Making Science Make Sense
| Home > Scientific Writing and Editing > Feature Article Sample | Home | About | Samples | Contact |
|
New Directions in Scientific Software Usability
Ben Shneiderman, a professor of computer science at the University of Maryland, is one of the preeminent researchers in user interface design and human/computer interaction. Christopher Ahlberg, a former doctoral student of Shneiderman, founded Spotfire, Inc. and serves as the company's CEO. Trevor Grayling manages technical communications at MDL and has examined the impact of user interface design on online help usability, particularly in MDL's recent ISIS releases. We began by asking Shneiderman to explain the shortcomings of modern software interfaces. Shneiderman: The term “user friendly” captures the problem with the way we perceive computers. We shouldn't want computers to be our friends; rather, we need computers to be tools that support our creativity. This means understanding the fundamental difference between computers and people. People communicate by talking. Computers, however, offer a screen -- a visual display that is rapidly updateable and user controlled. Yet we try to talk to them, typing in commands and clicking “Submit” as a way to retrieve information. We neglect that we are working in a visual medium. Think about it. The human visual system is remarkable. And the computer, for the first time in history, gives us a rapid display that potentially can amplify human abilities a thousand-fold. By organizing this information and offering potent control panels that allow the rapid exploration of data, we can get away from the information overload associated with scrolling through lists of items and instead think of harvesting ideas from an “information abundance.” Molecular Connection: How does the typical scientist approach software? Shneiderman: If a picture is worth a thousand words, an interface is worth a thousand pictures. For scientists, this means an interface geared for discovery rather than for locating specific items. Scientists aren't so much searching as exploring. Patterns, relationships, outliers, clusters, jumps, gaps -- all of these terms represent the data that scientists might find interesting. Grayling: Such free-form exploration is certainly one of the most important tasks that scientists need to perform, and it's also a task that software applications have neglected until fairly recently. But even more tactical processes, such as those addressed by most scientific software to date, can be better supported. When scientists need to search -- for a particular compound in a chemical inventory, for instance, or for the results of a specific assay, or a compound MSDS because they spilled a solvent in the fume hood -- their highest priority is getting their work done. Years of customer surveys and usability testing at MDL have told us that scientists want to go into an application, do a specific task, and exit. The tools should match, not mask, the tasks. M.C.: Let's talk a bit about these more specific tasks. More often than not, the software really does support the jobs scientists have to do. So where do the frustrations lie? Grayling: One of the main benefits of windowed applications is that they have taken some of the guesswork out of using software. On the whole, our customers can get into an application like ISIS and accomplish quite a bit through experimentation. But playing around can only get you so far. The prevailing model of application software, be it ISIS or Microsoft Word, has been to supplement interfaces with help files or manuals. What we have observed, though, is that while researchers may be highly motivated to achieve a given task, they are also, paradoxically, biased against reading the very help systems that are there to assist them. When trial and error fails -- that is, when the interface doesn't supply the needed information -- users resist going into help, to the point of repeating scenarios that they have already attempted in the hope of getting different results. When they finally decide to try the help system, they tend to read hastily, misunderstand instructions, and make abrupt or poor hyperlink choices. So they often leave the system without the information they need, which means that the queries they perform may be ineffective and inefficient. Ahlberg: Another frustrating aspect of the way we communicate with computers, and a sign that something about the communication is unnatural to us, is that so much of what we do generates error messages. I have a Ph.D. in computer science, but eventually even I get intimidated when I'm working in Windows and it keeps telling me that I'm doing something wrong. Shneiderman: For all of the advantages that the windowed environment offers in terms of usability, we are still in an archaic communication cycle where you issue a command, view the results, issue another command, and so on. This is not the way humans think, and it is not how we perform many tasks. Imagine driving a car and having to type in “left 37 degrees” to make a turn. You could do it, but it wouldn't make for very good driving. The very design of computers -- a keyboard and a display -- has forced us into a particular communication model, one that works, yes, but isn't necessarily that conducive to the different tasks for which we now use computers. M.C.: So how can we begin to change the way we relate to computers? Grayling: One of the first things we can do is learn more about how we interact with computers and what we expect from them on the job. Throughout the software industry, developers are starting to focus on usability testing and usability engineering. MDL is no exception, and with many of our recent products, most notably Reagent Selector and SMART, we have reengineered our development practices to find out earlier in development how well the products we are designing actually work in real-life scenarios. The information collected during this testing not only helps improve a particular product prototype, but also impacts the development of future products. Ahlberg: Doing this type of testing also gives us the information we need to make more far-reaching changes in interface design. While working with Ben in the summer of 1991, I did a usability study with 18 chemistry students using a periodic table application that had dynamics similar to those now found in Spotfire. We did one test on an interface with graphical output and input, then tested one with text output and graphical input -- iteration upon iteration, in an effort to see how these changes affected the way our test group worked. The main lesson from many of these studies has been that the more visual and interactive an interface is, the easier it is for scientists to use. Shneiderman: I coined the term “direct manipulation” in 1982 to describe a visual software environment where typing is replaced with actions like pointing, dragging, and selecting. Consider a situation where a user wants to view information along a historical timeline. Instead of asking users to enter a start date and stop date for a timeline, why not replace the entry box with a slider that lets users move along the timeline, stopping and starting at particular dates to get information along the way? It's an interface change that can be implemented fairly easily and delivers information to users quickly and more accurately. Ahlberg: One of the advantages of a visual interface is that the tools are in harmony with the task. Instead of learning a complex command language, you interact with tools that visually represent queries. Pulling a slider immediately impacts the data. And you never get an error message saying, “Sorry, you have dragged the slider out of range.” With Spotfire, we had the opportunity to build an application from scratch on this new paradigm. We embrace Ben's theory of direct manipulation. The objects of interest (in this case, chemical compounds and associated properties), the actions you are performing (narrowing a search, clustering a list, determining the impact of a particular chemistry on activity), and the impact that the action has on the data are all available on a single screen. The fact that you are doing all of this visually may seem trivial, but not when you consider the sophisticated data structures needed to provide this functionality. M.C.: What will the software interfaces of the future look like? Shneiderman: I'm a big fan of evolutionary design, whereby incremental improvements in software make the systems more usable. Spotfire really is just a first step in trying to present vast information fields to users for visual exploration. Improvements to handle larger displays and more rapid data updates will challenge us for the next decade -- maybe even longer. For me, the next direction is a notion that I call GenEx, building tools that generate excellence and support creativity. Right now, users need to work hard to integrate the multiple tools they need in their jobs. I see a situation in the future where digital libraries, collaborative tools, and creative practices would be supported much more potently with computers. But that's a direction that needs more research. For now, it's rewarding to see computers becoming more like tools than friends; it's an evolution that I think will make all users, and especially scientists, more productive and creative. |