June 25, 2025 • 37 mins
When Yasmin Green joined Google almost two decades ago, she knew she wanted to focus on the complex problems facing humanity in a technological world – from violent extremism to political censorship. As the CEO of Jigsaw, a unit within Google that’s dedicated to understanding global challenges and applying technological solutions, she’s been able to do just that. Jigsaw’s most recent project took Green and her team to Bowling Green, Kentucky to figure out how they could use AI to improve a civic cornerstone: the town hall. Green sits down with Oz to talk about the project, What Could BG Be?, and what the role of Jigsaw is in today’s rapidly shifting, revenue-oriented tech landscape.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:13):
Welcome to tech stuff. This is the story today. A
conversation with Yasmin Green, the CEO of Jigsaw. Jigsaw is
a unit within Google that focuses on finding ways to
harness technology as a solution for societal issues at a
global and local scale. The Jigsaw team doesn't have any
revenue goals, and Google grants it the time and resources
(00:35):
to study intricate problems around the world, from scient extremism
to political censorship, to determine why it's happening and what
could be done about it, all from a technology and
solution oriented point of view. Jigsaw's most recent project took
Yasmin and her team to a small city in Kentucky
called Bowling Green. The challenge was to use AI tools
(00:58):
to create a town hall meeting for the technological age,
civic engagement that can be seen, heard, and considered all
beyond the walls of city hall. I've only has been
green for many years, and in an age where big
tech consistently drives negative headlines, I wanted to hear from
someone on the inside working to find best case scenarios
(01:18):
for the deluge of new tools and technologies that we're
now contending with. We sat down to talk about Bowling
Green and Jigsaw's past and future ambitions. He has been Green.
Welcome back to Well, last time we spoke was on Sleepwalkers.
This is tech stuff, but welcome back to the studio.
Speaker 2 (01:34):
Follow you around from podcast to podcasts.
Speaker 1 (01:36):
I think it's the other way around. Tell me a
bit about Jigsaw and your mission there.
Speaker 2 (01:43):
Yeah, Digsaw is an incubator inside Google. We developed technology
to give people voice and choice in the world around them.
And I have the privilege of being the CEO.
Speaker 1 (01:55):
And last time we met you at the director of research,
and now you're the CEO.
Speaker 2 (02:00):
I have broad a shoulders. Now my voice is a
little lower.
Speaker 1 (02:05):
But what have you brought to the new role? Like,
how have you put your stamp on the organization?
Speaker 2 (02:10):
So I came to New York actually fourteen years ago
to help start Jigsaw. So I feel like I've been
while I bleed Jigsaw. I've been with jigsaf from the
beginning and then three years ago I took over as
CEO and it just coincided with a time of incredible
change and volatility. But one of the things is the
kind of mainstreaming of conversational AI and both I think
(02:33):
the power of the air models and also the kind
of public discourse and public awareness about AI, and that
really changed. I think also the tempo inside all of
the tech companies around releasing and iterating and you know,
the near term gain is really clear, and then maybe
the long term pain is less in focus. But that's
(02:54):
kind of what we think we exist to do. A
jigsaw is be where the major tech companies and Google
aren't what our kind of highest and best use would
be are the things that other people aren't doing right now.
Speaker 1 (03:04):
Last time we met, we talked about a program you
ran to serve targeted ads to people considering the path
of terrorism.
Speaker 2 (03:13):
Right, yeah. Yeah. When we started working on radicalization, which
was one of our first focus areas over a decade ago,
the idea that the Internet would have anything to do
with why anyone would go and join the majorhaden in
Afghanistan was so unfathomable, both to people who are experts
about you know, Islamism and the people who are experts
in the Internet. Now, those you know, areas of kind
(03:35):
of extremism and hate, and you know, those things are
they are not solved problems, but they're also not in
anyone's behind spots like they are very crowded that the
subject of regulation, et cetera. And now we're much more
interested in, like, what if this all goes right and
we do build this incredibly powerful technology and it's prevalent,
how do we make sure that people have a voice
and choice in the world around them, because it's not
(03:57):
inevitable that that will come along with the intelligence.
Speaker 1 (04:01):
How did your sort of personal experience prepare you for
this role? I know you worked to Google beforehand, I
think in sales in Middle East and Africa and Europe,
and I know that you left Iran I think as
a young child as well. I mean, how how has
your kind of personal experience contributed to you being the
technologist that you are today based.
Speaker 2 (04:21):
In Paris for Google, managing strategy and ops for the
sales teams. And I got, of course, thing, do you
want to come to New York and help start Jigsaw?
And the original frame of Jigsaw was actually kind of
looking at geopolitical threats. The revolution that happened in Iran
which led to me leaving, where the religious extremists took over.
They were those were violent extremists, you know, and sometimes
(04:41):
they do they make big moves and they take over land,
and they can you know, oppress people. For such a
long time, we had this view that was prevalent in
tech then, which was, you know, the Internet will be
a democratizing technology. I don't have to worry it's going
to be bumpy along the way, but like, one thing
you can be sure is this is going to bring
people together and connect them to information. And I was like, yeah,
(05:02):
I think there are a lot of people going to
have their own idea about that, and so I wanted
to work on things to do with the Internet that
were very kind of discerning about what might happen. So
I think this idea of alays looking around the corner
and saying where's where's crowded now and where people aren't looking,
and having the luxury to kind of have this incredible
group of people be thinking about what's going to happen next,
(05:25):
and you know, having the resources of Google to try
to affect that.
Speaker 1 (05:29):
Jigsaw is a division of Google with I don't think
has any revenue. Doesn't have a revenue line, right, It's a.
Speaker 2 (05:34):
No, it does not.
Speaker 1 (05:36):
And Google's now in this moment where I think we'll
have a lot more questions about the larger role of
technology companies in our society. But then also the business
model with search and with the rise of AI is
more threatened. And so how does an organization like the
one you lead deal with those two things? On the
one hand, maybe being odds or surfacing things at the
(05:57):
parent company doesn't want to be surfaced. On the other hand,
operating in a more economically constrain the environment where there
are cuts and buyouts and stuff going on. How do
those two things affect you up day to day.
Speaker 2 (06:08):
I mean, there are so many parts of all the
tech companies that work on the things that are challenges.
You know, they will have these massive trust and safety
machinery that are actually looking all the time and like
a trillion pieces of content and everything from child sexual
assault imagery to the you know, the worst of it
and scams. So the tech companies are full of people
who are looking at where things aren't going well and
(06:30):
trying to affect them. I think maybe the thing that's
different is that Jigsaw is providing a public commentary and
there I think it probably helps that we have our
own brand, and so we go off and do our things,
and sometimes sometimes we try things that don't work, and
in which case Google's not too bollied. And then when
we do things that you really are very significant contributions
(06:51):
to helping make people better off with the Internet and AI,
then Google's like, this is Google Jigsaw everyone, we take
full credit. So it seems to be a good kind
of arrangement that really works for Google. But you're right
that we often are in that we are looking at
like what what is not necessarily going to go? Well,
it is in Google's long term interest to have a
group that does that and that makes significant progress that
(07:12):
they can rightfully take credit for.
Speaker 1 (07:14):
You're sort of in a sense you're not red teaming
particular products, but you're read teaming the cult that the
wider tech industry.
Speaker 2 (07:20):
Yeah, and e specifically with the view to like, what
are we going to do about it? Yeah. I assure
you that there are many many teams that are red
teaming internally. They just don't come and talk to you
about it, or maybe they can maybe actually got an
advocate work. It was so great. That was really great.
Speaker 1 (07:35):
Can you describe for our audience that text stuff, what
ethnographic research looks like and how you used it. If
you're projecting Bowling Green Kentucky.
Speaker 2 (07:45):
Yeah, depending on where you sit, you then press that
jigs or does it or you're kind of appalled that
not every part of a tech company does this. But
it's basically the very patient type of research that involves
going in and spending time with people and observing how
they live in their community and people are making sense
(08:05):
of using the Internet, which is the part we care
about as part of an existence that extends beyond technology obviously,
like who they are, what their identity is, what they do,
who their family is with their community is how they identify,
and so we bring that wide angle lens to how
we think about challenges. So we had built this big AI,
(08:25):
really helpful tool that was used by platforms and publishers
around the Internet to try to have conversations that serve
their communities. But last year we thought to ourselves, maybe
we could help enable conversations, large conversations that happen not
just online between people, but between policymakers and their constituents.
(08:47):
And we end up doing something which we've just completed,
which I'm excited to talk about in Kentucky, in this
town called Bonling Green. But we kind of understood that
they wanted to have a town wide conversation, but that
nobody participated. There was no public input when they were
given space to contribute.
Speaker 1 (09:06):
In fact, the only place you obviously a town hall
in America Arizona, Cinno, Fox News too use for the election, right.
The concept of the town hall is like more or
less a cliche. So how do you use technology to
actually have what would have the town hall in premodernaty?
Speaker 2 (09:20):
Actually? Okay, So we went to meet with the judge
executive of this county. It's like the mayor of this
county who's called Doug Norman. Incredible, I'm gonna call him Doug.
This town is a doubling in size over the next
twenty years. It's going from one hundred and forty eight manation.
Might not be getting those rights those numbers right, it
might be doubling even more. But it's basically like adding
another bowling green to the county, which sounds great like
(09:43):
growth and jobs and development. And at the same time
we're a mixed bag of emotions in the community. So
there's a big farming constituency and they are concerned about
land use. In the irreversible decision to let people develop
on your land, lifelong residents are concerned about the preserving
their culture and their building, and there's a lot of
concern about migration. And it's interesting that in Kentucky and
(10:05):
Bowling Green the concern is not about migration from foreigners
but from people from Nashville, Tennessee. Really, they're like, we
don't want to become like Nashville extended here. So there's
all like valid concerns that they have, and they're going
to have to have all you know, all this growth.
And actually the words from Doug capture it best. He
said to them, do we want this growth to happen
(10:25):
to us or do we want the growth to happen
for us? If we want to chappen for us, we
have to come together and set a vision for Bowling Green,
which is really amazing. But then you are similarly perfectly,
how's it going like with you know, Townholls, And he's like, yeah,
we get about eight or nine people, and so we're like,
maybe we can help because we have a lot of
experience with tech and large girl conversations. But it always
(10:46):
starts like product them always starts with like field work.
So we went to Bowling Green to go and spend
time with people, people who are really marginalized from politics,
like the elder, these students, recent migrants, et cetera, people
who've just been left incarceration, to figure out what what
would stop them from participating, And so we do things
like we took a couple to like a tree planting meeting.
(11:07):
We took like this college graduate to a city planning
commission get together about whether someone should be allowed to
add to the back of their pub. We took all
of them and so you can observe what is holding
them back, and it was the things that they said
resonated so much with me. They don't think that there
they have the expertise, they don't think they matter, they
(11:29):
don't think that if they use their voice it would
be listened to. And it's all stuff that who hasn't
been in like a large meeting where you're like, I
don't really understand what we're talking about, or I don't
understand my role in this discussion, or like if I
use my voice, you know my boss going to hear me,
or is it going to go anywhere? So if we
can relate to it, they get that level. The fact
that people really show up to their local community, whether
(11:51):
it's like their refugee community or their LGBTQ community, or
the sports club or the church. But then when you
like come to the town conversation, the civic oh, I
don't know, I don't look or sound the part. You know,
the college student we took to the city planning commission thing,
he was like, I can't go to that. I'm not
a lawyer. And of course when he went there were
no lawyers. But like, that's what we have in our head,
(12:13):
is like, I am the expertise.
Speaker 1 (12:16):
Can you tell me a little bit more about perspective API.
I know it's developed by Jigsaw and it's now used
by Wikipedia and Wall Street Journal, And essentially what it
does is regulate the comments sections using AI to understand
the impact any given comment might have on other commenters
or participants and score that impact. But I'm curious, is
(12:39):
the Bowling Green Project an extension of Jigsaw's work on perspective.
Speaker 2 (12:43):
It's hard to even remember that, you know, nine or
ten years ago, all these publishers are closing their comment
spaces because they couldn't even have a civil conversation, and
so Perspective was a you know, an AI that helps
with managing online conversations. They get ranks all different types
of things, everything from like ad hominem attacks too, is
something constructive and what explained like it scores them for
(13:05):
publishers to decide what kind of conversation you want to have.
But in the case of policymakers in there, you know
the people they represent, they need to be in conversation too.
There's this quote that I love, which is that conversation
is the soul of democracy. What gives democracies legitimacy is
our ability to have a free and open exchange of ideas.
And so the thing that we wanted to help Doug with,
(13:28):
which kind of informed what we developed, which we call
sense making, was he wants to tell his people come
and be in a conversation with me. People weren't showing up.
Speaker 1 (13:37):
So it wasn't like there's an existing conversation happening. How
do we make it more provatal inclusive? It was how
do we create the environment for a conversation to happen?
Speaker 2 (13:46):
Yes, and they need to believe that he will hear
their voice if they express it. People who study this say,
if there's nothing that's worse than not asking people for
their opinion.
Speaker 1 (13:55):
Asking people ignoring them because that's truly demotivating.
Speaker 2 (14:00):
So we realized that he would need to be able
to tell them that their voice would be heard, and
so we worked. We ended up working with a platform
called polist to solicit input and it was just open ended,
open ended like answers, so it was really in your
own words and on your own terms, So people were
free to say what they wanted. They're free to do
it at their own time, in their bedroom, you know,
(14:22):
an anonymously. So it was really like a lot of
what we heard in the ethnography we were trying to
account for. My team sat down with two sisters that
were recently relocated Bowling Green from Afghanistan, and they went
and in their home like with like over tree. Then
they were like you know, with the headscarves everything. They
were asking them like what what would it take for
you to come and be part of this conversation And
(14:42):
they were just like that, they had no idea what
civic participation referred to, the idea of like even attending
a town hall, never mind being asked to speak, was
a complete anathema to them. So this conversation that we
set up, which is called what could Bowling Green be,
which we did with this incredible partner on the ground
called Innovation an Engine. The conversation was like designed to
(15:02):
be like you are free to say what you want,
where you want, whenever you want. And it was a
month long conversation and we did it through this POSE
platform and the idea was speak and you'll be heard.
And the part that we used AI for was making
sense of what everybody said in a way that could
be shared back to them, that they would see their
voices in the conversation and could be shared back with
(15:24):
the policymaker so he could take an action.
Speaker 1 (15:37):
After the break, the largest town hall in US history
and how Jigsaw made it happen. Welcome back to tech stuff.
Before the break, we were talking with Yasmin Green, CEO
(15:59):
of Jigsaw, about how the initial pieces of an ambitious
project using AI to create a more inclusive and active
town hall came together in the town of Bowling Green, Kentucky. So,
taking a few steps back, how does this come about? Like,
how do you choose a community to partner with? And
I guess what are the stakes for you to be
working with government versus just in a purely online arena,
(16:22):
Like how much pressure and also opportunities.
Speaker 2 (16:24):
Does that add? Kentucky's actually a really pioneering state, and
partly a little bit by it because now I spent
much time there, but the cross even not in Bonling Green,
other places do. They're doing a lot of civic tech
work and they're very forward. And Bowling Green in particular
was known because they had done something several years ago
where they had done public participation with the local newspaper. Okay,
and we had heard that they're putting together the stake
(16:44):
vision for twenty fifty.
Speaker 1 (16:46):
So what do we want Bowling Green? Yeah, so you'd
heard about this, you knew that they would take forward.
Speaker 2 (16:50):
Yes, and we're like, maybe we can actually they come
and give them something really useful. Maybe we can co
develop something with them. And so they said, okay, let's
do this month and conversation to hear what people have
to say. And then they actually did like super local
like ground game marketing campaign.
Speaker 1 (17:04):
Where you involved with a marketing campaign or.
Speaker 2 (17:06):
Yeah, we partnered with them, but it was like local
creative agencies that came up with the creators.
Speaker 1 (17:10):
So basically get people to be aware and think it
was fun, evangelize interact with the platform.
Speaker 2 (17:15):
Yeah, yeah, and exactly, And the main thing was just
to be local. So things were translated into nine languages.
They they put them in you know, they fly at
international supermarkets, they went to make a church is, they
went to basketball games.
Speaker 1 (17:28):
So you created awareness and then whether like QR codes
that people like clicked on their phones and they would
take it into the conversational platform or like how what
was the.
Speaker 2 (17:36):
Yeah, basically you look, yes, they're working on code. Do
you go to what could be GB? In fact, you
can stop it. What could Bowling Green be? Is the
website dot com and you can see the report. But
people would log in and then you just get like
a question prompt what do you want for Bowling Green
in the future, Which that's that depended you put your
answers in and.
Speaker 1 (17:55):
Only one single prompt.
Speaker 2 (17:56):
There's one question and could you could put you could
keep putting more of what you want. You probably have lots.
It's pretty open. You might have ideas for education or infrastructure,
you know, restaurants. But then there's voting, so you you
vote on other people's input. And that was a really
important signal to see like where, if anywhere, do we
agree right and where we put so much effort into
trying to make a like make this local and be
(18:18):
make it accessible and still like you know, the beginning
of the month long conversation, we were like, is anyone
is anyone going to turn up to this? Yeah? Because
maybe we do all this effort and it's eight people.
Speaker 1 (18:29):
Did you show a tumble.
Speaker 2 (18:31):
Right? Right? So the baseline's eight okay, So we were like,
thirty will be good. Thirty will be good. Now, so
we're like, if we could get a one thousand or
two thousand, and so it was eight.
Speaker 1 (18:44):
Thousand, eight thousand respondents.
Speaker 2 (18:46):
This kind of like AI enabled town hall was a
thousand times larger than the regular town hall, and one
in every ten residents basically the equivalent of one every
ten residents participated. So as far as like that's what
the folks on the ground and say, is that it
meant that people were discussing this topic even outside of
the actual platform because there was that much participation, So
(19:07):
it kind of pas the tipping point. So it was
really nice to see the numbers go up. They did
go up, they did spike a bit towards the end,
and it was actually the largest digital town hall in
US history. So it's the most that anyone's come together,
and that's US partnering with a town in Kentucky. So
if you like imagine what is possible at the state
level or across states if you bring people together to
(19:28):
share their view. Although there was before like advocating for
more of those type of CIF conversations, I think the
first big apprehension was is anyone could show up? Ye?
The second was, well, you know, if this is a
scaled version of an in person town hall, we might
kind of be in trouble because the eight people that
come to an in person town hall detractors detracted typically, right,
(19:51):
the ambivalent people don't go and show up, or people
who are like quite like the idea and also could
see why not the idea? That's not the people that
get you motivated to get off the counch and drive
all the way and wait there. It's usually people who
really don't want the thing to happen. So I think
local leaders were like, well, we're just going to get
eight thousand of them entrance, you know, because they don't
want to scale in that sense, And that was why
it's so important to do like such a broad marketing
(20:13):
campaign to like all these different constituencies, but the bigger
side of relief from local leaders when they looked at
their feedback and a lot of stuff they anticipated. There
was like in a growing town, you need more infrastructure,
you need more roads. Lots of really interesting thoughts on education,
on like restaurants. There was just one submission on the
restaurant and interesting kind of like unexpected proposals on a
(20:36):
number of fronts, but the one that makes me giggles
in the area of like food and restaurants. One person
submitted an idea to pass a law to mandate restaurants
to include ketchup saches in the takeaway like the takeaway
bags of like fast food restaurants. So you know, that's
how you know that people are really expressing their views.
And we've got good mixed views but didn't have much support.
Speaker 1 (20:56):
How did people express support for one another? Like I
go on the platform, answer this is you know, I
want more free ketchup saches? Is that that Now I
don't live in Boling. I read that it was Geo
Geo fence. Otherwise it would have been me. But then
is that then like available for other people to comment on? Okay,
so it's like a message boarders, it's not just a one.
Speaker 2 (21:16):
So they're routing so you you put your idea and
then they're routing ideas for you to vote on. And
it's actually much You get a much higher percentage of
people like the The votes are much higher than the
actual submissions. Yeah, because it's much easier to vote and
you're kind of in there and it's a little addictive.
And I think that's one of the really really cool
things about this conversation is that it was designed to
(21:39):
advance a very large conversation. It wasn't designed to have
like eight thousand and one on ones. That's not what
we're trying to do when we're trying to have a
long conversation. And that that's interesting because there isn't the
option to reply the way that purchases platform is set up.
You have an idea, and you can vote on other
people's ideas, and if you really don't like the catch
up idea and you actually think it should be mustered,
(21:59):
put your idea in the area. Yeah, but we're not
you're not telling you know, you're not going to get
into some you know, back and forth with the person
who suggested Ki jump is you know, from the wrong
part of town, and everyone's asking that part of town.
You know, here's no point going down that you know,
spiral of the ad hominym attacks, et cetera. That happens
a not on social media. So the conversations dis like
designed differently from the get go that you put your ideas,
(22:21):
you vote, if you want to keep going after you've
given your thumbs up your thumbs down, you're welcome to
put your idea in there. And it well. So then
when we had to take stock of like, well did
any of these ideas they resonate with other people? So
there were four thousand policy proposals and over half of
them had near universal support.
Speaker 1 (22:38):
Near universal support over half. So in other words, what
you take from it is the community knows what's good
for it.
Speaker 2 (22:46):
Well yeah, and and that you know, there are so
many opportunities for us to hear from the people who
are the angriest or at least well informed.
Speaker 1 (22:54):
Or did you know the voter registration of the participants
or I mean.
Speaker 2 (22:57):
Was this no, but it's it's it is a it's
a purpolish place, so it's quite a mix. And you
saw some of the things that you see you know, nationally,
but most of the things that local people care about
are you know, what is going to happen to their
K through twelve education. It turns out they want more
vocational skills, you know, like they want their university, which
is the pride of like you know, Western Kentucky University,
(23:19):
they want it to be more integrated into their economy
and their like workforce development. And there's a lot of
things that they want. And the perspective of not just
dog actually but even like private section leaders was like,
that's great. Now we have now we can go and
lead and build and make changes knowing that we have
broad public support. Otherwise, if you're just based basing you know,
(23:41):
your like sentiment gauging on public calls and social media,
just thinking that like no one wants anything to happen,
you know, or that like you might have some supporters,
but also like everyone else is an avid, you know,
avid distractor. So so I think those are two really
important essons. One people really will show up, they really
do want to shape their future, you know, And two
(24:03):
that there is more that unites us then divides us,
which is kind of imperative for us to keep you know,
in our minds and cultivate at this moment.
Speaker 1 (24:12):
And what was Jigsaw's role or technology deployment because the
marketing campaign was like panel by local agencies, the polished
platform was it's not a Jigsaw platform, right, So what
did what was the layer that you guys added It was.
Speaker 2 (24:27):
The making sense of the conversation. So if you can
get lots of people to contribute, how can you guarantee
that their voice will be heard? And so it really
is using this case, Google's aigemini using an eye for
what it does best, which is they take large amounts
of information that it expressed in a very local and
personal way, understand it, organize it, they sort it, prioritize it,
(24:49):
and then give it back to people. So like, you
know the thing if you go to what could bowlinggreenpeed
dot com, you'll see the report and it's like, you know,
Google Maps, we think at Google mapps, you started very
high level. If you'll look at like the states of
in America, that's the first thing that you see, and
then you're like, actually, I'm kind of interested in New York.
And then you're like interest in my neighborhood and they're interesting.
You're like, oh, how can you see my front door?
(25:10):
You know, And that's how it is with this map
of opinions, like not geographical terrain, but opinion terrain. But
you start the top and you're like, what do people
have born and green care about? And then it has
those you know, instructural education, et cetera. And then you're
driven You're like, actually, what I really care about is
you know, education, and then go through, you know, you
just keep digging and most people, most people have specific
areas of interest. But the truth is people want to
(25:31):
know where they sit relative to their neighbors, and so
people want to actually find where they are. Because there's
this plot graph on there that has a distribution of
all the proposals and how much agreement they got, and
so you can see am I like on the am
I the fringe opinion holder even though like me and
all my two friends agree with me, or actually is
something that I've been scared to say out loud, something
(25:52):
that a lot of other people believe. So you can
go in there and see that, which is really powerful.
And of course most people agreed with most things. So
I think it does restore our sense of you know,
like it's a funhouse of mirror sometimes social media where
you're like, oh my goodness, everyone's extreme. It's just me
here that doesn't want to have like an extreme mistake
on things.
Speaker 1 (26:12):
What are the key policy ideas that have emerged and
will you measure success based on whether or not they're
enacted or are you, in a sense now passing the
bat on the lull.
Speaker 2 (26:22):
It's definitely in the court of their leaders, and it's
their process so to begin with. So I think they
genuinely do want people's input because the process kind of
proceeded our involvement. But one example was they had wanted
to do a riverside development. They built building up the
area along the riverside, and when they had got tried
to get public input, very few people pay and they
(26:44):
were detractors. But then if you look at the feedback
on that one topic in the reports, like seventy five
plus percent of people support it. So I think they're
now going to accelerate that. What we did, in terms
of like really trying to gauge people's sense of like
whether it mattered for them, was we say it on
whether they felt their voice mattered. We asked them whether
they felt they understanded other people's perspectives better than they
(27:06):
did before. And then third was where they think they
believe their input will lead to a better outcoming. On
all of those, it was like eighty percent plus satisfaction.
Speaker 1 (27:15):
And were you surprised by the outcome? And how do
you how do you kind of measure the success of
a project like this.
Speaker 2 (27:23):
I think it was that that people wanted to come
out and they did. I think that there's two There's like,
there's two parts of my brain. There's a part of
brains like did we achieve our you know, like our
product development goals or are like whatever engagement goals? And
then there's an other part of me that's like can
I retain my hope for like society thriving as we
(27:45):
get more at Amize And so I feel like on
both of those I felt really good. I felt good
because obviously the technology wor I also, you know, the
sense of the team is that we can push to
do more. And so now we're exploring what other types
of conversations in the conversation space. How could we Maybe
it's a different scale, you know, maybe it's a different
nature of conversation. Maybe we're using different capabilities of Gemini
(28:08):
to do more to bring people together, to have them
understand each other more and then having it count for them.
So that's the kind of fun and exploration phase that
we're in. Now.
Speaker 1 (28:18):
There's another thing that came out of the research I
think called grounding.
Speaker 2 (28:21):
Yeah, can you talk a bit about rounding. Groundings their
sister to hallucinations, So, which is a funny term that
has widespread adoption to describe when air models may make
stuff up.
Speaker 1 (28:33):
Because of course that the very bad scenario. Who would
be that the air was making up policy of pro
puzzles and no one asked for.
Speaker 2 (28:39):
Oh my goodness, imagine that in such a high stakes
context where people are trying and talk to their policy maker.
And there were some funny things that in our early
testing we were finding it would kind of conflate two
topics that seem like they're about the same thing, or
that like you would it would have only one. There'd
be like one post about you know, there should not
(28:59):
be some any sweet tuitets offered at in kindergarten. And
then the summarizing AI would be like the people of
Bowling Green, you do not want to have like cakes
and chocolates and candy and they would rather have like
keep compisation. It's just like wait, whatever that wasn't said, yos,
that's out of lines. So part of kind of making
sure that we understood the ways that the AI might
(29:21):
misfire was doing a ton of testing with humor reviews
and on previous data sets, and then we did it
kind of kind of like a small private trial a
few months earlier with Bowling Green and the local partners
reviewing them and just seeing like how might it go wrong?
And then some of the tweets we made with to
make sure that we batched appropriately. So grounding is every
(29:41):
output that the AI produces is that it's like appended
with citations. So everything that the AI says was a
theme in the report, you can say, you know, show
me receipts and it will take you to the underlying
policy proposals or words, the actual words of the people
at Bowling Green, so you don't have to you know,
(30:03):
you actually don't have to trust the AI.
Speaker 1 (30:05):
Super interesting. Could grounding be used beyond this context? Do
you think?
Speaker 2 (30:09):
Yeah? Yeah, and it is it's used broadly. Have you
ever used notebook M Yes, So they use grounding. So
when they you know, when they summarize a you can
give it like one hundred reports or whatever backad Democrat.
Of course, if you're don't research and it'll tell you
the themes and then it gives you like like footnotes
so that you can see where the source material and
have confidence that at accurate.
Speaker 1 (30:31):
What's the wider hope with a project like Bowling Green?
I mean, do you is that, like, is there an
ideal community size? Why this is will be perfect for
towns that are growing of around one hundred thousand people
because like you have enough people in the community, you're
invested in the future because it's growing, it's not so
large that you get so many proposals that it's like
impossible to sort. And therefore maybe there's like another one
(30:52):
hundred Bowling Greens in the US, or is it more like, Wow,
the greatest problem in US politics today is that participation
and is low and conversation is toxic, and therefore, like
this could be a much wider solution, Like how do
you between those two things? Like where do you feel
for me?
Speaker 2 (31:09):
The more conversation, the better. I think the reverse is
also true, the less conversation the worst, the worst we
ended up. I understand each other, and that was kind
of where we had taken perspective for over the last
couple of years, was actually not just helping publisher score
things that they might want to see less of, like toxicity,
but actually doing the bridging attributes as well, which was
the things that kind of help people who disagree stay
(31:32):
in the conversation together. So I would like to see
more of it. And I think we've also shown we
show that the appetite is there from people and that
it can go really well. That I think the judge
executive Doug Gorman in our case, was really really brave
to try and do that. Why brave because the safe
nobody shows up, or say if they show up and
you know it's Cook Coffinus. But now seeing what Doug's
(31:56):
that actually Doug came to So Google has a developer
conference equal to I Air. That just happened, and Google
invited Doug to go and participate in the local leaders track,
and he went and told all these like other mayors
and local leaders and they were They really lit up.
And so from what I understand, I think there is
a lot of appetite to do this, and I think
him going first and having the bowling green kind of
(32:17):
proof of concept will inspire others to adopt it to.
Speaker 1 (32:21):
Is that part of your work or in a sense
you like the fire and it's of other people to
kindle it bad metaphor, but I mean otherwise do what
are you going to do the same experiment again in
another community or a larger community? Or do you feel
like you have the proof point that you think you know?
Speaker 2 (32:36):
I think it's you know this question, like what is
Jigsaw's highest and best use. I think if we if
now we've done the concept, if there's another proof like
another place that we'd want to do it, I think
we would be up into that if it felt that
it was some additive like materially additive, if it were
you know, maybe at the state level or you know,
some other kind of complex conversation that we felt that
(32:58):
we could we could advance. I think there will be others,
like kind of implementers who will I think this will
become a kind of a commercial opportunity for others to
help with. I don't think we do you know, that's
the best use of Jigsaw to do many many more
of the same thing in towns.
Speaker 1 (33:17):
So in other words, like a polling firm could take
it over as a business. Liane Audrey Tang, who I
know that you were, you know, to spend time with.
Recently an interview with Nick Thompson and Audrey spoke about
the Bowling Green project and in the context of how
in California post fires, Newsome is sort of using this
(33:38):
tech enabled participatory democratic technology to understand what people want
from the rebuild of like burnt areas of LA. Where
does this fit in with the kind of larger tech
enabled digital democracy trend.
Speaker 2 (33:56):
I think it is. I think they're pioneering there because
they're doing at the state level in California's a massive state.
And interestingly, when we did it in Bowling Green, which
was you know, Well or Warren County, much much smaller,
super open question. In the case of California, that asking
a really narrow question. I mean, it's big in itself,
but it's much more focused question, which is, you know,
(34:16):
how are we going to where do we go from
here after the wildfire specifically, So I think that's really instructive,
Like doesn't once you go to the state level, doesn't
need to be so focused And they've already got quite
a bit of engagement and is she going to talk
to them about how they're going to make sense of
it because I don't know what they do. Well, yeah,
we're in touch with them. And Audrey, who's you know,
it was the former the kind of in agural like
(34:38):
I guess did a digital minister for Taiwan is like
this just she's just a global matchmaker. So she's advising
California as well, and she's advised us a lot. She's
been very instructive for our work.
Speaker 1 (34:50):
Two questions are close, I guess the first. Obviously, it's
a moment where Google is being you know, the subject
of the Justice Department and Chrome. All this kind of stuff,
in other words, is a target of federal investigation when
you work with government, like do you are there people
who say this is a conflict of interest or Jigsaw
(35:10):
being in bed with local estate governments is somehow advancing
Google's corporate motives elsewhere, and that's a problem.
Speaker 2 (35:18):
In this case. We didn't actually have any commercial relationship
with the government. We were actually partnered with the group
the I told you about Innovation Engine and they did
the kind of execution. But we did care a lot
about making the tech work for the hotesty maker and
the people because it seemed it seems like such an
important thing for us to get right. And if we
(35:40):
want policy makers to make good policy, even about tech,
they should kind of understand it and see how it
shows up for them. And in this case, I think
Voluningroom was so attractible was so because it was so
mixed in terms of politics. So, yeah, we've always kind
of been in and we've always been around governments, and
actually still this point you haven't actually directly partnered with
(36:02):
the government. But it felt really gratifying to feel like
you are helping that dynamic work better.
Speaker 1 (36:09):
What do you still have to achieve a Jigsaw and
what kind of future signature projects can you tease for
us Today.
Speaker 2 (36:17):
We're noticing a change in the culture at Jigsaw internally
that actually mirrors what's happening externally. That things are moving
so fast, like the technology is moving so fast, and
we are inside this incredible tech company, and so I
think we sometimes maybe things appear to us before they
mate to others. I think we could help a lot
(36:38):
by kind of sharing as we're going. So I think
the change in the coming year that I'd like to
see is more rapid iteration and more, sharing more frequently,
and being more public about what we're doing.
Speaker 1 (36:50):
Well, you know where to come. Thank you, Thank you
us for tech Stuff. I'm oz Vaaloshian. This episode was
(37:16):
produced by Eliza Dennis and Adriana Topia. It was executive
produced by me Karaen Price and Kate Osborne for Kaleidoscope
and Katrina Norvel for iHeart Podcasts. Jack Insley mixed this episode,
and Kyle Murdoch Rodelphime Song join us on Friday for
the Week in Tech, when we'll run through the tech
headlines you may have missed. Please rate, review, and reach
(37:39):
out to us at tech Stuff podcast at gmail dot com.
Welcome to tech stuff. This is the story today. A
conversation with Yasmin Green, the CEO of Jigsaw. Jigsaw is
a unit within Google that focuses on finding ways to
harness technology as a solution for societal issues at a
global and local scale. The Jigsaw team doesn't have any
revenue goals, and Google grants it the time and resources
(00:35):
to study intricate problems around the world, from scient extremism
to political censorship, to determine why it's happening and what
could be done about it, all from a technology and
solution oriented point of view. Jigsaw's most recent project took
Yasmin and her team to a small city in Kentucky
called Bowling Green. The challenge was to use AI tools
(00:58):
to create a town hall meeting for the technological age,
civic engagement that can be seen, heard, and considered all
beyond the walls of city hall. I've only has been
green for many years, and in an age where big
tech consistently drives negative headlines, I wanted to hear from
someone on the inside working to find best case scenarios
(01:18):
for the deluge of new tools and technologies that we're
now contending with. We sat down to talk about Bowling
Green and Jigsaw's past and future ambitions. He has been Green.
Welcome back to Well, last time we spoke was on Sleepwalkers.
This is tech stuff, but welcome back to the studio.
Speaker 2 (01:34):
Follow you around from podcast to podcasts.
Speaker 1 (01:36):
I think it's the other way around. Tell me a
bit about Jigsaw and your mission there.
Speaker 2 (01:43):
Yeah, Digsaw is an incubator inside Google. We developed technology
to give people voice and choice in the world around them.
And I have the privilege of being the CEO.
Speaker 1 (01:55):
And last time we met you at the director of research,
and now you're the CEO.
Speaker 2 (02:00):
I have broad a shoulders. Now my voice is a
little lower.
Speaker 1 (02:05):
But what have you brought to the new role? Like,
how have you put your stamp on the organization?
Speaker 2 (02:10):
So I came to New York actually fourteen years ago
to help start Jigsaw. So I feel like I've been
while I bleed Jigsaw. I've been with jigsaf from the
beginning and then three years ago I took over as
CEO and it just coincided with a time of incredible
change and volatility. But one of the things is the
kind of mainstreaming of conversational AI and both I think
(02:33):
the power of the air models and also the kind
of public discourse and public awareness about AI, and that
really changed. I think also the tempo inside all of
the tech companies around releasing and iterating and you know,
the near term gain is really clear, and then maybe
the long term pain is less in focus. But that's
(02:54):
kind of what we think we exist to do. A
jigsaw is be where the major tech companies and Google
aren't what our kind of highest and best use would
be are the things that other people aren't doing right now.
Speaker 1 (03:04):
Last time we met, we talked about a program you
ran to serve targeted ads to people considering the path
of terrorism.
Speaker 2 (03:13):
Right, yeah. Yeah. When we started working on radicalization, which
was one of our first focus areas over a decade ago,
the idea that the Internet would have anything to do
with why anyone would go and join the majorhaden in
Afghanistan was so unfathomable, both to people who are experts
about you know, Islamism and the people who are experts
in the Internet. Now, those you know, areas of kind
(03:35):
of extremism and hate, and you know, those things are
they are not solved problems, but they're also not in
anyone's behind spots like they are very crowded that the
subject of regulation, et cetera. And now we're much more
interested in, like, what if this all goes right and
we do build this incredibly powerful technology and it's prevalent,
how do we make sure that people have a voice
and choice in the world around them, because it's not
(03:57):
inevitable that that will come along with the intelligence.
Speaker 1 (04:01):
How did your sort of personal experience prepare you for
this role? I know you worked to Google beforehand, I
think in sales in Middle East and Africa and Europe,
and I know that you left Iran I think as
a young child as well. I mean, how how has
your kind of personal experience contributed to you being the
technologist that you are today based.
Speaker 2 (04:21):
In Paris for Google, managing strategy and ops for the
sales teams. And I got, of course, thing, do you
want to come to New York and help start Jigsaw?
And the original frame of Jigsaw was actually kind of
looking at geopolitical threats. The revolution that happened in Iran
which led to me leaving, where the religious extremists took over.
They were those were violent extremists, you know, and sometimes
(04:41):
they do they make big moves and they take over land,
and they can you know, oppress people. For such a
long time, we had this view that was prevalent in
tech then, which was, you know, the Internet will be
a democratizing technology. I don't have to worry it's going
to be bumpy along the way, but like, one thing
you can be sure is this is going to bring
people together and connect them to information. And I was like, yeah,
(05:02):
I think there are a lot of people going to
have their own idea about that, and so I wanted
to work on things to do with the Internet that
were very kind of discerning about what might happen. So
I think this idea of alays looking around the corner
and saying where's where's crowded now and where people aren't looking,
and having the luxury to kind of have this incredible
group of people be thinking about what's going to happen next,
(05:25):
and you know, having the resources of Google to try
to affect that.
Speaker 1 (05:29):
Jigsaw is a division of Google with I don't think
has any revenue. Doesn't have a revenue line, right, It's a.
Speaker 2 (05:34):
No, it does not.
Speaker 1 (05:36):
And Google's now in this moment where I think we'll
have a lot more questions about the larger role of
technology companies in our society. But then also the business
model with search and with the rise of AI is
more threatened. And so how does an organization like the
one you lead deal with those two things? On the
one hand, maybe being odds or surfacing things at the
(05:57):
parent company doesn't want to be surfaced. On the other hand,
operating in a more economically constrain the environment where there
are cuts and buyouts and stuff going on. How do
those two things affect you up day to day.
Speaker 2 (06:08):
I mean, there are so many parts of all the
tech companies that work on the things that are challenges.
You know, they will have these massive trust and safety
machinery that are actually looking all the time and like
a trillion pieces of content and everything from child sexual
assault imagery to the you know, the worst of it
and scams. So the tech companies are full of people
who are looking at where things aren't going well and
(06:30):
trying to affect them. I think maybe the thing that's
different is that Jigsaw is providing a public commentary and
there I think it probably helps that we have our
own brand, and so we go off and do our things,
and sometimes sometimes we try things that don't work, and
in which case Google's not too bollied. And then when
we do things that you really are very significant contributions
(06:51):
to helping make people better off with the Internet and AI,
then Google's like, this is Google Jigsaw everyone, we take
full credit. So it seems to be a good kind
of arrangement that really works for Google. But you're right
that we often are in that we are looking at
like what what is not necessarily going to go? Well,
it is in Google's long term interest to have a
group that does that and that makes significant progress that
(07:12):
they can rightfully take credit for.
Speaker 1 (07:14):
You're sort of in a sense you're not red teaming
particular products, but you're read teaming the cult that the
wider tech industry.
Speaker 2 (07:20):
Yeah, and e specifically with the view to like, what
are we going to do about it? Yeah. I assure
you that there are many many teams that are red
teaming internally. They just don't come and talk to you
about it, or maybe they can maybe actually got an
advocate work. It was so great. That was really great.
Speaker 1 (07:35):
Can you describe for our audience that text stuff, what
ethnographic research looks like and how you used it. If
you're projecting Bowling Green Kentucky.
Speaker 2 (07:45):
Yeah, depending on where you sit, you then press that
jigs or does it or you're kind of appalled that
not every part of a tech company does this. But
it's basically the very patient type of research that involves
going in and spending time with people and observing how
they live in their community and people are making sense
(08:05):
of using the Internet, which is the part we care
about as part of an existence that extends beyond technology obviously,
like who they are, what their identity is, what they do,
who their family is with their community is how they identify,
and so we bring that wide angle lens to how
we think about challenges. So we had built this big AI,
(08:25):
really helpful tool that was used by platforms and publishers
around the Internet to try to have conversations that serve
their communities. But last year we thought to ourselves, maybe
we could help enable conversations, large conversations that happen not
just online between people, but between policymakers and their constituents.
(08:47):
And we end up doing something which we've just completed,
which I'm excited to talk about in Kentucky, in this
town called Bonling Green. But we kind of understood that
they wanted to have a town wide conversation, but that
nobody participated. There was no public input when they were
given space to contribute.
Speaker 1 (09:06):
In fact, the only place you obviously a town hall
in America Arizona, Cinno, Fox News too use for the election, right.
The concept of the town hall is like more or
less a cliche. So how do you use technology to
actually have what would have the town hall in premodernaty?
Speaker 2 (09:20):
Actually? Okay, So we went to meet with the judge
executive of this county. It's like the mayor of this
county who's called Doug Norman. Incredible, I'm gonna call him Doug.
This town is a doubling in size over the next
twenty years. It's going from one hundred and forty eight manation.
Might not be getting those rights those numbers right, it
might be doubling even more. But it's basically like adding
another bowling green to the county, which sounds great like
(09:43):
growth and jobs and development. And at the same time
we're a mixed bag of emotions in the community. So
there's a big farming constituency and they are concerned about
land use. In the irreversible decision to let people develop
on your land, lifelong residents are concerned about the preserving
their culture and their building, and there's a lot of
concern about migration. And it's interesting that in Kentucky and
(10:05):
Bowling Green the concern is not about migration from foreigners
but from people from Nashville, Tennessee. Really, they're like, we
don't want to become like Nashville extended here. So there's
all like valid concerns that they have, and they're going
to have to have all you know, all this growth.
And actually the words from Doug capture it best. He
said to them, do we want this growth to happen
(10:25):
to us or do we want the growth to happen
for us? If we want to chappen for us, we
have to come together and set a vision for Bowling Green,
which is really amazing. But then you are similarly perfectly,
how's it going like with you know, Townholls, And he's like, yeah,
we get about eight or nine people, and so we're like,
maybe we can help because we have a lot of
experience with tech and large girl conversations. But it always
(10:46):
starts like product them always starts with like field work.
So we went to Bowling Green to go and spend
time with people, people who are really marginalized from politics,
like the elder, these students, recent migrants, et cetera, people
who've just been left incarceration, to figure out what what
would stop them from participating, And so we do things
like we took a couple to like a tree planting meeting.
(11:07):
We took like this college graduate to a city planning
commission get together about whether someone should be allowed to
add to the back of their pub. We took all
of them and so you can observe what is holding
them back, and it was the things that they said
resonated so much with me. They don't think that there
they have the expertise, they don't think they matter, they
(11:29):
don't think that if they use their voice it would
be listened to. And it's all stuff that who hasn't
been in like a large meeting where you're like, I
don't really understand what we're talking about, or I don't
understand my role in this discussion, or like if I
use my voice, you know my boss going to hear me,
or is it going to go anywhere? So if we
can relate to it, they get that level. The fact
that people really show up to their local community, whether
(11:51):
it's like their refugee community or their LGBTQ community, or
the sports club or the church. But then when you
like come to the town conversation, the civic oh, I
don't know, I don't look or sound the part. You know,
the college student we took to the city planning commission thing,
he was like, I can't go to that. I'm not
a lawyer. And of course when he went there were
no lawyers. But like, that's what we have in our head,
(12:13):
is like, I am the expertise.
Speaker 1 (12:16):
Can you tell me a little bit more about perspective API.
I know it's developed by Jigsaw and it's now used
by Wikipedia and Wall Street Journal, And essentially what it
does is regulate the comments sections using AI to understand
the impact any given comment might have on other commenters
or participants and score that impact. But I'm curious, is
(12:39):
the Bowling Green Project an extension of Jigsaw's work on perspective.
Speaker 2 (12:43):
It's hard to even remember that, you know, nine or
ten years ago, all these publishers are closing their comment
spaces because they couldn't even have a civil conversation, and
so Perspective was a you know, an AI that helps
with managing online conversations. They get ranks all different types
of things, everything from like ad hominem attacks too, is
something constructive and what explained like it scores them for
(13:05):
publishers to decide what kind of conversation you want to have.
But in the case of policymakers in there, you know
the people they represent, they need to be in conversation too.
There's this quote that I love, which is that conversation
is the soul of democracy. What gives democracies legitimacy is
our ability to have a free and open exchange of ideas.
And so the thing that we wanted to help Doug with,
(13:28):
which kind of informed what we developed, which we call
sense making, was he wants to tell his people come
and be in a conversation with me. People weren't showing up.
Speaker 1 (13:37):
So it wasn't like there's an existing conversation happening. How
do we make it more provatal inclusive? It was how
do we create the environment for a conversation to happen?
Speaker 2 (13:46):
Yes, and they need to believe that he will hear
their voice if they express it. People who study this say,
if there's nothing that's worse than not asking people for
their opinion.
Speaker 1 (13:55):
Asking people ignoring them because that's truly demotivating.
Speaker 2 (14:00):
So we realized that he would need to be able
to tell them that their voice would be heard, and
so we worked. We ended up working with a platform
called polist to solicit input and it was just open ended,
open ended like answers, so it was really in your
own words and on your own terms, So people were
free to say what they wanted. They're free to do
it at their own time, in their bedroom, you know,
(14:22):
an anonymously. So it was really like a lot of
what we heard in the ethnography we were trying to
account for. My team sat down with two sisters that
were recently relocated Bowling Green from Afghanistan, and they went
and in their home like with like over tree. Then
they were like you know, with the headscarves everything. They
were asking them like what what would it take for
you to come and be part of this conversation And
(14:42):
they were just like that, they had no idea what
civic participation referred to, the idea of like even attending
a town hall, never mind being asked to speak, was
a complete anathema to them. So this conversation that we
set up, which is called what could Bowling Green be,
which we did with this incredible partner on the ground
called Innovation an Engine. The conversation was like designed to
(15:02):
be like you are free to say what you want,
where you want, whenever you want. And it was a
month long conversation and we did it through this POSE
platform and the idea was speak and you'll be heard.
And the part that we used AI for was making
sense of what everybody said in a way that could
be shared back to them, that they would see their
voices in the conversation and could be shared back with
(15:24):
the policymaker so he could take an action.
Speaker 1 (15:37):
After the break, the largest town hall in US history
and how Jigsaw made it happen. Welcome back to tech stuff.
Before the break, we were talking with Yasmin Green, CEO
(15:59):
of Jigsaw, about how the initial pieces of an ambitious
project using AI to create a more inclusive and active
town hall came together in the town of Bowling Green, Kentucky. So,
taking a few steps back, how does this come about? Like,
how do you choose a community to partner with? And
I guess what are the stakes for you to be
working with government versus just in a purely online arena,
(16:22):
Like how much pressure and also opportunities.
Speaker 2 (16:24):
Does that add? Kentucky's actually a really pioneering state, and
partly a little bit by it because now I spent
much time there, but the cross even not in Bonling Green,
other places do. They're doing a lot of civic tech
work and they're very forward. And Bowling Green in particular
was known because they had done something several years ago
where they had done public participation with the local newspaper. Okay,
and we had heard that they're putting together the stake
(16:44):
vision for twenty fifty.
Speaker 1 (16:46):
So what do we want Bowling Green? Yeah, so you'd
heard about this, you knew that they would take forward.
Speaker 2 (16:50):
Yes, and we're like, maybe we can actually they come
and give them something really useful. Maybe we can co
develop something with them. And so they said, okay, let's
do this month and conversation to hear what people have
to say. And then they actually did like super local
like ground game marketing campaign.
Speaker 1 (17:04):
Where you involved with a marketing campaign or.
Speaker 2 (17:06):
Yeah, we partnered with them, but it was like local
creative agencies that came up with the creators.
Speaker 1 (17:10):
So basically get people to be aware and think it
was fun, evangelize interact with the platform.
Speaker 2 (17:15):
Yeah, yeah, and exactly, And the main thing was just
to be local. So things were translated into nine languages.
They they put them in you know, they fly at
international supermarkets, they went to make a church is, they
went to basketball games.
Speaker 1 (17:28):
So you created awareness and then whether like QR codes
that people like clicked on their phones and they would
take it into the conversational platform or like how what
was the.
Speaker 2 (17:36):
Yeah, basically you look, yes, they're working on code. Do
you go to what could be GB? In fact, you
can stop it. What could Bowling Green be? Is the
website dot com and you can see the report. But
people would log in and then you just get like
a question prompt what do you want for Bowling Green
in the future, Which that's that depended you put your
answers in and.
Speaker 1 (17:55):
Only one single prompt.
Speaker 2 (17:56):
There's one question and could you could put you could
keep putting more of what you want. You probably have lots.
It's pretty open. You might have ideas for education or infrastructure,
you know, restaurants. But then there's voting, so you you
vote on other people's input. And that was a really
important signal to see like where, if anywhere, do we
agree right and where we put so much effort into
trying to make a like make this local and be
(18:18):
make it accessible and still like you know, the beginning
of the month long conversation, we were like, is anyone
is anyone going to turn up to this? Yeah? Because
maybe we do all this effort and it's eight people.
Speaker 1 (18:29):
Did you show a tumble.
Speaker 2 (18:31):
Right? Right? So the baseline's eight okay, So we were like,
thirty will be good. Thirty will be good. Now, so
we're like, if we could get a one thousand or
two thousand, and so it was eight.
Speaker 1 (18:44):
Thousand, eight thousand respondents.
Speaker 2 (18:46):
This kind of like AI enabled town hall was a
thousand times larger than the regular town hall, and one
in every ten residents basically the equivalent of one every
ten residents participated. So as far as like that's what
the folks on the ground and say, is that it
meant that people were discussing this topic even outside of
the actual platform because there was that much participation, So
(19:07):
it kind of pas the tipping point. So it was
really nice to see the numbers go up. They did
go up, they did spike a bit towards the end,
and it was actually the largest digital town hall in
US history. So it's the most that anyone's come together,
and that's US partnering with a town in Kentucky. So
if you like imagine what is possible at the state
level or across states if you bring people together to
(19:28):
share their view. Although there was before like advocating for
more of those type of CIF conversations, I think the
first big apprehension was is anyone could show up? Ye?
The second was, well, you know, if this is a
scaled version of an in person town hall, we might
kind of be in trouble because the eight people that
come to an in person town hall detractors detracted typically, right,
(19:51):
the ambivalent people don't go and show up, or people
who are like quite like the idea and also could
see why not the idea? That's not the people that
get you motivated to get off the counch and drive
all the way and wait there. It's usually people who
really don't want the thing to happen. So I think
local leaders were like, well, we're just going to get
eight thousand of them entrance, you know, because they don't
want to scale in that sense, And that was why
it's so important to do like such a broad marketing
(20:13):
campaign to like all these different constituencies, but the bigger
side of relief from local leaders when they looked at
their feedback and a lot of stuff they anticipated. There
was like in a growing town, you need more infrastructure,
you need more roads. Lots of really interesting thoughts on education,
on like restaurants. There was just one submission on the
restaurant and interesting kind of like unexpected proposals on a
(20:36):
number of fronts, but the one that makes me giggles
in the area of like food and restaurants. One person
submitted an idea to pass a law to mandate restaurants
to include ketchup saches in the takeaway like the takeaway
bags of like fast food restaurants. So you know, that's
how you know that people are really expressing their views.
And we've got good mixed views but didn't have much support.
Speaker 1 (20:56):
How did people express support for one another? Like I
go on the platform, answer this is you know, I
want more free ketchup saches? Is that that Now I
don't live in Boling. I read that it was Geo
Geo fence. Otherwise it would have been me. But then
is that then like available for other people to comment on? Okay,
so it's like a message boarders, it's not just a one.
Speaker 2 (21:16):
So they're routing so you you put your idea and
then they're routing ideas for you to vote on. And
it's actually much You get a much higher percentage of
people like the The votes are much higher than the
actual submissions. Yeah, because it's much easier to vote and
you're kind of in there and it's a little addictive.
And I think that's one of the really really cool
things about this conversation is that it was designed to
(21:39):
advance a very large conversation. It wasn't designed to have
like eight thousand and one on ones. That's not what
we're trying to do when we're trying to have a
long conversation. And that that's interesting because there isn't the
option to reply the way that purchases platform is set up.
You have an idea, and you can vote on other
people's ideas, and if you really don't like the catch
up idea and you actually think it should be mustered,
(21:59):
put your idea in the area. Yeah, but we're not
you're not telling you know, you're not going to get
into some you know, back and forth with the person
who suggested Ki jump is you know, from the wrong
part of town, and everyone's asking that part of town.
You know, here's no point going down that you know,
spiral of the ad hominym attacks, et cetera. That happens
a not on social media. So the conversations dis like
designed differently from the get go that you put your ideas,
(22:21):
you vote, if you want to keep going after you've
given your thumbs up your thumbs down, you're welcome to
put your idea in there. And it well. So then
when we had to take stock of like, well did
any of these ideas they resonate with other people? So
there were four thousand policy proposals and over half of
them had near universal support.
Speaker 1 (22:38):
Near universal support over half. So in other words, what
you take from it is the community knows what's good
for it.
Speaker 2 (22:46):
Well yeah, and and that you know, there are so
many opportunities for us to hear from the people who
are the angriest or at least well informed.
Speaker 1 (22:54):
Or did you know the voter registration of the participants
or I mean.
Speaker 2 (22:57):
Was this no, but it's it's it is a it's
a purpolish place, so it's quite a mix. And you
saw some of the things that you see you know, nationally,
but most of the things that local people care about
are you know, what is going to happen to their
K through twelve education. It turns out they want more
vocational skills, you know, like they want their university, which
is the pride of like you know, Western Kentucky University,
(23:19):
they want it to be more integrated into their economy
and their like workforce development. And there's a lot of
things that they want. And the perspective of not just
dog actually but even like private section leaders was like,
that's great. Now we have now we can go and
lead and build and make changes knowing that we have
broad public support. Otherwise, if you're just based basing you know,
(23:41):
your like sentiment gauging on public calls and social media,
just thinking that like no one wants anything to happen,
you know, or that like you might have some supporters,
but also like everyone else is an avid, you know,
avid distractor. So so I think those are two really
important essons. One people really will show up, they really
do want to shape their future, you know, And two
(24:03):
that there is more that unites us then divides us,
which is kind of imperative for us to keep you know,
in our minds and cultivate at this moment.
Speaker 1 (24:12):
And what was Jigsaw's role or technology deployment because the
marketing campaign was like panel by local agencies, the polished
platform was it's not a Jigsaw platform, right, So what
did what was the layer that you guys added It was.
Speaker 2 (24:27):
The making sense of the conversation. So if you can
get lots of people to contribute, how can you guarantee
that their voice will be heard? And so it really
is using this case, Google's aigemini using an eye for
what it does best, which is they take large amounts
of information that it expressed in a very local and
personal way, understand it, organize it, they sort it, prioritize it,
(24:49):
and then give it back to people. So like, you
know the thing if you go to what could bowlinggreenpeed
dot com, you'll see the report and it's like, you know,
Google Maps, we think at Google mapps, you started very
high level. If you'll look at like the states of
in America, that's the first thing that you see, and
then you're like, actually, I'm kind of interested in New York.
And then you're like interest in my neighborhood and they're interesting.
You're like, oh, how can you see my front door?
(25:10):
You know, And that's how it is with this map
of opinions, like not geographical terrain, but opinion terrain. But
you start the top and you're like, what do people
have born and green care about? And then it has
those you know, instructural education, et cetera. And then you're
driven You're like, actually, what I really care about is
you know, education, and then go through, you know, you
just keep digging and most people, most people have specific
areas of interest. But the truth is people want to
(25:31):
know where they sit relative to their neighbors, and so
people want to actually find where they are. Because there's
this plot graph on there that has a distribution of
all the proposals and how much agreement they got, and
so you can see am I like on the am
I the fringe opinion holder even though like me and
all my two friends agree with me, or actually is
something that I've been scared to say out loud, something
(25:52):
that a lot of other people believe. So you can
go in there and see that, which is really powerful.
And of course most people agreed with most things. So
I think it does restore our sense of you know,
like it's a funhouse of mirror sometimes social media where
you're like, oh my goodness, everyone's extreme. It's just me
here that doesn't want to have like an extreme mistake
on things.
Speaker 1 (26:12):
What are the key policy ideas that have emerged and
will you measure success based on whether or not they're
enacted or are you, in a sense now passing the
bat on the lull.
Speaker 2 (26:22):
It's definitely in the court of their leaders, and it's
their process so to begin with. So I think they
genuinely do want people's input because the process kind of
proceeded our involvement. But one example was they had wanted
to do a riverside development. They built building up the
area along the riverside, and when they had got tried
to get public input, very few people pay and they
(26:44):
were detractors. But then if you look at the feedback
on that one topic in the reports, like seventy five
plus percent of people support it. So I think they're
now going to accelerate that. What we did, in terms
of like really trying to gauge people's sense of like
whether it mattered for them, was we say it on
whether they felt their voice mattered. We asked them whether
they felt they understanded other people's perspectives better than they
(27:06):
did before. And then third was where they think they
believe their input will lead to a better outcoming. On
all of those, it was like eighty percent plus satisfaction.
Speaker 1 (27:15):
And were you surprised by the outcome? And how do
you how do you kind of measure the success of
a project like this.
Speaker 2 (27:23):
I think it was that that people wanted to come
out and they did. I think that there's two There's like,
there's two parts of my brain. There's a part of
brains like did we achieve our you know, like our
product development goals or are like whatever engagement goals? And
then there's an other part of me that's like can
I retain my hope for like society thriving as we
(27:45):
get more at Amize And so I feel like on
both of those I felt really good. I felt good
because obviously the technology wor I also, you know, the
sense of the team is that we can push to
do more. And so now we're exploring what other types
of conversations in the conversation space. How could we Maybe
it's a different scale, you know, maybe it's a different
nature of conversation. Maybe we're using different capabilities of Gemini
(28:08):
to do more to bring people together, to have them
understand each other more and then having it count for them.
So that's the kind of fun and exploration phase that
we're in. Now.
Speaker 1 (28:18):
There's another thing that came out of the research I
think called grounding.
Speaker 2 (28:21):
Yeah, can you talk a bit about rounding. Groundings their
sister to hallucinations, So, which is a funny term that
has widespread adoption to describe when air models may make
stuff up.
Speaker 1 (28:33):
Because of course that the very bad scenario. Who would
be that the air was making up policy of pro
puzzles and no one asked for.
Speaker 2 (28:39):
Oh my goodness, imagine that in such a high stakes
context where people are trying and talk to their policy maker.
And there were some funny things that in our early
testing we were finding it would kind of conflate two
topics that seem like they're about the same thing, or
that like you would it would have only one. There'd
be like one post about you know, there should not
(28:59):
be some any sweet tuitets offered at in kindergarten. And
then the summarizing AI would be like the people of
Bowling Green, you do not want to have like cakes
and chocolates and candy and they would rather have like
keep compisation. It's just like wait, whatever that wasn't said, yos,
that's out of lines. So part of kind of making
sure that we understood the ways that the AI might
(29:21):
misfire was doing a ton of testing with humor reviews
and on previous data sets, and then we did it
kind of kind of like a small private trial a
few months earlier with Bowling Green and the local partners
reviewing them and just seeing like how might it go wrong?
And then some of the tweets we made with to
make sure that we batched appropriately. So grounding is every
(29:41):
output that the AI produces is that it's like appended
with citations. So everything that the AI says was a
theme in the report, you can say, you know, show
me receipts and it will take you to the underlying
policy proposals or words, the actual words of the people
at Bowling Green, so you don't have to you know,
(30:03):
you actually don't have to trust the AI.
Speaker 1 (30:05):
Super interesting. Could grounding be used beyond this context? Do
you think?
Speaker 2 (30:09):
Yeah? Yeah, and it is it's used broadly. Have you
ever used notebook M Yes, So they use grounding. So
when they you know, when they summarize a you can
give it like one hundred reports or whatever backad Democrat.
Of course, if you're don't research and it'll tell you
the themes and then it gives you like like footnotes
so that you can see where the source material and
have confidence that at accurate.
Speaker 1 (30:31):
What's the wider hope with a project like Bowling Green?
I mean, do you is that, like, is there an
ideal community size? Why this is will be perfect for
towns that are growing of around one hundred thousand people
because like you have enough people in the community, you're
invested in the future because it's growing, it's not so
large that you get so many proposals that it's like
impossible to sort. And therefore maybe there's like another one
(30:52):
hundred Bowling Greens in the US, or is it more like, Wow,
the greatest problem in US politics today is that participation
and is low and conversation is toxic, and therefore, like
this could be a much wider solution, Like how do
you between those two things? Like where do you feel
for me?
Speaker 2 (31:09):
The more conversation, the better. I think the reverse is
also true, the less conversation the worst, the worst we
ended up. I understand each other, and that was kind
of where we had taken perspective for over the last
couple of years, was actually not just helping publisher score
things that they might want to see less of, like toxicity,
but actually doing the bridging attributes as well, which was
the things that kind of help people who disagree stay
(31:32):
in the conversation together. So I would like to see
more of it. And I think we've also shown we
show that the appetite is there from people and that
it can go really well. That I think the judge
executive Doug Gorman in our case, was really really brave
to try and do that. Why brave because the safe
nobody shows up, or say if they show up and
you know it's Cook Coffinus. But now seeing what Doug's
(31:56):
that actually Doug came to So Google has a developer
conference equal to I Air. That just happened, and Google
invited Doug to go and participate in the local leaders track,
and he went and told all these like other mayors
and local leaders and they were They really lit up.
And so from what I understand, I think there is
a lot of appetite to do this, and I think
him going first and having the bowling green kind of
(32:17):
proof of concept will inspire others to adopt it to.
Speaker 1 (32:21):
Is that part of your work or in a sense
you like the fire and it's of other people to
kindle it bad metaphor, but I mean otherwise do what
are you going to do the same experiment again in
another community or a larger community? Or do you feel
like you have the proof point that you think you know?
Speaker 2 (32:36):
I think it's you know this question, like what is
Jigsaw's highest and best use. I think if we if
now we've done the concept, if there's another proof like
another place that we'd want to do it, I think
we would be up into that if it felt that
it was some additive like materially additive, if it were
you know, maybe at the state level or you know,
some other kind of complex conversation that we felt that
(32:58):
we could we could advance. I think there will be others,
like kind of implementers who will I think this will
become a kind of a commercial opportunity for others to
help with. I don't think we do you know, that's
the best use of Jigsaw to do many many more
of the same thing in towns.
Speaker 1 (33:17):
So in other words, like a polling firm could take
it over as a business. Liane Audrey Tang, who I
know that you were, you know, to spend time with.
Recently an interview with Nick Thompson and Audrey spoke about
the Bowling Green project and in the context of how
in California post fires, Newsome is sort of using this
(33:38):
tech enabled participatory democratic technology to understand what people want
from the rebuild of like burnt areas of LA. Where
does this fit in with the kind of larger tech
enabled digital democracy trend.
Speaker 2 (33:56):
I think it is. I think they're pioneering there because
they're doing at the state level in California's a massive state.
And interestingly, when we did it in Bowling Green, which
was you know, Well or Warren County, much much smaller,
super open question. In the case of California, that asking
a really narrow question. I mean, it's big in itself,
but it's much more focused question, which is, you know,
(34:16):
how are we going to where do we go from
here after the wildfire specifically, So I think that's really instructive,
Like doesn't once you go to the state level, doesn't
need to be so focused And they've already got quite
a bit of engagement and is she going to talk
to them about how they're going to make sense of
it because I don't know what they do. Well, yeah,
we're in touch with them. And Audrey, who's you know,
it was the former the kind of in agural like
(34:38):
I guess did a digital minister for Taiwan is like
this just she's just a global matchmaker. So she's advising
California as well, and she's advised us a lot. She's
been very instructive for our work.
Speaker 1 (34:50):
Two questions are close, I guess the first. Obviously, it's
a moment where Google is being you know, the subject
of the Justice Department and Chrome. All this kind of stuff,
in other words, is a target of federal investigation when
you work with government, like do you are there people
who say this is a conflict of interest or Jigsaw
(35:10):
being in bed with local estate governments is somehow advancing
Google's corporate motives elsewhere, and that's a problem.
Speaker 2 (35:18):
In this case. We didn't actually have any commercial relationship
with the government. We were actually partnered with the group
the I told you about Innovation Engine and they did
the kind of execution. But we did care a lot
about making the tech work for the hotesty maker and
the people because it seemed it seems like such an
important thing for us to get right. And if we
(35:40):
want policy makers to make good policy, even about tech,
they should kind of understand it and see how it
shows up for them. And in this case, I think
Voluningroom was so attractible was so because it was so
mixed in terms of politics. So, yeah, we've always kind
of been in and we've always been around governments, and
actually still this point you haven't actually directly partnered with
(36:02):
the government. But it felt really gratifying to feel like
you are helping that dynamic work better.
Speaker 1 (36:09):
What do you still have to achieve a Jigsaw and
what kind of future signature projects can you tease for
us Today.
Speaker 2 (36:17):
We're noticing a change in the culture at Jigsaw internally
that actually mirrors what's happening externally. That things are moving
so fast, like the technology is moving so fast, and
we are inside this incredible tech company, and so I
think we sometimes maybe things appear to us before they
mate to others. I think we could help a lot
(36:38):
by kind of sharing as we're going. So I think
the change in the coming year that I'd like to
see is more rapid iteration and more, sharing more frequently,
and being more public about what we're doing.
Speaker 1 (36:50):
Well, you know where to come. Thank you, Thank you
us for tech Stuff. I'm oz Vaaloshian. This episode was
(37:16):
produced by Eliza Dennis and Adriana Topia. It was executive
produced by me Karaen Price and Kate Osborne for Kaleidoscope
and Katrina Norvel for iHeart Podcasts. Jack Insley mixed this episode,
and Kyle Murdoch Rodelphime Song join us on Friday for
the Week in Tech, when we'll run through the tech
headlines you may have missed. Please rate, review, and reach
(37:39):
out to us at tech Stuff podcast at gmail dot com.
TechStuff News
Hosts And Creators
Oz Woloshyn
Karah Preiss
Popular Podcasts
New Heights with Jason & Travis Kelce
Football’s funniest family duo — Jason Kelce of the Philadelphia Eagles and Travis Kelce of the Kansas City Chiefs — team up to provide next-level access to life in the league as it unfolds. The two brothers and Super Bowl champions drop weekly insights about the weekly slate of games and share their INSIDE perspectives on trending NFL news and sports headlines. They also endlessly rag on each other as brothers do, chat the latest in pop culture and welcome some very popular and well-known friends to chat with them. Check out new episodes every Wednesday. Follow New Heights on the Wondery App, YouTube or wherever you get your podcasts. You can listen to new episodes early and ad-free, and get exclusive content on Wondery+. Join Wondery+ in the Wondery App, Apple Podcasts or Spotify. And join our new membership for a unique fan experience by going to the New Heights YouTube channel now!
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!
Fudd Around And Find Out
UConn basketball star Azzi Fudd brings her championship swag to iHeart Women’s Sports with Fudd Around and Find Out, a weekly podcast that takes fans along for the ride as Azzi spends her final year of college trying to reclaim the National Championship and prepare to be a first round WNBA draft pick. Ever wonder what it’s like to be a world-class athlete in the public spotlight while still managing schoolwork, friendships and family time? It’s time to Fudd Around and Find Out!