Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Creating bespoke data visualizations with Alyssa Fowers
140 Plays6 years ago
Alyssa Fowers is a Ph.D. student at the University of Miami where she studies research methods, applied statistics, and data visualization. She writes about graphs (and sometimes swords) at Data & Dragons. Before returning to graduate school, she worked in...
The post Episode #146: Alyssa Fowers appeared first on PolicyViz.
Recommended
Transcript
Introduction of Alyssa Fowers
00:00:11
Speaker
Welcome back to the Policy Vis podcast. I'm your host, John Schwabisch. Welcome back to the show, everyone. I'm really excited for this week's guest, Alyssa Fowers, who is a PhD student at the University of Miami, where she studies research methods, applied statistics, and data visualization with...
00:00:26
Speaker
our good friend Alberto Cairo. Alyssa and I sat down at the tapestry conference back in December to talk about her interest in school and her work in the research methods department and in the communications department and talk about how she visually communicates both large and small numbers and her blog, Date and Dragons, which I've been really enjoying over the last few months.
00:00:48
Speaker
Just a couple of quick announcements before I get to the interview.
Book Giveaway Announcement
00:00:51
Speaker
The first thing, if you're interested in getting some free books, go check out a blog post I wrote a few weeks ago. I've linked to it in the episode notes.
00:01:00
Speaker
I'm giving away some extra books I found in my office while I was cleaning out some bookshelves. I have a couple extra copies of Cole Nussbaum or Naflix book, Storytelling with Data. I have an extra copy of Andy Kirk's book, extra copy of the Big Book of Dashboards, extra copy of John Tukey's book. So if you want to try to win one of those free books, go over to the blog post that I've linked to in the show notes and see what you need to do to win those books.
00:01:27
Speaker
It's been a couple of months, I'm just going to keep extending that contest until we reach the threshold that I can start mailing those books around the world.
Data Visualization Inventors
00:01:35
Speaker
Second announcement is again another blog post I wrote a couple weeks ago on data visualization inventors, founders, and developers. I've been doing some writing lately about different types of graphs and I got to thinking about who are the people who created and invented and developed some of the
00:01:51
Speaker
graphs that we use all the time, not just the line graphs and the pie charts and the area charts, but who created the Venn diagram, for example, that has a capital V in it, or who created the first Sankey diagram or the Gantt chart. So I wrote a blog post and opened up a Google Sheet where I'm asking people to put in some names and links and original articles from people who they think may be the original inventor or creator of some of our favorite data visualization that we use day-to-day.
00:02:20
Speaker
So that blog post is also linked on the show notes page. So go on over there and check it out and see if you can contribute to this open project. And maybe we can have a nice catalog of people who invented some of our favorite graphs. So anyway, on this week's show, I'm really excited to chat with Alyssa Flowers.
Alyssa's Transition to Data Visualization
00:02:37
Speaker
And here is that interview. So do you want to start by talking a little bit about your background and how you got to University of Miami? Sure.
00:02:45
Speaker
So I sort of arrived through a circuitous route, which I think a lot of people do. I got into this by falling backwards through a door and sliding into a new dimension. Seems to be the common thread. Yeah. So my undergraduate degree is actually in social psychology and I was a little bit of a lab rat. I did a lot of research, which meant I had data, which meant I had to analyze it.
00:03:09
Speaker
applied to and was accepted into and was considering going to some PhD programs in social psychology research. I was studying implicit bias and what wound up happening was that I realized that what had me leaping out of bed at two in the morning and running across campus was not that I had a new idea about how implicit bias worked. It was that I had a new idea about a potential interaction effect.
00:03:33
Speaker
in my data. So that got me thinking that maybe I wanted to work in data and analysis maybe more than I wanted to. So you were really more like excited about working and getting your hands dirty in the data. I just wanted to see what was going on. I like exploring, I like making things. So I decided to sort of lean into that curiosity. So from there I actually graduated, moved to Washington DC and worked in a mix of
00:04:02
Speaker
data analysis management, a little bit of teaching, a little bit of database design for about five years. So I worked for a consulting company, as many people who graduate and move to DC do. And then I worked for a nonprofit as sort of their, my title was data manager. Anything that could wind up in a graph was my job. So the very wide range of things.
00:04:28
Speaker
And what I realized, I realized sort of two things. One was that people kept coming up to me and asking me questions about, is this legitimate? Can I say this? How far can we push this? What does this mean? Is this a real difference? And I wound up saying, I don't know a lot. And the second thing was that when
Understanding Data Responsibly
00:04:50
Speaker
I gave people data, people treated it as the hard and fast truth. So it was this, I think Mona talked about this a little bit earlier, the terror of your word being gospel and not really having those checks and balances. And I realized that I just didn't know enough to do the job that I was doing in a responsible way.
00:05:11
Speaker
So I started thinking about going back to graduate school and actually, as fate would have it, saw Alberto Cairo speak at a conference I was at for work and thought, what if I did that also? Right, right, right. So let's interrupt real quick. So the people at the nonprofit were coming to you. Those are like senior researchers.
00:05:32
Speaker
It was, I wouldn't say it was a research team, but it would be like the CEO would come to me and say, yeah. And so they didn't have like a data background. No. Okay. They were very smart, very analytic people who did their own analysis and would go into the data and do stuff and say, wait, can I actually do this? Right, right, right. And so they were asking you to double check, fact check, but asking more, sounds like asking more of the harder analytical questions.
00:05:58
Speaker
Yeah. And we also, we worked with a research company at one point where they did a little bit more in-depth analysis for us. And I wound up doing a lot of the like translation back and forth and that kind of work also. Gotcha.
Research and Communication in Data
00:06:13
Speaker
Okay. So you're doing the data analytics stuff. You're doing some of the biz work. You're doing the communication bridging sort of
00:06:21
Speaker
thing that's amorphous. And then you come down here to Miami. But you're not in the communications department. No. Right. I studied in the program of research, measurement, and evaluation in the School of Education. So that's research methods. That's some applied statistics. So it's
00:06:42
Speaker
both finding things out and then thinking really hard about how we find things out. And I wanted to add in that third piece, which was telling people about the things that we found out, which seems like a very natural fit to me. Yeah, absolutely. And so just quickly on how the program works. So you have, I would guess, your core part of the program in the School of Ed, and then you also have classes in the School of Communication.
00:07:06
Speaker
So, I have classes in the School of Ed, and then I have elective slots that I use in the school community. And then, so the reason that I wanted to chat with you is because you wrote one particular post that caught my eye, but also sort of a back series about getting people to connect with data. Do you want to maybe walk people through the core argument that you're making here?
Space and Time in Data Visualization
00:07:30
Speaker
Sure. Yeah. So the blog post is called Time and Space. It's on my blog, Data and Dragons. The core of the argument is that the space that is spent on figures in a chart, the time that is spent in sort of scrolling through something that's interactive and digital is also a form of encoding that we use these things to communicate priorities
00:07:57
Speaker
as well as communicating just straight information. So I have some examples in the blog post where there's just a blank page.
00:08:08
Speaker
So, I think it's like the Wall Street CEOs from the Great Recession. And it's just a blank page. And so, if you think about what the New York Times can put on a page in print, it's an enormous quantity of information. And to them, having this blank page, having that shock value of they're just nobody, there's nothing, was as important as everything else that they could have put on that page.
00:08:34
Speaker
There are some other examples too where it's sort of the opposite, where rather than having a lot of blank space, you take something and you give it a lot of space. So there is part of the post that includes a visualization about the victims of the Las Vegas shooting. And what that does is it takes, I think it's three numbers.
00:08:57
Speaker
And it expands that into an individual image of each person. So you could say that in three numbers. It's very, very small. People injured, people shot,
Impact of Scrolling Visualizations
00:09:07
Speaker
people killed. That could be a single sentence. It could be a very, very tiny table. And instead, what they've done is they have sort of
00:09:14
Speaker
blown that up into an individual outline, a sort of silhouette for each person. And it's not like the kind of thing that you often see where there's a single very uniform like bathroom sign person. Right. Or like a bar chart or like, yeah, right. Yeah. It looks like people are milling around talking to each other like they would have been just before the massacre. So I think that
00:09:39
Speaker
that's giving space to each of those individual people and making them into people again for the next step. But it's also interesting because you talked about there's sort of two parts here. So there's the print part and then there's the digital part. So I want to come with the digital part because you have made it interesting to comment how our perception, I guess, of the story is in some ways driven by the order in which things are sharp. In the print side,
00:10:05
Speaker
when it comes to the CEO example you talked about earlier is a good one, right? So, we're looking at it here. But like, it's basically, it just says in the Times, I think this is like 10, the CEOs of Wall Street sent to jail, and it's just a blank page. Now, that seems to work. From my perspective, that works in print, but I don't think it would work on the website
00:10:30
Speaker
only because there's like, there's stuff on the top and there's stuff off to the side and then there's ads popping up. And so I wonder if you've thought about like this distinction between this approach print versus online. Yeah, I think that's where the sort of the time piece comes in in the title. Um, so this particular example where it just, it's a blank sheet might not work, but there was one, um, another gun violence post, um, that was extremely striking to me. Um,
00:10:56
Speaker
I believe it was also in the New York Times. It's a graphic of legislation, gun control legislation that's been passed over, I believe, between the Sandy Hook shooting and the Parkland shooting. And it's a series of tiny calendars, and each day is shown in gray. It's month by month.
Online vs Print Data Communication
00:11:15
Speaker
There's the number of mass shootings above each one. There's little notes about what might have been legislation. And I just remember scrolling and scrolling and scrolling, and it's all blank.
00:11:25
Speaker
Every single day is gray because nothing was ever passed. And so it's this really remarkable experience because you keep expecting something to happen and nothing happens. And it's sort of violating the expectation that you have of this calendar. And it's one thing to say,
00:11:42
Speaker
no gun control legislation was actually passed or very minor gun control legislation was actually passed. And another thing to just see where something should be and have nothing be there at all. So you're sort of taking the viewer's time to communicate an absence.
00:11:57
Speaker
Yeah, I guess it is also true that there are pieces like there's this piece from the Washington Post where it's the maps of segregation, which is sort of now famous, but like it is its own thing. It's like a black background and there are no ads. So I guess there are ways in which in which they do that.
00:12:14
Speaker
There was actually another sort of scrolling effect from that post about the Las Vegas massacre, or the article about the Las Vegas massacre. There was an online version of it as well, and they didn't just reproduce the same graphic, which was part of what was so striking about it to me. When you scroll through, at some point you get to where I think this graphic was in the print version, and the screen just goes white. And then a sentence pops up that says, here's the number of people
00:12:42
Speaker
that were injured, you just keep scrolling, here's the number of people that were, you know. So it forces you to sit, to sort of sit with each of those ideas while you're scrolling, because there's this really, really compelling personal narrative going on on either side. To see what happened, you have to put your attention on these things. Yeah, right. It takes your attention and then you have to, you're forced to wait for the next piece. As opposed to that bar chart where it's like, oh, there's three bars.
00:13:06
Speaker
That's what I think is so striking to me about it because in print, the limitation is really, I believe it's space, whereas online, the limit is really attention. You have potentially unlimited space. You can make your website or your article or your post or whatever as big as you want, but you're really limited by how long people want to look at it. It sort of comes back to that thing about values and priorities where whoever was designing that visualization
00:13:32
Speaker
or that scrolling experience decided that it was worth potentially losing people. It was worth spending the time on it, because this is very important. So that was a lot of what was very powerful to me about it. Have you thought about the change in direction, as it were? So most of these sort of scrolling time things are all vertical. And we used to have a lot of steppers that were horizontal. And there only seems to be like the rare occasion where they're kind of blended. Like I remember The Guardian did this story on the Mekong River.
00:14:01
Speaker
a few years ago now probably where, at least in the mobile version, there was a combination of you would scroll down and then you could scroll across and then it would bring you back. I don't want to put you in the spotlight. Have you come across any of those projects or just generally the differences in direction in that way?
00:14:21
Speaker
And maybe it's something that's just different between a mobile platform and a desktop or a laptop where, especially if you don't have a touchscreen, it's just a different interaction, right?
00:14:32
Speaker
Yeah, I think that part of my guess would be that the move towards having things be vertical scrolling is because people are looking at things on mobile. From my own experience, working with spreadsheets, I don't like going sideways, so I have an instinctive desire to not go sideways. I think if something is stepping across a screen and I don't have to actually scroll to see it,
00:14:57
Speaker
where it's a physical step or the chapter or the page term. It's less aversive to me than having to scroll right further because I'm just like, oh my god, why isn't this in a database? Why isn't this long for that? Which lets you know how my brain works a little bit.
00:15:12
Speaker
I think also that I'm going to take back what I said. There is a little bit of more space limitation when you look side to side. Yeah, absolutely. Just because of how people are used to consuming information digitally. If you're reading a document, you're scrolling down. Yeah, not scrolling across. Unless, like you said, you're in like an e-reader or something where you have this vision of turning pages. But a map, for example. I mean, we obviously looked north-south, but also east-west. And so if the story was
00:15:39
Speaker
to go from west to east like the uh the washington post in that story on the eclipse and i'm trying to remember how they set it up i think it was his own sort of thing but i don't remember it going like
00:15:53
Speaker
where it was showing the path of the eclipse across the country, but I don't remember swiping or something like left and right. My guess if I remember it is that you just zoom. It's not like you have to go in a specific order. Well, anyway, so that's sort of on the side because that's interesting, but I want to get back to this
00:16:13
Speaker
this idea of using the silhouettes of the people.
Using Icons in Visualizations
00:16:15
Speaker
Because in this particular example of the Vegas shooting, it's silhouettes of people, it's not dots, it's not bars, it's people. And so I guess that's just a broader question about using icons in data visualization and your thoughts on how that helps a reader or hinders a reader and how it's useful for data visualization creators and also potentially a challenge for creators.
00:16:41
Speaker
I think that an icon communicates a lot. One of the things that is really interesting to me about icons is how important the level of detail is.
00:16:56
Speaker
in this Las Vegas massacre image, they're clearly people, they're sort of leaning, they're grouping in kind of organic ways, so they're very clearly people, but it's really hard to tell what gender they might be, you can't tell what race anybody is, you really can't tell anything about them except that they're probably all adult people.
00:17:16
Speaker
different levels of detail that communicate different things. So like I mentioned, this sort of bathroom sign, icon, all that says is person. It's pretty generic. It's instantly recognizable. People aren't going to spend a lot of time looking at it and going, what is this? Or absorbing the details. But as soon as you put more detail into that, the more specific you make an image, the more specifically people are going to read it. So if you take that sort of generic human
00:17:46
Speaker
outline and you make it more gendered or you make it into a child or something like that, people are going to assume that what you're showing is sort of relevant to everything included in your visualization. And as you sort of add more detail, the message that you're sending gets more and more specific.
00:18:09
Speaker
Because you can get, of course, all the way down to a photograph of someone. But then it's extremely specifically that person. So I think that it's something
00:18:18
Speaker
to be aware of where the less you tell you have, the faster it is to process, but the less specific it's going to be and the less sort of immediate resonance it's going to have with the people who are looking at it. So that the gender icons, for example, like the bathroom icons, there's like basically the icon, which we all sort of recognize now as male, only because I think the female icon, the sort of standard one has like a triangle dress sort of thing, right? And so that directs us in the two
00:18:48
Speaker
The two directions, but we also have like gender is a is a spectrum and so that adds an additional and races You know, you know people can be very have various backgrounds So like, you know, how are we as designers or maybe better question actually I said you as a designer like communicating this like how do you start thinking about this? I mean, I don't there's I don't think there's a right answer an answer to this really but
00:19:14
Speaker
You're clearly thinking about it, so I'm curious, you know, if someone were to bring a graph to you and to say, you know, I've got income for men and income for women and I want to use icons, like how do you start thinking about the limitations of that?
00:19:29
Speaker
So I think that in a case where you really have this categorical information in your data, and it's really important to the analysis, it makes sense to show it as these two categories. But I think it's also important to recognize and acknowledge that you are going to have non-binary people in a lot of data sets, and that not all women look a certain way, not all men look a certain way, not all non-binary people look a certain way.
00:19:56
Speaker
And I don't have a great answer for how do we have imagery that is instantly recognizable that appreciates that gender is not necessarily this hard and fast binary. But I think it's really important to keep it in mind and to just be mindful about what you're doing while you're doing this work.
00:20:21
Speaker
And if there are non-binary people in your data set, don't forget about them. If they're there, make sure they're sort of visible. Yeah, I mean, it's difficult, right? Because there certainly, almost certainly are non-binary people in a lot of the data sets we use. But they're either, you know, for whatever reason, they're not identified. Yeah. And so I think the general message, right, is to just be aware, or try to be aware, which is what your implicit bias is.
00:20:50
Speaker
Try to be aware of these things, even though it doesn't show up in the data necessarily. It's like an explicit tag. Yeah, yeah. Yeah, so I should probably clarify a little bit. I do mean that one should always be aware of it, but if in your data set you have this explicit tag, I do sometimes see people saying, well, we have non-binary people in this data set, but it's like 1% of the data. So we're just going to exclude them from everything. Right, right.
00:21:16
Speaker
So that's the kind of thing that is important to me to not just leave out or exclude. And yeah, that question of how do we do iconography when we don't have discrete groups is really hard. It's really hard. I don't have a perfect answer for it.
Second Order Meta-Analysis
00:21:33
Speaker
Can you tell us a little bit about your research? So you've got this interesting mix, which is like, I don't know, it's like the Renaissance person. I don't know, it seems like it's the evaluation side, but also then sort of the communications and data-based side. So can you talk a little bit about, I mean, a PhD is a long thing. But yeah, so what is your research about and how are you blending all this together? So right now, I'm working with my advisor,
00:22:00
Speaker
on second order meta-analysis, which is an abstraction of an abstraction, basically.
00:22:12
Speaker
A meta-analysis is a synthesis of lots of different effect sizes about the same phenomenon synthesized into one. So it's sort of averaging across a bunch of different studies to say, we have looked at a lot of research about this issue. If we combine everything together, here is our estimate of how big this is or what this relationship is. So first order meta-analysis, which is what most people think of, is summarizing primary research. So you have
00:22:41
Speaker
an individual study and then where you compare one group of people to another group of people and then you get a bunch of those and that's a first order meta-analysis. I'm studying second order meta-analysis at the moment, which is you're doing a meta-analysis of meta-analyses. Gotcha. So you have a hundred studies, a hundred point estimates of some relationship between x and y and ten people have done meta-analyses of those hundred studies and so you are doing the meta-analysis of the ten studies.
00:23:10
Speaker
Yes. In a particular area? I'm studying it from a methodological perspective. Right now what I'm doing is a simulation study. I make up data to study an abstraction of an abstraction, which is the most academia thing I think I've ever said. I go through all that to say that this has been a case study in how hard it is to talk to people about statistics. Oh, okay. Interesting. I'm trying to write a grant application or something, an article about this.
00:23:40
Speaker
It takes eight words just to say the two things that I'm talking about in people's eyes at the third word. So that's some of the research on the statistics side that I'm doing. I have an interest though.
00:23:55
Speaker
sort of more generally in this translation piece. I think that data can seem like a foreign country full of incredibly perfect hard truths that if you throw a lot of money at trying to get at it, you can come back with an ironclad path of action that will undoubtedly result in success. You can come back with pure hard truth.
00:24:15
Speaker
And having worked with data in a lot of different ways, sort of having my hands in databases, seeing how it's collected, seeing how it's analyzed, that's not how it works. Data is intensely human. And so to me, being able to communicate, being able to sort of accurately communicate what the data is actually saying, being able to
00:24:45
Speaker
being able to sort of translate it for people is immensely important. So you've got this methodological piece of your work here that is doing this second order meta analysis. But really at the core of what you, it sounds like what the core of what you really drives you in some ways is trying to take these hundred estimates and trying to communicate that to people
00:25:09
Speaker
even though there's various levels of uncertainty and distributions around within each study and then within, you know, across all of these studies. So is the sort of bottom line goal, I guess, to try to figure out a way to communicate relationships between variable A and variable B to policymakers and stakeholders and decision makers? Is that like where you're trying to get to?
00:25:37
Speaker
Um, I'm going to give you again, the most academic answer possible, uh, which is, well, never sitting at the university. We should have done this like at a coffee shop. I would have got a librarian. Um, right. So that is, that's one of the things that I really want to do. And that's very important to me. Um, I think that with my work here at the university, I'm very much a both and person. Um, I want to be able to, um,
00:26:05
Speaker
help people do this research in sort of the best way possible while also having this, being able to communicate it out. I am also really interested in applied problems in sort of getting into more actual data and doing analysis on stuff that actually sort of comes from people and is applicable to sort of more hands-on situations. I do think that part of doing a good analysis
00:26:33
Speaker
is being able to tell people
Making Data Comprehensible
00:26:35
Speaker
about it. Because if you do the most brilliant and cutting and insightful analysis in the world and nobody knows what you're talking about, it's the tree falling in the forest. If your analysis falls in a desk drawer, did you ever really do it? Absolutely. Hallelujah. Yeah. So for me, it's this combination of wanting to make sure that data is handled in a responsible way and also making sure that
00:27:03
Speaker
It's understood. And these things very much go hand in hand for me. And what's also very important, and I'm glad you mentioned it, is making sure that it's understood by the people who are going to use it. Not just having other academics or other visualization people understand what I'm doing, but making it more open to people. Which is one of the really big challenges, I think.
00:27:29
Speaker
As someone who really likes making things and experimenting and sort of throwing things on the wall and seeing what sticks and not falling, leaping down rabbit holes, it can be hard to sort of reel it back to what actually makes sense to other people. Yeah. Well, I think you've cut out a pretty big slice of that.
00:27:49
Speaker
graduate work pie. Well, this is great. Thanks and good luck. Thank you. I'll put the posts on the show notes so people can check them out and I'll put some pictures on the site. Great. Thanks a lot. Thank you.
00:28:09
Speaker
So thanks so much for listening to this week's episode. I hope you learned a lot. I really do hope you'll check out Alyssa's blog, Data and Dragons. She's writing about some really interesting things and some interesting observations, especially at the crossroads of data visualization and research methods and statistics. So please do check out the show notes. There's some good resources there, some old blog posts and some other things that I think you'll find interesting. So until next time, this has been the policy of this podcast. Thanks so much for listening.