Dynamic Networks: What’s That?

What are dynamic networks, and how can we use them?

Think about your own friendships throughout your life. You’ll meet new people, lose touch with others — the network of people you know is constantly changing. And that’s exactly what a dynamic network is. It’s a representation of how actors and their respective relationships change over time.

The ability to visualize this kind of data is obviously useful. But in practice, it is not so simple. There isn’t really a “right” way to visualize dynamic networks. But why is that? In order to answer that, let’s take a look at two of the main ways that experts are currently displaying dynamic networks.

The first is called interactive animation. This is just an interactive, continuous timeline where the user can choose which moment in time they want to look more carefully at. Giving the user this flexibility and control seems great, until they’re given a timeline that spans millions of years. Imagine how much time it takes for the user to go through all of that information and narrow down the exact moment in time that they want to look at. Spoiler alert: it’s a lot.

That’s why splitting up a timeline into multiple parts (also referred to as “timeslices”) is the more popular alternative. This process of splitting up the timeline is called “small multiples”. If you give the user the ability to choose from multiple timeslices, rather than choosing on the timeline itself, users are able to complete their tasks just as accurately and much more quickly.

Unfortunately, that doesn’t mean we can just split up the timeline and call it a day. Just because we know that splitting a timeline into multiple timeslices is the better way of viewing a dynamic network, doesn’t mean that we know how to split it up. If we have too few timeslices, we’re going to have way too much data in one timeslice. Not only is that going to be really hard for users to process, but we’ll also just lose a lot of subtlety when watching how the data evolves. But if we have too many timeslices, it’s going to be hard to navigate, and hard to remember what we’ve already seen.

Clearly, visualizing a dynamic network is complicated, and there doesn’t seem to be a right answer. So now what?

This question is exactly what researchers at Swansea University wanted to explore. They came up with three different ideas of how to visualize dynamic networks, and tested them out to see which performed the best. They then displayed these visualizations on a large touchscreen monitor.

Researchers gave 24 participants a series of tasks to complete, repeating these tasks on each of the three visualizations, and compared the results to see which visualization performed the best.

The three visualization ideas were interactive animation, small multiples, and interactive timeslicing. Let’s take a closer look at these ideas.

Interactive Animation

If you remember how we defined interactive animations before, then you can probably guess how the researchers chose to implement this. They gave participants a giant timeline, allowed them to choose which window of time that they wanted to look at, and then showed an animated graphic of how the nodes and relationships changed over time.

Small Multiples

In small multiples, researchers split up the timeline into uniform pieces, so when the user selects a period of time from the timeline at the top, they’ll be able to see visualizations of four consecutive timeslices at the bottom. In this case, you’d read the visualizations like a comic strip. Reading the frames from left to right will show you how the data changed in the time frame chosen by the user.

Interactive Timeslicing

In interactive timeslicing, it works pretty similarly to small multiples, except the user has full control over how to split up the timeline. They can create time windows anywhere they want on the timeline, and can change how long each time window lasts.

Results and Conclusion

We already knew that breaking up the timeline would be better than having it as one continuous interactive animation, but now that testing has finished, let’s go back to our original question — how should we break up the timeline?

Well, the answer is “it depends’’.

Researchers found that interactive timeslicing is great when you want to compare two specific points, or data at disjointed periods of time. But if you want to analyze a continuous period of time, you’re probably better off doing small multiples.

So while there never will be a clear “right” answer, many thanks to the folks over at Swansea University for offering us guidance as to when to use which visualization. 🥳

Citation

Read the full paper here!

• Lee, Alexandra, et al. “The Effectiveness of Interactive Visualization Techniques for Time Navigation of Dynamic Graphs on Large Displays.” IEEE Transactions on Visualization and Computer Graphics, vol. 27, no. 2, 2021, pp. 528–538., https://doi.org/10.1109/tvcg.2020.3030446.