

to those classic visualizations.

17:16

That at least don't have those same

17:19

misinterpretations. Again, I'm not trying to make

17:21

someone evacuate or not. I'm just trying to help

17:23

them understand where the storm is

17:25

gonna go. Successfully to

17:27

empower their decision making.

17:29

Let's talk for a minute about the

17:31

the dichotomy between time

17:33

perception and decision making.

17:36

So if you're making a decision

17:38

that it's what you're gonna

17:40

wear to work tomorrow, that's one thing. But if you're trying to

17:42

make a decision about how

17:44

we save the planet, you know, what what do

17:46

we do to stop global warming? People

17:49

have problem understanding something that is

17:51

so far out in the distance that they don't know

17:53

how to make any kind of decision about. What

17:55

can we do to get people to better understand

17:58

how to make those kinds of decisions effectively at

18:00

at this point in

18:01

time. You were hinting at it a

18:04

bit, Kim, of sort of making

18:06

that really far off abstract

18:09

thing much more concrete

18:11

and feeling real here.

18:13

There's some co work by

18:16

Hal Hirschfield that

18:18

talks about sort of imagining

18:20

yourself farther along.

18:22

And so he does he does cool stuff

18:24

like like aging, peoples,

18:27

faces in in photos and being like this

18:29

is you in fifty years. Now what

18:31

are your thoughts about sort of what's gonna happen

18:33

in fifty years? And you're like, oh, like me.

18:35

Like, I'm gonna be there. Like, it feels it feels more

18:38

concrete and so we think about it as feeling

18:40

sooner and

18:42

evaluating it. Like we would a

18:44

a more closer in time

18:46

event. So

18:48

taking steps like I said, like, you hinted

18:50

at taking steps to

18:53

to make that feel real now could

18:56

certainly be beneficial? So

18:58

it's personalizing it.

19:00

Yeah. So that that is one

19:03

way to do it. So certainly, you

19:05

could think of time.

19:07

So there's this there's this whole literature on construal level

19:10

theory that talks about

19:12

how things can feel very abstract

19:14

and concrete. And

19:16

that abstract versus concrete continuum

19:18

can emerge in time. Like,

19:20

things are far away in time versus

19:22

close in time. But they can also

19:25

emerge interpersonally like

19:27

things that matter to me versus things

19:29

that matter to people farther away from

19:31

me that are less related to

19:33

Concepts that are less related to me.

19:35

So psychology is very related

19:37

to me. Economics is maybe

19:39

next closer, but theoretical

19:42

physics much farther away

19:45

from me. I don't do

19:47

anything related to to theoretical

19:49

physics. And so by

19:51

by bringing it closer to me

19:53

on any of those dimensions can help

19:55

make it more more concrete. So

19:57

Bringing myself into

20:00

that situation can help.

20:02

Dr. Piediane, thoughts on how

20:04

to help people visualize things that

20:06

are so far in the

20:07

future, but help them make the right decisions

20:10

now. Yeah, I think

20:12

one of the things that data visualizations can

20:14

do is to remove some of that

20:16

obstruction. We can show

20:18

similar to what Perfecto was

20:21

indicating what the flood plan of

20:23

your home will look like in fifty years. And

20:25

you can use that information. You can just

20:27

see it to help you make decisions now

20:30

about getting flood insurance. Or

20:32

we can show you what it would look

20:34

like if we did forest restoration in

20:36

this area and how we can reduce the risk

20:38

over long term. And you can just

20:40

see it with data visualizations in

20:43

ways that you wouldn't be able to

20:45

before. So I

20:47

think making things more

20:49

concrete less abstract, especially

20:51

in these long term future projections.

20:53

I think that's really one of

20:56

the powers of data visualizations

20:58

that have just recently

21:00

been considered

21:02

as as a resource? So

21:04

with the advent of the Internet, we

21:06

now have access to huge

21:08

amounts of information in an instant,

21:11

whereas in pre internet days,

21:13

if we wanted to make a decision regarding,

21:15

say, which car to buy or what

21:17

college to attend. The research was

21:19

very labor intensive. You might have to go to

21:21

a bunch of dealerships. You talk to your

21:23

friends and neighbors. Maybe you go to the

21:25

library and look at books. Now,

21:28

you know, everything is is

21:31

immediately available on the wondering

21:33

is that making decision

21:36

making easier for

21:38

people or More difficult do

21:40

we

21:40

know? I know very

21:43

short, long time

21:46

psychologist, Swarthmore and

21:48

Berkeley, her talks about how when

21:50

we have so many options

21:53

available, the

21:55

the work on choice overload, whether,

21:57

like, having so many and I feel

22:00

I can't choose. I feel less happy

22:03

with that is

22:05

mixed. Still an

22:07

ongoing debate, something that's hard to study

22:09

in the lab that probably does

22:12

exist out in the field. But

22:14

very short is where it says, if we

22:16

if we have a lot of options, it feels like we

22:18

have a lot of potential, like, we should be

22:20

able to find the perfect thing.

22:23

We I'm in marketing, so I was couchated

22:25

products and whatnot. But if

22:28

I if I'm looking for a college, like, I

22:30

should be able to eliminate

22:32

my I have all these

22:34

forums with students I can talk

22:36

to. I have lots

22:38

of resources available at

22:40

at the click of a mouse that

22:42

to be able to learn about the school. So I should

22:45

feel really confident.

22:47

But that all

22:49

that information doesn't perfectly

22:51

solve the problem for us. And

22:53

so we might feel

22:55

less satisfied with

22:57

what we come up with in the end that I feel like I

22:59

should know exactly what I want. I

23:01

feel like I should find everything, but

23:05

I don't. And so now I'm

23:07

less less happy with what I

23:08

get. Dr. Padilla, easier or

23:11

harder. I

23:13

think more complex would be my

23:15

answer to that. I think

23:17

what happens partly

23:19

is that, you know,

23:22

we are information forwarders. We go online

23:24

and we kind of search for information. And

23:26

so the way we do that is biased. We're

23:29

kind of looking for things that affirm our own

23:31

beliefs most of the time. So

23:33

we are to have a tendency to find things

23:35

that affirm our own beliefs.

23:38

And within that, we're exposed

23:40

to lots of misinformation and being

23:42

able to identify what misinformation

23:44

is is very challenging, especially

23:46

if it your own more likely to believe it's

23:48

true if it matches

23:50

your own beliefs. So we have a

23:52

whole new set of problems that we

23:54

have no haven't,

23:56

you know, developed any skills at

23:58

combating, which makes the whole

24:00

thing very complicated. And

24:02

I think for me, it really

24:04

brings this new concept of trust to the

24:06

forefront. Now

24:08

we have to evaluate the information that we're taking

24:10

in and think about how much we trust it.

24:12

And trust is a whole complicated

24:16

area of exploration that

24:19

doesn't necessarily correspond

24:21

with how effective or

24:23

useful information is. Oftentimes, we

24:26

trust things that don't always help

24:28

us make the best possible decision.

24:30

At least that's what my work has been finding,

24:33

that they don't always go together.

24:35

So I think that that's a new problem that the

24:37

Internet is presenting

24:39

to us is that we now

24:41

have to evaluate trust, and we

24:43

have to forward

24:45

this information for us. And there's certain

24:47

people who are profiting off

24:50

of trying to get us to

24:52

trust things that aren't necessarily

24:54

true. So is

24:56

artificial intelligence going to

24:58

save us? If we can

25:00

use artificial intelligence to

25:02

sift through a lot of

25:04

this data and maybe identify

25:06

the things that are fake. You know, we're confronted with

25:08

a lot of information, as you just said, Dr.

25:10

Padilla, that isn't – it's just not

25:12

accurate. How will – what role

25:15

will AI be playing going

25:17

forward, do you think? I

25:18

mean, it's really gonna be playing a massive role.

25:20

Maybe not as much in our

25:22

own personal decisions, but certainly

25:24

decisions that are

25:25

made. On institutional government

25:28

level kind of high impact decisions.

25:30

And

25:30

I feel like I'm an outlier

25:32

in saying that AI can be

25:35

very useful. I've always every time I talk

25:37

with colleagues about this, I'm I'm the one who

25:39

thinks AI could be good. there's

25:42

this general apprehension to trust AI

25:44

to kind of give our decisions over

25:46

to artificial intelligence,

25:48

partly because there's a whole history

25:51

of AI also making biased

25:53

decisions because they're made by people

25:55

and people, you know, have biased input

25:57

data and have biased, you know,

25:59

different algorithms and so forth. And I

26:01

think that's a very real problem. And I'm very

26:03

excited that there's many ethicists

26:06

who are working on those

26:08

particular issues. But I do think that

26:10

we should start to get comfortable

26:12

with a future that is not far

26:14

off where many

26:16

decisions will be

26:18

curated in some capacity by AI.

26:20

Where rather than having thousands of things available

26:22

to us, there'll be some type of

26:25

decision scaffolding that

26:27

AI will be doing for us. And

26:30

that's the future that I see,

26:32

which I think could be a good one if, you

26:35

know, if there's a lot of people

26:37

involved in the AI development

26:39

process and not just certain groups of people

26:41

that don't necessarily

26:43

represent all of the interests.

26:46

Of the people making the decisions. Howard

26:48

Bauchner: Dr. PERFECTO, do you share

26:49

those thoughts, or do you come down on a

26:52

different side of the question? Yeah, I

26:54

think I think it's that it's that last

26:56

if that is the most important

26:58

component. Because as

27:00

Dr. Pete has said, it's And like we were

27:02

talking about with choice architecture, it's the

27:05

people are the ones making the

27:07

decisions, whether

27:09

it is in

27:11

terms of choice architecture actually forming the

27:13

decision that is being made for you

27:15

in that moment. The actual thing you're gonna

27:17

see when you're electing your

27:19

retirement plans or whether they

27:21

are farther upstream, creating

27:23

the thing that will eventually

27:26

help create a decision scaffolding,

27:28

as Dr. Fadia said, to

27:30

then present me with a

27:32

decision and its options down the road. So I

27:34

think that's a really big component to

27:36

consider. But I think everyone

27:38

is acknowledges

27:41

that may Dr. Pedia's main point

27:43

that it's happening. It's

27:46

AI and algorithms

27:48

are constantly increasing

27:50

every day their presence in our lives

27:53

and in our decision making.

27:55

And so there's a lot of work ongoing

27:58

in the field of decision making on

28:00

the academic side investigating how

28:03

people evaluate algorithms

28:06

and how we can

28:09

how people or how and

28:11

when people shy

28:13

away or shun completely

28:15

algorithm in certain context,

28:17

what happens when they make mistakes? How do

28:19

people respond to that? And how

28:22

we can improve people's

28:26

perceptions of them? And making

28:28

sure that they since

28:32

oftentimes, the the

28:34

algorithm is going to help you make a better

28:36

decision if the big if it's made

28:38

properly and has the right inputs.

28:41

And we want people to

28:43

be excited and

28:45

willing to go with this good decision

28:47

that that the algorithm is is

28:49

helping you make. So a lot of cool

28:51

work ongoing by

28:53

people helping learning

28:55

about ways to facilitate

28:57

that adoption of of algorithms in

28:59

people's decision making. So let me ask

29:01

you both. What are the big questions

29:04

you're working on

29:04

now. Padilla, what's

29:07

your research looking at and right

29:09

at this moment? I'm really interested

29:11

in the question of how

29:13

to facilitate trust and what

29:15

that means. I think

29:18

partly because I study a lot of

29:21

forecast visualizations, There's interesting

29:23

questions about how

29:26

much uncertainty is

29:28

useful to show, particularly

29:30

in facilitating trust. You

29:32

can imagine if you had a forecast of,

29:34

you know, that's, you know, flooding in California

29:37

right now. If there was a a

29:39

forecast that said

29:42

you know, every day this month, there was some potential

29:44

of flooding. That wouldn't be

29:47

correct. But in some ways, it would feel

29:49

like a cry wolf effect,

29:51

like it's it's saying that they'd be

29:53

flying too much going on. So

29:55

what you really want is a forecast

29:57

that gives you just the right

29:59

amount of information to make your decision and not

30:01

too much uncertainty because

30:03

then you can't, you know, make your decision with it.

30:06

So I'm looking at trade

30:08

offs and trying to identify the sweet

30:10

spot and how much uncertainty

30:12

we wanna show people and

30:15

how that can support both trust

30:18

and performance. I think there's

30:21

certainly thresholds there that that haven't

30:23

been carefully identified. And

30:25

Dr. Perfecto, what are you looking at

30:28

these days? I

30:30

teased it a little bit earlier. I'm

30:32

right now spending a lot of

30:35

time getting a sense of whether

30:37

and how we've been

30:40

creating these more abstract studies in

30:42

the lab, how

30:44

-- whether and how that influences the

30:48

likelihood of success down the

30:50

road. How closely is is

30:52

the lab matching the real

30:54

world. Since in the end, that's what we really

30:56

care about is is helping out the real world.

30:58

And so In

31:01

doing that, I am having lots

31:03

of fun going through, lots

31:06

of long standing literature, seeing

31:09

identifying common mistakes

31:12

made, rectifying those

31:14

mistakes, and getting a sense

31:16

of the impact that that

31:18

that might have in terms of

31:21

when this particular

31:23

error or bias might be

31:25

made out in the real world. And for

31:27

our listeners, any words

31:29

of wisdom? I think Dr. Perfecto, you talked

31:32

about not rushing to make decisions to even

31:34

just take take a few

31:36

seconds to think. Dr. Padilla, any other ideas

31:38

that you can offer to to our listeners?

31:40

You know, more practical advice on how to

31:42

how to make better decisions.

31:45

Findings situations in which

31:47

you can play around with the data

31:49

can start to give people an

31:51

intuitive sense of some

31:54

of the uncertainties and variability in the

31:56

data. I think what tends to happen

31:59

is people in the general

32:01

public they just get shown a forecast on the news

32:03

and they just have to

32:03

decide, do I trust it? Do I not? What do

32:06

I do

32:06

with it? And that's a very one-sided

32:10

path of the information. There's some

32:12

places that you can go online that allow

32:14

you to interact with the data and give

32:16

you the opportunity to say, what

32:18

if I extend this forecast a little or what if I change

32:20

the parameters of this forecast, and

32:22

then you can kinda get a sense of

32:24

what the different aspects are

32:26

affecting this particular model for really any

32:29

type of natural disaster

32:31

or no biological disaster. They

32:33

have these with COVID, have it

32:35

with fires and and hurricanes and so forth. So

32:38

if you're wanting to make a a better decision

32:40

for those types of things and you

32:42

sincerely care about it, consider

32:45

getting a little bit more comfortable with the data yourself and

32:47

finding those types of sites that

32:50

empower people with

32:52

the data that they can play

32:54

around with. That that could help to give people

32:57

this more intuitions about

32:59

what the forecast will do and and the uncertainty

33:01

that goes into those

33:02

forecasts. And doctor Perfecto,

33:04

last word, any practical advice?

33:06

In addition to hold on

33:08

a second. Hold on ten seconds.

33:10

And and think about

33:13

it. I would say

33:16

to make sure

33:19

you are bringing

33:21

the right tools to task that if it's

33:23

a very difficult decision, then you

33:25

should really take some time and sit with it. If

33:27

it's a really easy, like, what cereal am I

33:30

gonna have? Maybe don't. You

33:32

know, we only have so much time here on

33:34

this Earth. Just just grab grab

33:36

the cocoa puffs. Whatever. You can

33:38

you can take those specialty tomorrow. And

33:43

making sure that you are asked answering

33:45

the the question that is being

33:47

asked. So if if you are in

33:49

the supermarket and you're trying to

33:52

decide what to get. And

33:54

the easiest question is, what do I

33:56

want to buy a harder question is, what should

33:58

I be buying? What

34:01

what what is the best? What is the healthiest

34:03

option? What did my doctors say that I should

34:06

get? That's a that's a harder question to answer. Maybe when

34:08

you don't want to answer as much

34:10

as what do I

34:12

want. And so making sure that

34:14

you are actually

34:16

not substituting a different

34:18

question, an easier question for a

34:20

harder one can also

34:22

be helpful. Well, I wanna thank you both for joining

34:25

me today. This has been really interesting and

34:27

and enlightening. I appreciate your thoughts, and

34:29

I'm gonna grab the

34:29

CocoPuffs. Thank you. Enjoy.

34:32

It was great talking with you. Thanks

34:34

so much for having us.

34:36

For previous episodes of speaking of

34:38

psychology, you can visit us on our website

34:40

at WWW dot speaking of psychology dot org or you can find

34:43

us on Apple Stitcher or wherever you get

34:45

your podcasts. And if you like what you

34:47

hear, leave us a review. If

34:50

you have comments or ideas for future podcasts, you can email

34:52

us at speaking of psychology at

34:54

APA dot org. Speaking

34:57

of psychology is produced by Lee

35:00

Warnerman. Our sound editor is Chris

35:02

Condein. Thank you for

35:04

listening. To

35:06

the American psychological association. I'm

35:08

Kim Mills.