Designing air quality monitoring for citizen scientists: Interview with Karen Donoghue, Principal, HumanLogic
Introduction
Interaction Architect and Boston native Karen Donoghue has launched an air quality monitoring app called Local Haze as part of her global design practice focused on human computer interaction. A former cell phone designer, she continues to work on innovative networked products spanning wearables, robots and more. This past summer she co-wrote a book on early-stage product design with Craig Newell, a software architect and co-founder of SavaJe in Boston. I recently caught up with Karen to discuss her current work and the challenges of designing a product for citizen scientists.
Product Design and Product Iteration:
I met you at Mobile Monday in Cambridge, MA years ago. Since then you’ve worked on a number of interesting projects as part of your product design practice, including Local Haze, which developed out of your own challenges with Asthma. Can you tell us about Local Haze?
Local Haze is a free app for iPhone that crowdsources and aggregates air quality data from many different sources and delivers it in an easy to understand experience. Local Haze has grown to monitor crowdsourced data from around 30,000 sensors across six continents. The sensors range from different types of low-cost sensors to highly calibrated government ones. One of our goals in launching Local Haze was to help consumers and citizen scientists to be confident about their air quality data.
This video of you speaking about the Local Haze app to students at Tufts University Department of Computer Science was fascinating. It was interesting to hear your discussion of the design of monitoring large bodies of crowdsourced AQI sensor data and presenting the data in an experience that is easy to understand. Could you tell us more about some of the design features you iterated on — in terms of filtering and scrolling? What were some of the design principles behind that?
During that talk, we spoke to an audience of Computer Science students and faculty at Tufts University — so it was primarily a technical audience, though not necessarily one with an understanding of HCI (Human Computer Interaction) principles. We shared some of the challenges of monitoring air quality data worldwide at a scale of tens of thousands of locations while at the same time staying true to our goal of delivering an easy-to-use product experience.
To give the audience some background before talking about scaling a user experience to encompass many entities, we began by explaining basic HCI concepts and appropriate interaction models, starting with small numbers of entities. For example, we explained how list filtering works and how we initially used this approach for Local Haze. When the scale of entities becomes too unwieldy for list navigation, enabling search on the list is one approach that we explained.
We also reviewed the basics of HCI practices before talking about some of the challenges we faced when the number of entities (i.e. air quality sensors) we were monitoring began to grow to the tens of thousands. By that time, when the numbers of entities were at such a large scale, we had to move to more complex presentations using geographic locations to let users navigate increasingly larger numbers of entities. We iterated on using lists to present the sensor data and felt like it was too much scrolling, so we decided to add maps. This turned out to be a problem because we had so many monitors that the early map versions were kind of disastrous. There were so many monitors that the maps were too crowded, too unclear and unusable. We wanted to maintain ease of use as a core design principle, so we developed clustering algorithms to present the sensor data in a way that made sense during the map interaction using the pinch and zoom gesture in the native iOS UI. Enabling this type of interaction was very technically challenging, and it took us a long time to get right in order to deliver a performant experience on a phone. In earlier map experiments (at left in the image below), you can see how the entity crowding made the map difficult to navigate. On the right you can see the results after we figured out how to display the summary information in a way that made sense for our users in terms of map navigation.
We also wanted to explain that our efforts in designing Local Haze were not just focused on making pretty maps or visualizations, but to make air quality data easily usable and discoverable for a wide audience of citizen scientists — not only air quality experts. As part of this we attempted to explain the differences between graphic design and interaction design, as we often see a misunderstanding between these distinct design practices. We wanted to show how the visual display of many entities is one aspect of design, but that there are many challenges when designing the interaction models as part of the user experience.
Did you research other potential users of the app to understand their needs as well?
We had many constraints of time and resources in developing Local Haze, so for our initial release we stayed focused on our primary target persona and their pain points and needs. We refer to this target user as an “Air Quality Enthusiast.” Through our user research we have learned that this type of user is motivated to understand and analyze their local air quality sensor data because they are concerned about local air quality or may have a family member with asthma. They are also interested in the rise of low-cost sensors and buy these sensors for their homes in order to regularly monitor air quality and decide on what course of action to take based on their current local air quality readings.
We will likely need to serve other user types and will carefully research those personas as needed, but for the early releases of Local Haze we have remained focused on delivering a great user experience to our primary persona.
You indicated that users have requested a number of other features. How are you prioritizing those requests? And, how do you plan to incorporate those features while still maintaining the simple user interface of the app?
It’s a good question, Perry. From your experience as a Product Manager, you’ll recall the art of PM requires dealing with the challenges of balancing the needs of users with the constraints of what you can build and deploy at the level of quality you want. Prioritizing is always a challenge and in my opinion, one of the most important skills for a PM.
Most of our feature requests come from users, but we also communicate with sensor manufacturers and others in the ecosystem. We analyze incoming requests and weigh them against the value we think we can deliver in the product experience and any potential risks in delivery. One of the biggest challenges has been managing requests for more granular data and details to be added onto the screens while maintaining our design goals of delivering an easy-to-use and glanceable user experience.
As I mentioned earlier, we also had to put a lot of effort into developing the clustering algorithms for the map interface. This map clustering feature resulted from a combination of user feedback and our observations and perspective on what it means to deliver a useful geographic view of air quality data. We had to figure out that balance of listening to users and delivering what we felt was the right user experience for the mapping feature — a real interaction design challenge!
What’s coming on next for Local Haze?
I’m glad you asked Perry! We’ve got lots of feature requests on our roadmap, and recently launched a notification feature for alerts from specific (PurpleAir) sensors. The ability to set notifications on a sensor is our first subscription feature. Now Local Haze users can set alerts to receive notifications from specific PurpleAir sensors with one tap on the sensor details screen.
In the screenshot below you can see an example of a notification for a PurpleAir sensor located at Nob Hill in San Francisco. When there is a substantial update in the air quality reading for the sensor, Local Haze sends a notification to the user’s phone.
To get this feature right, we had to do a lot of experimentation with the cadence of the notifications. Users told us in our beta testing that they like the notifications, but this feature has surfaced challenges of delivering a great time-based (or “temporal”) experience. To address this we are working to “tune” the algorithms to deliver the air quality notifications at the right cadence — this has been a challenge and we were careful to do a beta release before launching the feature. From our beta users, we have learned that notifications delivered too frequently can be perceived as a “noisy” user experience.
It’s always a challenge to strive for a clean and easy to use experience while taking into account new feature requests from our users. We are always clear about our objectives and intent in our approach to product design. We strive to maintain the right user experience for our users and improve with every release. We want to deliver a positive and trustworthy product user experience, and we are continually working on improving how we do this.
Where can people find out more about Local Haze?
Local Haze can be downloaded for free from the Apple App Store, and we maintain a web site for the product at https://localhaze.humanlogic.com/
References
Download Local Haze for free from the Apple App Store
Additional papers and books by Karen Donoghue and Craig Newell
