Representing Patterns
COMMUNICATION DESIGN STUDIO: PROJECT THREE | PROCESS
Project Introduction
This post documents the third project for the Communication Design Studio taught by Stacie Rohrbach and Vicki Crowley at CMU. Representing patterns challenges designers to critically develop stories based on factual data. The objective is to realize that many different stories can be told with the same sets of data, therefore, the designer’s decisions about which story to tell and what information to highlight will shape a reader’s understanding of that data. I have been assigned the topic of AI and have been given some datasets related to that topic that I am tasked with shaping into a narrative. I will think critically about how readers and viewers will engage with that data as they read or see the story I shape with it.
I am investigating data related to AI or “Artificial Intelligence.” I’m interested in this topic because recent developments in the field, including generative AI platforms such as ChatGPT, Dall E, Midjourney and others have caused a lot of buzz around the world. People are both excited and concerned about AI, and it would be interesting to visualize how that excitement and concern has fluctuated over the past years, and to see what turning points in opinions there might be and what the causes of those shifts are.
Being at CMU — an institution that has been responsible for the development of much of the AI technology we have today, I would like to map out the history of AI development, particularly at CMU and understand what impact those developments have had on technologies used by the general population. Moreover’s how has people’s perception of AI changed across it’s development. What worried people about AI technology 20 years ago? What about 10 years ago? Or before ChatGPT? Are there patterns in those concerns?
Today I went to a talk given by Nita Farahany called “Chat GPT and your Brain.” Farahany is also the author of the book The Battle for Your Brain. One fact Farahany mentioned today is that only 33% of consumers think they are using AI platforms, when actual usage is 77%. This means that AI has crept its way into most of our technologies and we haven’t even noticed it. I am also interested in visualizing how AI is being used to influence human behavior in our media platforms. For example, we are now facing increasing amounts of AI generated fake news, deep fake videos and we are made addicted to media platforms through AI tactics such as recommender systems, autoplay, and more. The issues with AI are so pressing today, that President Biden has just issued an “Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence” and prominent voices in tech, such as Elon Musk have recently described AI as far more intelligent than us and suggest it is one of our most pressing issues of today.
After spending several hours perusing AI-related data on sites like Statista and Our World in Data, I have developed a new interest. AI and it’s implications are so broad, so it’s hard to gather a meaningful set of data to ask questions. I’m now interested in the intersection of AI and technolgy with tourism and travel. I have found various data sets related to AI and travel which I intend to work with. I became especially interested in this intersection when I learned that people had the hardest time determining if text came from ChatGPT versus humans when it’s content was travel (people had an easier time determining this when the content presented was about tech, entertainment, finance, and health). According to a survey conducted by tooltester, “travel content was the most undetectable when it came to GPT-4.0, as 66.5% of readers believed it to be human or human edited.”
So far, I am organizing my data thematically, trying to find links between certain pieces. For example, I found data sets on:
Top ways in which AI will improve online shopping according to youth worldwide 2023; Interest in AI-related products among U.S. adults 2023; Travelers expecting to use AI to plan trips in 2033 worldwide 2022, by aspect; Expected usage of ChatGPT to plan next trip in the U.S. 2023; Opinions on whether news written by AI is good or bad in the U.S. 2023, by age group; Reasons for feeling frustrated when planning trips worldwide 2023, by tourist type; AI-influenced revenue share of travel companies worldwide 2018–2024; and Estimated revenue of leading OTAs worldwide 2022, by device.
All of these pieces can be linked together. For example, some of the top interests in AI-related products in U.S. adults are flight and hotel recommendations, menu recommendations at restaurants, and virtual travel agents and customer support. People are planning to use AI to find accommodation, transport arrangements to their destination, local transport, passport renewals, personalized travel insurance, work events/meetings, meals and entertainment, experiential activities, and a trip’s environmental impact. Currently, people said they feel frustrated when planning trips because their searches don’t allow booking multiple options in single place, don’t include any alternative destinations/experiences; don’t include any alternative modes of transport; and don’t allow open-ended gsts that could inspire alternative options. Leading travel and tourism companies, such as Booking.com, Airbnb, and Expedia have recently invested in AI powered chatbots, and the share of travel companies’ revenue that was influenced by AI is expected to be up to 34% in 2024 since it was at 21% in 2021 and 9% in 2018. For now I am organizing my data by drawing blobs around related topics. I can see organizing it later in cartesian space. I think the most compelling thing I have discovered so far is the links between different datasets (like how people’s interest in AI can be related to improving everyday activities like travel planning, which travel frustration data suggests is a process that can be improved. Moreover, the leading travel companies are also jumping on AI and incorporating it into their product ecosystems and that in turn is expected to generate significant revenue). It might be interesting to visualize and highlight those links and see what emerges. For now, I think visualizing a web or network is most helpful. That will also help me ask questions about my data.
For now, am excluding AI data that does not relate to travel and tourism in some way, because the AI as a topic is extremely broad and I feel the need to focus in on something more narrow for now. There also appears to be a lot of data about AI, travel, tourism, and even the intersection of AI and travel, so it’s a niche yet rich area for exploration. I will group my data thematically and will begin making scales and ranges that show progressions. For example, scales showing AI revenue growth; I may need to add some more data because I want to think about how AI has changed and is going to change the travel industry over time. I had previously looked at some data that considers the projected use of AI tools over time and people’s attitudes about that - I may need to bring that data back into the mix for a more robust output. Overall, I think time could serve as an anchor and could also be used to develop scales.
Finally, I don’t see it now, but I would like to bring a physical map or geographic reference into the data, but I’m not sure how to do this yet (or if it’s something I should try to do in this project). To me, it would make a lot of sense if a visualizaiton related to travel or tourism inluded maps, but so far none of the data I have looked at mentions physical locations. I am wondering what kind of geographical data might be interesting to stir into this. What are the current issues with travel algorithms? Could AI improve that? Might AI be better at offering a wider range of individualized trip recommendations than the algorithms we’ve used in the past? What new geographical places might AI recommend that a online search in the past might not have? Perhaps these are ways to bring maps into the mix…
Change in Direction: Working with solid data.
After eagerly downloading a bunch of data from Statista on topics related to AI and travel and AI in travel and arriving at several interesting questions, I realized that most of the data sets I downloaded were very small and mostly ordinal, there were a lot of surveys of opinions. This data was interesting, but I was having a hard time putting the different surveys together in an interesting visual way. I went back to look at earlier data in AI that had been collected by the course instructors, Stacie and Vicki. I realized that those data sets were more robust and much richer for working with, so I decided to totally shift my approach and questions. From now on, I was going to find questions directly related to the rich data sets I had, because I understood that I needed rich data for a rich visualization. This realization led me to look at one of the largest data sets available. It’s about global private investments in AI from 2010 to 2021. After some browsing and quick visualizations, I was struck by the enormous increase of private global investments in AI. Below is my first visualization from this data.
My new questions were:
A quick visualization of the change in global investment in AI over the past decade:
I decided to sort this data by year. I would be visualizing it by country.
I could get detailed looking at the investment data per country.
Then I began to imagine how I might narrate this story.
Below is my first pass at a data visualization.
I have now taken a pass at some more visualizations, and added more data to my dataset that relates to what countries invested in AI each year as well as what types of industries those investments were made in. I have adjusted my color palette to be black and white with one accent color. I think green makes sense since we are talking about dollars invested.
I still need to refine these and add some information at the bottom. I want to make note of the top 3 countries investing each year and label what that investment is. With this visualization it’s very clear to me that the US has invested way more than any other country in AI. I will check with classmates to see if this is clear to them to, or if I need to add more labels and graphics.
Once my map was getting in better shape, I began the making the graphics below. The following graphics visualize the type of industries invested in each year (from 2017–2021; not 2010–2021 like for countries). I have highlighted the top 3 industries invested in each year and describe those with labels and icons at the top. It was interesting to see not only how much the amount invested grew, but to see that the top industries invested in remain largely the same. I am wondering if it makes sense to look at more than the top 3 for this visualization or if I should move on to another question or topic.
Finalizing the visualization — December 11–14
Above is a video showing the my data visualization as an interactive prototype that someone would be able to view, toggle and interact with.
This visualization of global investments in artificial intelligence investigates the following questions:
- How has global investment in AI changed from 2010 to 2021?
- Which countries invested most in AI each year? What are the top five countries investing in AI each year?
- What AI industries were most invested in from 2017 to 2021?
I used both geographic and cartesian coordinate systems to plot and represent the data I was working with. A geographic coordinate system made sense for me since I was looking to visualize global investments and to compare the investment amounts across various countries in AI over the course of several years. I used an absolute position for the positioning of the countries, and even have the centers of the circles that represent their investment amounts located at the exact geographic center of each country. I could have used relative position, however I chose to use absolute because I wanted to take advantage of using QGIS and it’s mapping abilities. QGIS is a mapping software, and it allowed me to create circles at the center of each country whose size varied depending directly on the amount of money that that country invested in AI that year. The QGIS mapping software doesn’t have a relative map projection (it does have many other kinds of map projections though!) so changing the country position from absolute to relative would have created more work, and I thought the data was very legible on a traditional map projection. This also allowed me to work with a lot more of the data that I had available in the set. In fact, I have visualized every single investment per country each year in AI that the original data set included. I was happy to be able to do that. It took a lot of processing the data and required a lot of work when putting together the interactive visualization — but I think that it’s valuable in that the visualization feels complete. I had a large data set to work with, and I was able to take advantage of that. To visualize the AI industries that were most invested in, I used a cartesian coordinate system, and I also did this when comparing the countries to each other at the end.
I used categories to group countries together in one part of my visualization. I colored coded the countries in that instance because it afforded a very quick comparison. I also added some text to hightlight key take-aways.
I used years as buckets. Time is the biggest toggle factor for my interactions. People can see how investment changes over time in AI globally and per country from 2010–2021 and can see some additional layers of information about AI investments from 2017–2021 where the data on the industries that are invested in is also available.
The narrative path I went for in my interaction is highly indexical with some linear components. People interacting with my visualization can toggle the year slider to see how much each country invest in AI in a given year, and at each year they have the option to go deeper and learn more about which countries invested most in AI that year — for example they can see the top five countries that invested in AI during that year and can also see how much each of those countries spent on AI investment during that year. People interacting with the visualization are also able to jump to a specific year without needing to go through other years — they could go from 2012 to 2021 and back to 2014 smoothly. I thought this would make the visualization more user-friendly since people have more control over what they want to see.
Below I’ll describe the overall narrative journey that someone might experience in interacting with this visualization, while recognizing that its non-linear format could lend to many different orders in the steps described.
In the first step, we toggle the year slider to see how AI Investment has changed from 2010–2021. In the visualization it is apparent that the global investment in AI has been constantly increasing and really took off in 2016. It is also apparent that the United States has invested significantly more than any other country (and more than most other countries combined) each year. It’s also interesting to see that in the most recent years, China and India have emerged as other countries with leading investments in AI.
In the second step, we check to see which countries invested most in AI each year. We can check an ordered list showing the top five countries that invested in AI for the given year we are on. It’s interesting to see how the top 5 countries that invested in AI each year changes some. While countries like the United States, United Kingdom, China, and India are big players throughout, it’s interesting to note which other countries are up there. It’s also interesting to see that India and China have been investing much more in recent years. I was also surprised to see that in 2021, Israel was in the top 5 — having not been on the top 5 list previously.
In the third step, we move over to a new visualization that displays the amount invested in AI per industry (globally) in billions of USD from 2017–2021. On the top of the screen we have an icon and labels that highlight the top 3 industries invested in for a given year. Below is a bar chart that is plotting the total amount of investment in AI for our time frame (2017–2021). The given year’s investment is highlighted in green on the bar. From this visualization we can see that data management, medical and healthcare, and financial technology are the top three AI industries that have been invested in. For the most part, those industries are at the top of the list each year — although I was interested to see that educational technology was on the top of the list during 2020 and third in 2018. This makes me interested to know what exactly AI is in educational technology and what companies are being invested in that are classified as creating educational technology.
Finally, we move onto a visualization that shows the top 5 countries that invested in AI each year side-by-side. This makes it easy to see that the United States is always at the top. It’s also easy to see which countries are repeatedly on the top 5 and which ones have been on the list just once or a few times.
The visual forms and graphic style that I am using is mostly simple. The choice of a standard/traditional geographic world map makes it clear that we’re talking global. The circles whose area varies based on the AI investment is easily legible, and the shifting scale of the circles makes it clear that the amounts invested are changing (increasing) over time. The more challenging visualization was the one looking into the main AI industries that were invested in. It was somewhat difficult to find good representations for industries such as “educational technology,” “medical and healthcare,” “data management” and others. The icons I selected work alright and I’m happy with how the bar chart is able to show the change in investment per industry over time, but I think this graphic could be pushed further. Finally, I just want to say that I think the dark background, radial gradient and bright green and white colors make the graphic look sort of tech-y, which is the vibe I was going for! I hope this came across to others.
Reflection
The biggest take-away I had from working on this data visualization project is that, when working with data, your story should come from inside the data, not from outside — meaning, you should look directly at the data to start forming questions, hypothesis and stories. When initially approaching this project, I had preconceived questions, hypothesis and stories I was interested in telling about AI, however I did not have data that directly related to the questions or narratives I was interested in. For the first couple weeks of the projects I was spinning my wheels looking for data that matched my specific interests. If from the start I had been looking at the rich, complete and clean data sets that were provided to me, I could have easily found the story I made the final visualization for. With more time, perhaps I could have layered some additional information or pushed my visual, aural or temporal styles in ways that I was not able to do for this project. Overall, I want to try visualizing data again after looking deep at the data, without preconceived questions. I think the magic in data visualization happens when you really let the data speak. When you are able to uncover interesting insights that are already there. While I think I did come away with some interesting findings — for example:
- The United States is by far the top investor in AI, every year.
- China and India have emerged in second and third place, respectively as AI investment players. This is interesting because in the early 2000s neither was on the top 5 list.
- No southern hemisphere countries have been on the top 5 list.
- The top AI industries invested in are data management, medical and healthcare, and financial technology.
I think upon further inspection and layering I could have come up with even more interesting take-aways. Overall, I’m looking forward to my next data visualization project and feel like I have a much better understanding of how I could approach one next time.
Sources:
Icons:
simple-illustration-that-prevent-heart-attack-heart-health-cardiology-outlined-icon-free-vector.jpg
