The Dark Side Of Usability

In Phaedrus, Plato documents a dialogue between Socrates and Phaedrus. Socrates shares a story on the subject of writing. The story takes place in ancient Egypt, ruled at the time by the god king Thamus. One day Thamus was approached by Theuth, an old god, inventor of many arts and sciences including arithmetic, geometry, astronomy, and most important, the use of letters–writing. Theuth wanted to share with Thamus his discoveries so that the people of Egypt could benefit from them.

The two gods discussed Theuth’s discoveries, and Thamus gave his opinions on each. When it came to the subject of writing, Thamus expressed his disapproval. While the discovery was ingenious, he believed that it would make people more forgetful. As people would begin to write things down, they would rely less and less on their memory. People would appear omniscient, yet in reality they would know nothing, having to rely almost completely on knowledge external to them. Writing would be an aid not to memory, but to reminiscence.

Socrates held a negative view of writing. He believed that putting thoughts down on paper would relieve you of committing them to your own memory, where they can be dwelled on and fully assimilated. Additionally, reading from paper only goes one way. You cannot ask questions and get feedback, something that can be done through spoken communication. He was not completely against writing, for example he thought it useful in old age as your own memory begins to fail, but his view was essentially that writing not only gives but takes away a lot in the process.

He was wrong to dismiss the value of the written word, as evident by the progress it has brought us over the last couple of thousand years, but this doesn’t mean he was also wrong about the its downside. Committing things to memory requires a lot of conscious effort, and writing things down is a way to relieve yourself of that effort. Of course there is no point trying to memorize everything–a lot of things are too trivial to warrant that, for example, keeping a calendar of appointments. The appointments are important, but it’s not important to memorize their dates when they can be stored and retrieved with ease from an external source. On the other hand, some ideas should be committed to memory, for example, that interesting new design approach, fully assimilated will give your mind opportunity to develop it further as well as use this knowledge in the context of everything else you do.

So how does Socrates’ story relate to usability? In 2008, Christof van Nimwegen published a paper called The paradox of the guided user: assistance can be counter-effective. The paper documents a study on user interfaces and pits two groups of interfaces against in each other in order to see how their users perform.

One group of interfaces is more user-friendly, in that it attempts to aid the user as much as possible in their given task, for example, indicating options available at a given moment of their task (e.g. greying out and disabling unavailable options, or highlighting available moves). This interface tries to offload as much “trivial” mechanical thinking as possible to the machine to give the user more time to think about the problem itself. The other group of interfaces don’t provide this level of help and hints, forcing the user to get better acquainted with the mechanics of each problem and put more thought into how they will want to complete their task.

In essence, one set of interfaces externalizes information, whereas the other internalizes it, in relation to the user. Just like in the story of Socrates, the user-friendly interface relieves its users from having to commit information to memory, and this information is in turn externalized. When the interface isn’t helpful, information is internalized by the user, meaning they have to think longer about the problem and learn more fully the inner workings of their task.

The findings are very interesting and go against the general trend of simply accepting usability of user interfaces as something completely beneficial to the user. Indeed, the opposite thing happened. People using the more difficult interfaces tended to perform better, were less fazed by distractions and were found more likely to transfer their skills to new interfaces or tasks. People using the user-friendly interfaces tended to rely on them too much, and so were never fully able to grasp the problem and come up with working strategies for tackling it. Even though the user-friendly interface was meant to relieve them of trivial tasks, they instead relied on the computer as a crutch, stumbling around on their way to an eventual solution with the help of the interface.

So what sort of tests did Nimwegen perform and were the tasks close to any real world work we may be doing? Nimwegen conducted experiments with 3 different applications. The first is a computerized take on the famous math puzzle Missionaries & Cannibals that was called Balls & Boxes. Perhaps not something we do every day, however, the other two tasks were much more “realistic”, and were called: Conference Planner and Ferry Planner.

The Balls & Boxes problem involved moving yellow and blue balls from one box to the other, whilst following a set of rules. On the screen users would see the balls in two boxes, together with a dish to transfer the balls between them. There was also a set of clickable arrows that let the users move balls and the dish around. Available arrows were highlighted and clickable, unavailable moves were greyed out and disabled. The “unfriendly” version was the same, except all arrows were “available”, giving the user no hint as to what moves they could actually perform in line with the rules of the task.

What was measured? Time to solve the puzzle, number of moves made (including number of superfluous moves), number of dead-end states reached, knowledge of the rules test conducted after the puzzle, as well as questions concerning perceived amount of planning and feeling lost during the puzzle. Additionally, a distraction was introduced after the participants solved a few puzzles, which involved a visual rotation task expected to erase Balls & Boxes related routines from working memory. This distraction task would go on for 10 minutes after which the participants would resume the Balls & Boxes puzzle.

Results? Those on the Internalization interface solved more puzzles. They initially took more time to solve them, but closed the gap quickly as they spent more time on the problem. While they also made a little more superfluous moves the start, they quickly overtook the Externalization interface participants with each experiment phase, and were not beaten on the number of dead-end states. Internalization participants also performed better on the after puzzle questionnaires.

It’s a revealing outcome that’s also mirrored in the more realistic Conference Planner and Ferry Planner experiments. The Conference Planner was a custom application designed to simulate planning of speakers for a conference, which looks a lot like a typical calendar app. Various constraints are introduced, for example rooms needing to fit a certain number of audience, projector requirements and time slots. The Externalization interface in this case would help by highlighting available slots when trying to place a speaker. Similar indicators to the first experiment were tracked: total time, time before first more, time between moves, duration of a move, superfluous moves, etc.

The results: Internalization participants took longer to start and longer to make each move, but the total time was similar to that of Externalization participants. They also made a lot fewer superfluous moves and came up with better working strategies for solving the task. Externalization participants tended to do more guesswork, relying on the interface to guide them along.

To test transferability of the skills the participants learned in the conference planning task, Nimwegen devised another experiment with a similar problem: the Ferry Planner. The interface looked different, but the fundamental mechanics behind the problem were the same as in the previous task. Results again were telling: the Internalization participants performed much better right away, completing the task quicker with much fewer superfluous moves.

The conclusion is that the experiments didn’t demonstrate the benefits of the Externalization of information–indeed, they showed that Externalization may not always be beneficial for the end user. Interfaces that didn’t give the user hints forced them to think more fully through the task ahead, and thus come up with a plan of attack. Internalization participants also tended to cope better with interruptions. Those relying on the screen for directions were more likely to go for guesswork and did not transfer skills to a similar problem. The interface, instead of acting as an aid, became a sort of impediment to efficient and smart problem solving.

Externalization makes users count on the interface and gives them the feeling (unrightfully so) that the thinking-work is done for them. This seduces them into more shallow cognitive behavior and discourages undertaking cognitive activities aimed at strategy and knowledge construction. Users who internalize information themselves behave more plan-based, invest more effort in cognitive processes, and are more proactive and ready to make inferences. This in turn results in more focus, more direct and economical solutions, better strategies, and better imprinting of knowledge. This knowledge is easier to recall at a future point in time, and is better transferable to transfer situations where the interface, the task, or both are different, and less vulnerable to a severe interruption.

Christof van Nimwegen - The paradox of the guided user

Does this mean usable software is bad? No. Software where usability is key, for example, ATMs, information websites or home video editing apps will not benefit from Internalization since the tasks involved will likely be casual and incidental. Remember, the tasks still get done just fine with the more helpful interfaces, they simply push more of the work on the interface and involve less learning, which isn’t required in the above circumstances. The findings of the study are more applicable to serious task environments where learning itself is the aim. Nimwegen suggests to take care with externalizing especially in tasks with: frequent interruptions, educational objectives, skill transfer, high costs, continuous attention and deep domain understanding.

By crafting simple and user-friendly interfaces we relieve our users of the need to think–or more accurately, to think about the more trivial and mechanical parts of the task, things which can be outsourced to the machine. But by doing so we are at risk of indadvertedly surrendering more than we bargained for, as we are lured into thinking that the interface will do our work for us–and so we end up spending less time thinking through the problem and less time planning.

Of course these conclusions are based on a limited experiment, with a limited sample size and test applications. The findings focused on only a certain type of interface. How well the experiments reflect the varied real world applications remains to be seen, but I think the real value of this paper is that of challenging common assumptions by presenting evidence on the contrary. We should not assume that a more user-friendly interface is necessarily better. Indeed–better for what, or for whom?

Socrates warned us about the dangers of writing. Today we face similar uncertainties with computers, eager to help relieve us of our burdens–but we have to ask: at what cost?