Google Search is an amazing tool. It works so well you don’t notice that you are using it, only that you have used it.
In my experience, there are a roughly three different categories of Google searches, and the third one highlights some room for (specific) improvement.
In the first case, what you want is a set of results, which can include a set with only one member. Like if you want to look up the phone number for a restaurant. In the second case, what you want is an ordered list of relevant links and information. The third type of search is when you are researching a topic and you find yourself edging close to the frontier of human knowledge. If you are at the frontier, what you want is a node-link graph.
Google currently works great for the first two types of searches. However there is room for improvement for the third type. To see how, let’s dive into a specific case, using a google product specifically optimized for research – Google scholar — and explore how its UX might be improved.
Say we use google scholar to look for research related to Edward Webb’s 1915 Ph.D. thesis about the measurement of “character” and “intelligence.”
What we get is a small set result:
Now, if we have read it, and we want to know more, we click on the “Cited by 340” link. Voila, we are presented with this screen, which looks like the ordered list result:
Yet we don’t really want an ordered list. Because essentially we have found one node in the entire graph of academic papers, and we are now looking at five of its connections.
If we were to click on the next “cited by ____” link, we’d be crudely traversing through parts of a node-link graph. If you click on “Cited by” enough times, you’ll travel through a small part of humanity’s structured scientific knowledge. And if you know what you are looking for, you’ll get close the frontier of human knowledge.
The screens that you see when you use Google’s search engine are, in a very real sense, data visualizations. (And the same is true for what you see when you visit the home-page of any academic journal.) As such, Google’s interface can be viewed through Edward Tufte’s information-to-ink ratio, or the idea that the best displays of data maximize the amount of information for the amount of “ink” that is used to make them.)
The Google Scholar interface robs me of two pieces of context which could be used to increase the information-to-ink ratio. First, there is no way one I have time to dig through all 304 articles that cite Edward Webb’s Thesis — some context about relevance to me would be nice. However, Google doesn’t know what I know of this frontier of human knowledge. Second, I’m also missing out on an important (albeit implied) network of scientific collaborators. If professors A & B write one paper and professor C writes a paper that cites A & B, it could be valuable to know that C collaborated with D in the past, and now, D & E have collaborated on a relevant paper that isn’t part of the immediate web of citations but important for understanding the topic.
If Google Scholar results could be viewed as a graph, I could immediately see which papers cited each other. If I could highlight the nodes (papers) I’ve already found to be useful, I could sort through the 340 other citations to help refine my search. With a graph-based UX, I could search and monitor the frontier of knowledge about a specific subject.
Given Google’s mission (“To organize the world’s information and make it universally accessible and useful”), this seems like something they could lead. Such tool could reduce the variation in quality of search, yet again making research more efficient. And if Google doesn’t want to do it, there are other search engines that could. Potentially every scientific journal could also do a version of this idea as well. PLOSone, Nature, and Science all could utilize interactive, possibly 3D, node-link graphs to help researchers find relevant information.