GDC2013 - Usability Is Not Random

• This talk was given by Mike Acton, director at Insomniac Games. The goal was to develop a mental model for recognizing and breaking down usability issues. It is not a numbers or deep talk.
• Usability should be applied to tools UI. Actually, usability should be applied to game, API design, and everywhere...
• The main lessons of this talk are...

1. ​Only way is to watch it live
2. Choice is bad
3. Choice is good
4. 20 questions is all you need
5. Learning curve is bad
6. Learning curve is good
7. When in doubt, be like Maya

• ​The five functions of usability are...

1. ​Context
2. Information density
3. Message creation
4. Message translation
5. Message response

• Usability is the "total time required to solve the specified problem at the desired level of quality." There are many sophisticated measuring tools, but the most important one is to watch your testers live and ask good questions.
• Context is "all agreed upon information." Agreeing on what we're trying to accomplish will add more context and usability. Take Google for example. When doing a search, you have to visually parse the results to find what you're looking for. Clicking the "Image" tab gives the search more context of what you're searching for. More constraints makes it potentially more usable.
• The upper limit of usability is determined by the amount of agreed upon informaton we start with. In this case, choice is bad. Google, with no context, gives us too many results that it becomes less usable. If you're specifically looking for computational knowledge, searching on Wolfram Alpha (more context) will directly give you the result you're looking for.
• There are 2 kinds of contexts: the hunt (where you know what the end result is) and play (where you don't know what the end result is). When you're "playing," you have an idea of what you want, but you don't necessarily have a specific thing you're looking for. In this case, choice is good. Common instances of play can be found with Amazon and Netflix, which present lots of relevant, comparable options quickly to the user, increasing its usability.
• Information density is "the amount of problems that can be solved with any particular message." It is the data that is associated with each step in the process. Every problem is just a game of 20 questions, narrowing down the list of possibilities and page results. For example, looking at Google again, at the start of a search, there are infinite possibilities to what you're searching for. The moment you type the letter "a", it reduces the possibilities to about 6,500. You add a second letter and it reduces that number of possibilities dramatically again.
• Message creation is "the amount of time required to react correctly." The things to measure here are parsing time, thinking time, and error rate. Common problems to message creation are the need to do calculations in your head, the need to choose between unknown options, and having options that are too granular or useless. Learning curve is bad, so reduce the time it takes to learn what to do.
• But initial difficulty isn't always bad. There are plenty of software, games, and activities that have high initial difficulty (programming languages, 3D modeling software, real-time strategy games, fighting games, learning a musical instrument). History has shown that dedication to a difficult task produces better mastery of the task. In this case, learning curve is good. People get better with practice.
• Message translation is "the latency to transform a message to a useful result." This is basically the time it takes between a user interaction and the result being shown, whether that is building lightmaps or compiling a program. Follow the 5 second rule. One iteration takes 5 seconds, so one minute wasted is a loss of 12 iterations. Make things more usable by reducing the time needed to work out what the right thing to do is.
• Message response is "the throughput of the message." It is the time until the next message can begin.
• When in doubt, be like Maya or a UI that users are already familiar with. The more different your tool is, the bigger the bridge your users have to cross in order to use it.
• Usability is a triage. You simply can't do everything because you don't have infinite resources or time. Don't try to maximize usability in everything.
• References

1. ​"Mathematical Theory of Communication" - Claude Shannon
2. "Design of Everyday THings" - Donald Norman
3. "Man Who Lied to His Laptop: What We Can Learn About Ourselves from Our Machines" - Clifford Nass, Corina Yen