These days I read some cool papers about information visualization, especially in scientific study.

Balancing Systematic and Flexible Exploration of Social Networks

Summary:

This paper discusses a system called SocialAction, which allows users to systematically analyze social networks. SocialAction allows users to use a variety of techniques, such as attribute ranking, filtering, and graphing, to better visual and understand large and complex networks of data. The purpose of this system is to help users make discoveries and find patterns in data through flexible and iterative measures.

More info about this system on UMD HCIL website

Cool demo video:

The social network of the U.S. Senators voting patterns in 2007, after Democrats took control. Republicans are colored red, Democrats blue and Independents maroon. Here, the partisanship of the parties appeared automatically (180 vote threshold).
The social network of the U.S. Senators voting patterns. Here, the threshold is raised to 290 votes. The Democrats’ relationships are much more intact than the Republicans. Details-on-demand are provided for Senator Whitehouse, the senator with the highest degree at this threshold.

Critiques:

I really like this system because it does help with cognitive load with the affordances it adds to the visualization. For example, the colors indicate the ranking and the size denotes the number of nodes in the group. Screen_Shot_20171014_at_3.19.43_PM.png

SocialAction_1.png

It reminds me of the data visualization talk “Visual Trumpery” by Alberto Cairo. He coined the term “graphicacy” and argued that data scientists and designer should check the accuracy and appropriation of data visualization before publishing them. One point is that sometimes over-aggregated data may be misleading. We want to reduce cognitive workload but should not do it at the expense of providing incomplete even inaccurate data.
The system in this paper both provides system enables users to rank and filter nodes, it did improve the effectiveness of conventional node. The flexibility provided by the system allows users to interpret the networks in a qualitative way. And the systematic visualization keeps the information complete as well as supports formal characterization.
This system can also be used in the analysis of the collective behavior of customers to see the correlation of social network and shopping, entertaining, or traveling choices.

Unfortunately, the author did not  have good maintenance of the system. 

Reference:

Adam Perer and Ben Shneiderman. 2006. Balancing Systematic and Flexible Exploration of Social Networks. IEEE Transactions on Visualization and Computer Graphics 12, 5 (September 2006), 693-700. DOI=http://dx.doi.org/10.1109/TVCG.2006.122

Motif Simplification: Improving Network Visualization Readability with Fan, Connector, and Clique Glyphs

Summary:

In order to simplify network diagrams, the researchers in this paper used “motif simplification”. Essentially, networks tend to have common structures called motifs. If you break a network down into its motifs, and give each motif its own symbol, or glyph, then you can get an abstracted and easy to understand the of the network structure. In this paper, glyphs were made for three high-pay motifs: fans, connectors, and cliques.

To test if the effect of motif simplification on user performance they ran a controlled experiment with 36 students where they asked them several questions about the networks. Unsurprisingly, they found that overview tasks were easier with less visually complex networks. They concluded that using glyphs for motifs makes the motifs easier to detect and measure.

Cool demo video:

Critiques:

The pros and cons vary based on the cases. Some drawbacks are evident such as high cost of user education. Users need more time and effort to understand the meaning of glyphs for motifs. Additionally, for overlapped motifs like cliques, the choice of which motifs to simplify can also lead to different result.

The controlled experiment itself showed some drawbacks of the design of motif. Results show that motifs are easier to detect and measure. But for the rest of the network,
Topology-based tasks are a mixed bag. Some tasks like finding cut nodes were more accurate, but path-based tasks showed different results in three circumstances. Even though the author emphasized the benefits of simplification, most tasks in topology did not represent the advantages of glyphs.

Perhaps for networks with a huge amount of nodes and complicated data structures like the web case, glyphs are less efficient than complex networks.

Reference:

Cody Dunne and Ben Shneiderman. 2013. Motif simplification: improving network visualization readability with fan, connector, and clique glyphs. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’13). ACM, New York, NY, USA, 3247-3256. DOI: https://doi.org/10.1145/2470654.2466444

Apolo: making sense of large network data

Summary:

Since the large network dataset has improved beyond the capability of many domains to extract useful language, the authors of this article created a system called Apolo, which combines visualization, users’ interaction and machine learning together to guide the users to incrementally and interactively explore large network data and make sense of it. The difference of this tool compared to other sensemaking approaches is that the users start small and build up the understanding in a bottom-up way. Also, Apolo helps users find relevant information by specifying exemplars and then uses a machine learning method to search for other related nodes. The authors evaluated Apolo with twelve participants in a between-subjects study, with the task being to find relevant new papers to update an existing survey paper. Using expert judges, participants using Apolo found significantly more relevant papers. And the feedbacks of Apolo are positive.

The authors believe that the ideas in Apolo can be helpful for many kinds of data-intensive domains.

Cool demo video:

Critiques:

This visualization tool is the most integrated in the three and it is pretty cool and quite efficient. From the experiment result of the controlled study, Apolo has high usability and good ratings. But its drawbacks are also evident. I regard it as a good tool for entry or intermediate level scholars.

With the support of machine learning, this system can give the result of most relevant articles quickly at the expense of providing the chances for users to find brand new things. Sometimes innovation happened from the collision of two irrelevant elements. Apolo is designed to support lookup instead of exploratory searches. Users are assumed to know a small number of documents within their interest. In this case, I would not totally rely on this system.

In addition, my concern is not only the restriction of creativity also the rich get richer phenomenon in academia. From the case of the paper, users can choose to show more articles, add suggested articles for each group. Newly added neighboring nodes ranked by citation count, which means that the more citations the paper have, the higher chance for it to be searched by users. Imaging all the scholars use this system, some good paper could be unknown forever.

Reference:

Duen Horng Chau, Aniket Kittur, Jason I. Hong, and Christos Faloutsos. 2011. Apolo: making sense of large network data by combining rich user interaction and machine learning. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’11). ACM, New York, NY, USA, 167-176. DOI: https://doi.org/10.1145/1978942.1978967