Interview with Irene de la Torre Arenas
How does data visualization work for journalism?
What can we learn from it as statisticians?
Today, we will be talking about data visualization in journalism with Irene. She has worked for BBC so she has a background with journalism and we will learn from her perspective.
We talk about the following interesting points:
- Who is your audience? Will they understand it?
- Is it accessible e.g to color blind people?
- What is your message?
- Use charts
- Curating information
- Include aggregated data
- Use a specific person as an example
- Data transparency to build trust
- Engaging data visualizations
Irene de la Torre Arenas
She is a data designer specialized in visualizing information in understandable and meaningful ways. She works as a Visualization Lead at UCB since February 2021. Before her current role, she worked at BBC News, where she was part of the UX&Design and the Data Journalism teams, and at the MIT SENSEable City Lab. She has also collaborated with other media and advertising organizations in Spain.
In 2017, she gave Data Visualization classes at the MIT School of Architecture. One year later, she designed the didactic program Communicating with Data that MIT developed for the Dubai Institute of Design and Innovation.
She graduated from the MFA Information Design and Visual Communication at Northeastern University. Her work has been recognized in the Kantar Information is Beautiful Awards and the WAN-IFRA European Digital Media Award.
Listen to this episode and share this with your friends and colleagues!
Alexander Schacht: You’re listening to the effective statistician podcast, a weekly podcast with Alexander Schacht and your MP Scott and Sam Gardner designed to help you reach your potential lead great science and serve patients without becoming overwhelmed by work. Today, I’m talking again with Irene about what we can learn from journalism. So stay tuned and now some music.
Today, we are talking again about visualization and says lots of content on the homepage for visualization. So check out the homepage under the effectivestatistician.com and find many more helpful resources in terms of improving your data visualization skills, and producing this podcast in association with PSI. A community dedicated to leading and promoting the use of statistics was in the healthcare industry for the benefit of patients. Join PSI today to further develop your statistical capabilities with access to the ever growing video on demand content library, free registration to all psi webinars and much, much more, head over to psiweb.org. To learn more about PSI activities and become a PSI member today.
Welcome to another episode. Today we are talking with Irene dela Torre Arenas about data visualization again, but more in a specific context. And she has worked for the BBC for some time, and has actually a background in journalism. It’s great to talk from that perspective, and what we can learn about it. So if you don’t know who your Irene is then scroll back a couple of episodes, we already discussed a couple of different topics and had lots of interesting insights on data visualization. So welcome again.
Irene: Hello, thank you so much for inviting me again.
Alexander: Well, great guests are always great to have a couple of times.
Irene: Thank you.
Alexander: Okay, so what’s the number one learning set we should take from journalism?
Irene: I really think about your audience, because that touches different points. The first one is, what you’re doing is accessible, readable, understandable. Sometimes when we do data visualization, and this happens to me a lot, you understand everything, right? Because you’ve done it. And then, you show it to people, you show it to colleagues. And even though they probably know what you’ve been doing, they don’t understand much of what is happening there. So it’s important to think about your audience, the person who is going to be on the other side, is she or he going to understand information? Is what you want to say coming across in the correct way? Because sometimes, actually, that doesn’t really happen. So I would say that the most important thing come in a second point. Is it accessible? So if it’s a static graph, is it readable? If we are using a specific color, is it readable, for people with color blindness, which is something that is very important in journalism, I was aware of it already because there are more colorblind people than we think and I had a couple of bosses who were color blind. And then at the BBC, being a public company, it was requested it was everything had to be accessible. So I learned that it’s really important, it’s basic that absolutely everyone can read it and this is just like if you’re walking in the street and you’re using a wheelchair that you’re still able to walk around it is the easiest basic, right? So check that our colors are understandable so that people can differentiate them because that will make the chart much better for absolutely everyone not only for people who are colorblind.
Alexander: In terms of that. I think the first is of course, in these cases, you need to choose your color yourself. So what tools do you recommend for people to use to choose colors, or color schemes.
Irene: I use a couple of tools I started with Adobe Color Wheel just to check and see. It already has included an accessibility option where you can check how it works. And if it’s accessible. And from there, I started using other tools. One is, I think it’s a BS palette, created by Susie and Elijah Meeks. They did it for Netflix, they worked for the visualization team at Netflix. And basically what you do is you copy your colors, and then you can see how they work in the shapes because sometimes even when we see all the colors, right, let’s think about, for example, blue and gray or blue and purple when they are displayed in very small shapes, or lines. They actually look the same.
Alexander: Yeah. If you have very thin lines, yeah.
Irene: So what this tool allows you is to see how these colors look in shapes. If they have enough contrast, then you can see it in the different colored lines. So basically, what you do is you start with your palette, in, for example, Adobe Color Wheel, and then you just run it with a different website.
Alexander: What I really enjoyed using is Adobe Capture, it’s an app you can install on your phone, and with that you can kind of point to different backgrounds, let’s say in your garden, and then it will pick up a color scheme from your garden. So you know, the greens and you know, maybe if you have red flowers you will take up some red and create a color palette from there. Or if you pointed let’s say to a more industrial area yet we’ll have more of the blue and gray and dark colors or if you pointed let’s say to your kid’s room, it will take up you know all these different colors, depending whether you have boys or girls. Yeah, different color palettes will come up. But that’s really fun to play with. And so you can capture different kinds of moods. So to say, yeah, depending on where you go.
Irene: Yeah, that tool is included in the Adobe Color Wheel at their website. So you can upload it and then you can say, I want a lot of contrast. So it picks up the strongest colors or muted or if you want more or less the same shades, etc. It’s very interesting when you use images from movies, because the color is used to create a specific feeling. So you start seeing the blues and the oranges working together, or blues and purples. It’s really fun.
Alexander: Yeah, that’s cool. So you can upload some kind of pictures and some you can create kind of more like, maybe 20s style kind of color, or maybe an 80s color scheme or something like this. Okay, what’s the next point that you think is really important to learn from journalism?
Irene: While there are two things, there are several things but the one that comes to mind is kind of creating a story, which is speaking is related to what we were saying before, or what is the message that you want to come across, right?
Of course, this really depends on the context where you’re displaying your visualization is not the same. If it’s, let’s say, a poster in a conference. It’s grabbing just a scientific journal or if it’s an add on in a scientific journal, where it’s like a video or a visualization, that interactive visualization that is supporting that paper, right. But really think about what is the message, structure everything around it and create a story also in it and also think about it again, is it understandable? Are you making it as easy to understand as possible in journalism? There’s this thing of thinking of the audience as people who don’t have time. So the message has to come across really quickly. And you need to have labels, your title needs to be very descriptive, then you repeat the information in the subtitle, you have a legend, but you also have a notation so that everything is repeated and the information is just really clear.
Alexander: In terms of descriptive title or telling title. Say it often says pushback, well, doesn’t fit data need to speak for itself. What do you think about some pushback?
Irene: I understand the point. But sometimes, comparison between drag x and y is not telling me what the graph is showing comparison of exactly what is it performance? Is it the speed? Is it affecting a specific indicator in the patient? That is what I mean by descriptive is not enough to say comparison of x and y. That is not really telling me anything.
Alexander: Okay, so But if I say, let’s say response rates in question, explain that. of truck A versus truck B. Is that descriptive? Is that sufficient?
Irene: I had the same conversation last week. I guess it really depends on the context again. I come in from journalism again. It was really strange to me, Or it was something that I learned that you cannot say the performance of drug A is X percent better than drug B, right? Which is something that you do in imedia. That’s because that is being impartial, you’re not doing anything, it’s just a number. And it is what the data is showing, right. But in this context, I learned that you cannot do that. And it was kind of striking to me. That’s why I say it really depends on the context. If you cannot do that, because that’s not expected, then try to really be as descriptive as possible.
Alexander: I like in these contexts to ask a specific question. So that’s the visualization that gives you a specific answer. Yeah. So if you say, how much better is Truck A versus B, in terms of response rates? Then I think, see, until you can show a bar chart, and you know, you can see, okay, there’s a difference is that say 10%? Yeah. And then you can also have your inferential statistics around that confidence interval whatsoever. Yeah. And I think it also tells a nice story, because first asking a question creates tension, and then the visualization basically closes the loop. Yeah.
Irene: Yeah. that’s exactly the strategy
Alexander: That’s it. Yeah. I think that that is really kind of nobody, you know, not even a compliance organization can you know, say something against that? So if you ask a question in a, in a, in a title, and that is exactly the kind of successions that the visualization is supposed to, to answer. Yeah. So I think that’s completely fair.
Irene: Yeah. It’s also me, when I was preparing our workshop I went back to a research from a former professor of mine, Michelle Barkin. And she studies how people read charts, how they remember them, what makes charts memorable. And in the first, in the first research that she did, they basically showed charts for like 10 seconds to two people, and then they asked them what they remembered. And of course, they remember infographics, anything that had icons, photographs, etc, for very traditional charts. They were not very memorable, I think, because they just look very similar. So they didn’t remember the topic. And then for the next one, they showed the chart for I think, 30 seconds or one second, or one minute. And what he showed is that if you include labels, if you include annotations, if you include a really good title and subtitle Then those chairs that were not very memorable before, become very efficient. And people remember what they are about. And people remember what the message is? So I think titles are more important in any context than what we think of. For sure.
Alexander: Yeah. And if you use even a bar chart you can create or make also more accessible and more understandable if you can afford it, for example, for every, instead of having a bar, yeah, you could, let’s say, icons of people on top of each other, or next to each other. Yeah, so you can directly see, let’s say each person stands for 10%. Yeah. And you have five persons for drug x and six persons for drug y you have your direct seal was 50 to 60 comparison. Yeah, it’s something that is, you know, much more easy than to remember and you know five centimeters for six centimeters
Irene: Yeah, I also like using slope charts to highlight increase or decrease between two variables, I think, because you do basically connect to two lollipops, with a line and then you see if the line is going up or down. So it’s also highlighting that message of difference. So that’s something that I really like. And I haven’t seen many slopecharts in pharma yet.
Alexander: Yeah. Yep. Slope charts are so basically that, if you have, you know, basically two time points. Yeah. Start and End. before and after. And you have two groups or three groups or four groups, Yeah. And you just have on the horizontal axis, the time and on the vertical axis response rate, or the mean score, or whatever. Yeah, I think that’s really nice. And you can also have the confidence bands next to each other, or maybe you just chose the difference to placebo. Yeah If you have multiple groups or multiple treatments, that way, it becomes really nice and accessible
Next point on the list that we discussed before was creating information. I think this is a really good point. What does it mean for you?
Irene: For me, creating information is really thinking again, about the message and including only those things that are relevant for the message and the information and not because sometimes we want to add so much in our graph. And I understand we want to kind of highlight many things at the same time, but there’s a limit to what you can do. And sometimes if you include too much you understand a lot from it, well at least in my opinion.
Alexander: Yeah. Yeah. And this gets very easily cluttered. Yeah, yeah. And if you go back to the audience that you talked about earlier, if you think of an audience that has little time, too much information is distracting. And then your main point doesn’t come across.
Irene: Again, this really depends on the context, I think, if you need to show a specific result really quickly, then be careful with how much information you’re including right? Then if it’s something that you should read with more time, that is more carefully done, etc, then you can still show everything but slowly, like, create, create a story around it and build everything around it slowly and then step by step. So that is understandable.
Alexander: And actually, data visualization already gives us the opportunity to show much more data in a smaller time frame compared to a table. So you can show the richness of the data in terms of, let’s say, if you think about a scatter plot, or something like this, Yeah. You already show lots of lots of data points, but very fast or if you show lines over many visits?
you should actually show lots of data points. Yeah. But in a, you get some main points at a glance. And that makes it so fast. I think coration is also a service to the audience, in terms of set, you make sure that they get some main thing, and some minimum of time. And you basically serve your audience. Of course, on the other side, a push back set, that you become selective, or let’s say biased in terms of what you present.
Irene: And that happens a lot with time series data, right? Like, Where do you cut it? etc. I think for example sometimes these might happen when there’s this discussion around, should you show overall results or aggregated results or patient level data, right? If you have a clinical trial with hundreds of people, for example, and you want to use a heat map, depending on the size of the graph, it might not be very understandable, but you’re still seeing all the information at the same time. And maybe you’re even displaying three different variables and attributes at the same time. But is it? Is it easy to understand, is it fast at telling you what is going on? whereas an aggregated graph, it might be showing less information, but I suppose, is it much efficient, not telling you a specific thing?
Alexander: Yeah. Yeah. And I think that it’s different layers. Yeah. Like you all, I also haven’t done journalism. Yeah, you have the title, which is the most condensed part, yeah, then you have maybe, you know, a short, very short paragraph, just three sentences below it, which is maybe your executive summary. And then it becomes, you know, more and more broader. And I think that kind of layering, and filtering, or information set way, also are some things that you can very easily do with data visualization, in terms of set, you have maybe the most important results directly available. And certainly you can filter down and click down in an interactive graph and show the subgroups of maybe, you know, the patients and additional sensitivity analysis and all kinds of other things. But at least you get some main point across very fast.
Irene: Yeah, there are different theories around these, I think one of the most important ones, I don’t remember his name, I think it’s Ben Schniderman, but I will check it out later.
He says that usually visualizations do either start from a very general view of the data, and then you slowly go into the individuality of it. So using means or using the opposite. So you start from the individual and then you go to the general aspect. And I think this usually touches also with the structure of journalism, and how do you start the story?
Do you start by using a person as an example? And then you kind of go on to speak about what is happening to a more general group, but then you have kind of connected the person to the protagonist.
And then you also have the other side of stories that start with the general impact that is something in the size of the population, and then you find people who are examples of this situation, right? In databases, exactly the same, Either you’re starting with a general view, and then use the mean or you do the opposite and you start with the particular and then you sum it out, but it needs to have it cannot be you go from an individual then to general into an individual than a general, because it becomes a little bit confusing.
Alexander: Yeah. So that is basically deductive versus inductive, in terms of how you approach things. And I think generally in clinical research, we have this inductive approach, yes, this kind of bottom up approach.
And so I think it’s great to from time to time, use the other one around and start with the answer, and then show how you got to it. Rather than, you know, start with his baseline characteristics here, see, patient disposition, and so on. And by the time we actually get to the results, nobody is listening anymore. Yep. Another point that we had talked about in the prep is results versus process.
Irene: Well, this also touches on that point, as you were saying before, from bottom up, right, it’s just similar I think. And if from a journalism point of view, sometimes you also want to speak about the process of how you came to those results, just your strength and that your conclusion is not biased, that it’s coming from my specific workflow process of thinking etc. I also like it here, for example, when I think about the process, I’m probably thinking of all those patients participating in clinical trials who might want to know, what is the process of launching a drug? What is the process of how the industry is taking decisions? Or how governments are taking decisions? That is basically transparency? And that is, that’s what it means to me at least
Alexander: Yeah. I think we can do more transparency in the pharmaceutical industry as well. say is, and here also, data visualization plays another big role, because if we can show much more kinds of individual patient levels, which you can really only find difficult. Yeah. It cannot really show in tables. Yeah. So if you, for example, think about clinical trials, about coughing. It’s all tables, one table offers. Yes. It’s all aggregated data. I think it would be much more accessible, much more transparent. If we would have more data visualizations based on set. And if there would be more kind of interactive data visualization, if there would be data that shows the individual patients levels. Now, of course, there’s some kind of restrictions in terms of privacy, especially when it comes to more rare diseases and things like that. But for lots of studies, we could be much more transparent. And I think that would also help with building trust. Because I think that is one of the great things we can also learn from journalism. If you’re building trust, if you’re building a kind of fear showing, you know, a good story, if you’re kind of knowledgeable about it, if you have done your own work, right? Then, you can build a lot of trust. And that helps you to get your message across. And if you don’t do that, and you lose the trust, it takes quite a long time to get it back. And I think for newspapers or for any kind of media company, it’s nearly fatal to lose us.
Irene: Yeah. it all depends. You know, journalism is credibility. People go to you because they trust that what you’re saying it’s right and it’s done. It has a specific quality. And it’s very obvious in the media, but I think it happens to absolutely every company. We buy because we trust that the company is following specific quality control processes, and of course in industries related to our health, that is even more important, right?
We want that, The drugs happen in a specific process. That is they are checked that there’s Control, etc. So having that transparency out there of how that medication came to be and how it happened, or how it was put in the market, I think it’s extremely relevant, both for the company showing that they appreciate that patients want to know more about it, and also for the patients who are on for the doctors who are taking those decisions.
Alexander: Yeah, I completely agree. Last bullet points that we had on the list. And that is something that we can actually learn quite a lot from journalism is about engaging data visualization. Can you give me an example of what sets our engagement data visualization for example? What does it look like?
Irene: I think it looks in different ways, it looks like different shapes depending on context. You can have a very static graph and it is still very engaging, because it raises questions, it makes you wonder what this is about, it makes you want to know more that I would say that is very engaging. And that is a very effective graph. Because it’s making you do more actions, request more information, etc. And then you have more interactive visualizations that basically what they do is that they are so interesting and they look so amazing that they make you stay with them for a really long time. And, and that is what they engage in visualization really wants. It wants you to stay with it playing around and getting the information for a long time.
Alexander: Yeah. So the first kind of you needs to grab your attention. And then you need to keep your attention. I think there’s also this concept of a scrolly telling set. I really, really love it, it’s becoming more and more popular, especially on mobile devices. They kind of scroll down and then the graph is changing. And you know, the small blurb is coming up that tells you what’s happening. And then kind of you want to further scroll and scroll and see how the story is continuous?
Said I absolutely love these kinds of things. And it would be great to have these kinds of things much more.
Irene: Yeah, I think scrolly telling what happened. Because before I think early, in the beginning of this decade, well, 10 years ago, basically.
We had many interactions in the news where users were asked to click interact with things and then you were expecting them to get the information by themselves. What happened is suddenly, especially with phones, they realized that the people were not clicking, it’s just too much.
So it’s really scrolly telling what it allows you to have that interactivity, that feeling that you’re editing the story by yourself as a user. But that’s actually a false impression, because it’s already created and is already a structure of the journalist. And the team has already created the story for you. And this slowly, is like an onion. So you scroll and there are different layers of information. And you only need to scroll, you don’t need to click. It is already created there for you. So it’s a great mix between interactivity. But it’s still a very readable story so that you don’t lose interest.
Alexander: Yeah, I think that’s really interesting. I never thought that it was coming because people weren’t clicking. Yeah, I can see that you notice a scroll is kind of what you do anyway, if you scroll down my page. And so it’s, it doesn’t require you to do something different to change. And it speaks to these very little hurdles. Set you need to kind of think about to make it very easy for your audience or interact with the data. I’m just thinking of another example I was looking at once, the very early version of Sir John Hopkins dashboard. And I couldn’t see that I could click on the individual countries to show me the by country data. Because these countries don’t have some kind of, you know, say little tap around it or this kind of button like view that suggests that you can actually interact with it. And only by chance I clicked on one of these, and so “Oh!”, actually there’s an interaction here, and really thinking about these small hurdles can make a big difference in how you engage with your audience.
Irene: Well, that’s user-ex design. That’s if you speak about that, with user experience design there, that person will tell you well, that’s really bad design. Because if something is clickable, it should look clickable. This is also something I forgot to mention when we were speaking about accessibility, right? If you’re doing a visualization, everything and you’re including interaction, it should look like there’s interaction in it. So buttons should look like buttons. sliders should look like sliders.
And there should be, like, user-ex design, which sounds really flashy, depending on where you are. It’s really important because it’s what makes a visualization usable.
Irene: Yeah, I usually distrust websites that have text that is very small for me to read, that means that they don’t want me to read something. It’s like when you buy something like food and suddenly you check the ingredients, and they’re very small, right? It’s very similar.
I, this is something that I learned at BBC, that text sizes are also accessible. Like there are some people who, and you should always give the power to the user to increase the font sizes or diminish them and make them smaller. And I was aware of it in a more kind of intuitive way. But then there are rules, right? So you couldn’t go beyond 12 points in an interactive, or the graphics have to have a specific font size so that they are readable in mobile, etc. So it’s just a tiny detail that makes a huge change.
Alexander: Yeah. Thanks so much, Irene. That was an awesome discussion, we talked about accessibility, creating a story, and curating the information. We talked about, you know, great titles and telling titles, questions as titles, and about different approaches on how we can structure all our data visualization, kind of inductive or deductive.
We talked about how we can make graphics more engaging, So that it’s really memorable, and it is also something that people want to play with. Want to learn more about it. Is there any one point that you would say is the most important thing that you learned from your time in journalism?
Irene: I think that for me the key point was to think about your audience. That’s really because it touches everything else. At the end of the day and there’s another point that I forgot to mention which is reusability
and speed, which is extremely important there. Think about ways of making charts reusable modules, reducible, having templates, make charts fast, so that you can then you can spend more time with the fancy and interesting things too.
Alexander: Yeah, and I was really amazed about the presentation from the financial times data visualization people at the PSI conference in 2021 the keynotes where one designer said, did he kind of had only seven minutes or something like this to produce the graph for the front page, for the cover page and it kinda blew my mind. Oh my God, you need to have really good systems so that you can produce high-quality data visualization at that time. But you know, generally they wouldn’t spend more than a day. For it was not because the news is already known. In Pharma we spend weeks and months on it. So I think there’s just a whole lot of room for improvement. So to speak.
Irene: In the data journalism team at the BBC. they created their plot theme, just so that they could. So, usually the owners, which is a statistical office in the UK. They publish data. I don’t know, 8 a.m. For example, there’s a huge competition and the beauty of it of course, is this pressure of being the first publishing the results, Etc. So they created the scripts, prepared them before and then the day the data came. They really had the scripts in our ggplot theme, Etc. So that in less than 10 minutes, they could publish the article. So yeah, that I don’t think that happens in Pharma.
Alexander: Well, but of course in Pharma, it’s something similar if you have, for example, your database look, that you have a big readout, let’s say on the 20th of June next year. You know that well, maybe it shifts a little bit forward or backwards but, you already know what data to expect. So you can prepare for lots of these things in advance. So you just kind of plug them in at times. Of course, then we always need to. Let’s fine-tune a little bit because I think that’s another point. You always need to fine-tune data visualizations. You never kind of completely have them pre-programmed, because there’s always these small things that you need to change or there’s a difference that is so big or small. So you need to adjust something or some labels don’t fit or whatsoever. Something’s always happening. But if it’s already, let’s say 90% there. Then you have the time and the processes to actually get to the end. Yeah, anytime. Thanks so much Irene. Thanks for a very great discussion again. And yeah, there’s surely much more coming in terms of data visualization, there’s also the special interest group on data visualization that we are both working on and just check out the PSI homepage to find that there is much more content related to data visualizations there. And surely more things are coming at the next conferences, as well. I think data visualization was a huge topic at the 2021 PSI conference and the growing number of Content. Also on video on demand Content Library from PSI. So check that out. And you can learn much more and apply what we just talked about.
This show was created in association with PSI. Thanks to Reine who helps with the show in the background and thank you for listening. Head over to theeffectivestatistician.com to find more resources about data visualization and much, much more to boost your career. As a statistician in the health sector, reach a potential, lead great science and serve patients, Just be an Effective Statistician.
Never miss an episode of The Effective Statistician
Join hundreds of your peers and subscribe to get our latest updates by email!