Published in


Visualizations by Sam Pottinger, details below.

Natural Language Processing: A new way to listen to people about health

This interview series is a part of a collaboration between IDEO, Robert Wood Johnson Foundation, and Building H to imagine how we might design health into everyday life.

To make everyday life healthy by design, we need to understand what happens not just at the individual level but also what the trends are at the scale of communities and populations.

In this interview, we’re speaking with Sam Pottinger, a data scientist and data visualization artist at IDEO, to learn about a form of data science called natural language processing and how listening to public conversations happening online helps build empathy at scale.

As a data scientist, how do you look at systems as complex as health or as big as everyday life? Where did you start your exploration?

Health is massively complex but it’s also something to which all of us have a relationship. So, the internet is full of stories and perspectives if you know how to look for them. Researchers can learn a lot about a problem space by listening to and honoring those experiences. This broad perspective lets more voices influence design and offers an additional important source of inspiration. I start with questions like, “what topics are discussed in different parts of the internet? How are topics interconnected? What can those connections tell us about systems and opportunities for design?” Some of what we found in qualitative interviews influenced this. The team heard how COVID is prompting people to think about their health and values in a new way. This inspired analysis to focus on not just what is in these discussions but ask how has conversation changed from before COVID to during?

As you were researching this topic, where do you look for data?

In the context of everyday health and COVID, we looked at public information in a bunch of places, trying to see how conversation differs across different parts of the internet and for different topics. All that said, the three that really had an impact on my thinking in the space were Tumblr, Crunchbase, and a sample of news media (see data note for rationale & limitations) because they give you a really rich window into very different communities.

What did you learn about how the public thinks about health?

We learned a lot, but something that stuck with me is that, while many efforts focus on education around the importance of “key inputs” like diet and exercise, analysis suggests that the public already discuss these ideas frequently. This is important because these are the behaviors that can help reduce many chronic illnesses. Therefore, I think this observation challenges us to build on this valuable awareness by helping individuals on the health journeys they are already taking instead of adding an additional program to lecture on the importance of those key inputs. Also, in addition to information about frequency of topics, the connection between topics also informed our thinking. For example, encouragement and community become valuable elements for design because the data suggest they may have a connection with diet and exercise. Furthermore, similar overlaps between loneliness and mental health in the data further emphasize that nurturing human connection may help foster health in everyday life.

Visualization from Sam’s research. Frequency of different topics in health-related Tumblr posts as measured by hashtags. Arcs represent frequency with which these topics overlap with line width representing the amount of overlap (see data note for methodology). Arcs are semi-transparent but connections to top 5 hashtag groups made opaque for emphasis.

Going back to what you said earlier, what has changed since COVID-19?

Examining health-related posts but comparing data from before the pandemic to during, analysis highlights important changes in the public conversation, especially in mental health. The list of hashtags with significantly heightened volume on Tumblr includes entries like “anxiety” and “depressed” (see data note). In fact, mental health related tags account for 1 in 3 of the hashtags showing significant increases in volume on the site’s health related posts, a finding in line with other early research into mental health during the pandemic. This may be social media underscoring the importance of mental health during COVID. Given what I mentioned earlier with loneliness, there’s hints that there’s a resounding need for human connection, especially now.

What are some emerging themes that we should all pay more attention to?

In terms of themes, I think the observations around mental health and loneliness make me wonder how we might encourage greater human connection, especially during COVID. Meanwhile, the findings around inspiration make me wonder how we can care for and motivate each other.

That said, in addition to these themes, I also saw “gaps” in the data that might tell us where we need to be paying more attention. For example, I analyzed how the topics found in the Tumblr research compare to similar analysis in news articles and Crunchbase. Here, the data show that the “key inputs” of diet, exercise, and mental health each appear relatively more often in Tumblr discussion than in both news media and industry (see data note). This might suggest that the public are already thinking about health in an everyday context but that potential lies in bringing more institutional or entrepreneurial efforts to bear in these key inputs.

Visualization from Sam’s research. Comparison of how different groups are considering health, showing how they compare and contrast. Top 5 from public conversation made opaque for easier reading. Powered by News API.

Going forward, how do you see data science changing how we design systems?

These kinds of data science-based explorations broaden the perspective of teams and, for this project, I think help us look at a space from “above” in a way that is difficult through other ways of seeing. It lets us look at a system more broadly. For this project, for example, all of this helped build towards a deeper understanding of a health system and economy that understands the importance of community and nurturing human connection.

Are there limitations to this approach or your findings? And given more time, what are you curious to explore next?

Absolutely! Just like in traditional qualitative research, we have to consider where we are listening and who does or does not have access to those spaces. During design, we have to be thoughtful about those blind spots and consider how different methods of research can complement each other to increase inclusivity and broaden our perspective. To that end, I have a data note on some of the limitations and assumptions around the findings I mentioned earlier. That note discusses some future work opportunities in greater detail but, in general, I’d definitely spend some time in more communities and more languages.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store