Visualizing NYT Discourse – Design Iteration 3

tags: info graphics NYT API process visualization vox populi

posted by Zeke Shore on Mar 1st, 2010

The last design iteration I wrote about a couple weeks ago started to take a departure from earlier iterations by exploring the idea of representing the personality of every comment on the New York Times website (that relates to any given topic) as it’s own entity, and visually describing it’s sentiment or personality.

After reflecting back on our original reasons for wanting to visualize online discussions, our thesis question really centers around how can the ‘Vox Populi‘ still be heard as reader participation in the journalistic process scales to hundreds of thousands of comments spread across hundreds of articles and blog posts for even just one news source.

So this resulted in a design prototype that involved rendering comments for a given topic as balls swarming around the article that seeded the conversation, representing sentiment with color, opacity, and speed of movement, describing each comment’s polarity (how positive or negative), strength (strong or weak) and activity (active or passive) respectively.

While this iteration was both readable and interesting to look at, it suffered in terms of scalability. We could realistically only look at a couple conversations at a time for any given topic. So the next phase of the design process involved trying pull some of the more successful aspects of this iteration into a more real estate friendly composition. The logical progression of this involved breaking conversation into a linear organization (all of the following mock ups are not visualizing real data, but rather serving as design explorations).

Of course horizontal flows of information are rarely web-friendly, despite it being a logical way to organize content chronologically. So this quickly evolved into a vertical orientation, and opened the door for exploring the concept of possibly showing when commenters reference each other within a conversation.

While this started to maximize screen real estate a bit better, it also presented the opportunity to scale our level of abstraction out an additional level to average the sentiment of entire conversations (rather than individual comments) and theoretically see stronger sentiment trends for a given topic over time.

Now we started to realize the true challenge of finding a balance between abstracting the data enough to see trends over time, but still keeping the full visualization tied to the contextually relevant points of information that where creating the image in the first place. The problem with averaging all of the positive and negative sentiment of a full conversation is that it actually starts to cancel itself out… if we depict conversations along a color spectrum between yellow (positive) and blue (negative), they all hover around various tones of green, and we loose all sense of the sentiment dynamic that exists within each conversation.

At this point revisiting the successful  aspects of earlier iterations became imperative. Going back to where we left off in the previous round of design iterations, simply by breaking our earlier swarm of comment bubbles to live above or bellow a seeding article based on a median neutral sentiment level helped distinguish the true polarity of a conversation by layering the color distinction with a spacial distinction.

Another idea that had been floating in and out of our repeated design evaluations was hopefully providing the ability to filter a search topic by sub-topics, and perhaps even be able compare these sub-taxonomies with each other. While looking at the sentiment surround “Obama” might be interesting, seeing the sentiment surrounding “Obama and Healthcare” could be far more revealing.  Taking that idea a step further, it might be even more revealing to be able to contrast one filter with another, for example seeing the chronological sentiment surround “Obama and Healthcare” in contrast to the sentiment surrounding “Obama and Iran.”

This lead to a final concept for this design iteration. First, this design attempts to help better describe the true polarity of a conversation by proportionally showing the amount positive sentiment vs. the amount of negative sentiment as two separate bars extending above and bellow our axis of seeding articles. Second, we maintain a level of abstraction that allows dozens of conversation surrounding a subject to viewable at one time, allowing trends over time to potentially be revealed. Third, this design begins to explore the possibility of allowing users to group multiple keywords into one timeline, as well as group multiple timelines into one visualization, thereby allowing the sub-topics of larger topic to be compared (again, this is not visualizing real data).

While this final concept does indeed solve a lot of our previously established issues, as well as exploring some new possibilities, it is far from being fully resolved. We have unfortunately fallen back to our horizontal orientation (which is not very web-friendly) in order to accommodate these multiple timelines. Additionally, while splitting our conversation sentiment into ‘positive’ and ‘negative’ bars extending out in opposite directions from our article axis does help distinguish conversation polarity, we are still failing to represent actual voices (or individual comments) at a larger scale, which is our underlying goal in the whole project.

Filtering topics by additional keywords and comparing timelines seems like it could  have some really compelling use cases. However, as we become more familiar with the content we are dealing with, the prospect of just having users stack random keywords and timelines at-will raises some big usability read flags. Namely, we are able to know what the popular topics (and sub-topics) of discussion are from constantly parsing through and aggregating API outputs. but expecting a user to correctly guess what the active sub-topics of discussion surrounding “Haiti” are is just not realistic. If we want to provide the ability to compare the sentiment surrounding the various sub-topics of conversation within a larger topic, we will need to explore some additional solutions to guiding user in to what that active breakdowns of discussion are.

Coming Soon: Design Iteration 4