Continuing from the previous module, we will cover some theories and concepts in this lecture note. We will discuss models of people's abilities like perception (e.g., seeing) and cognition (e.g., memorizing), topics that have been studied by cognitive psychologists.
So why should we, people studying how to design a system and its user interface, care about models of people’s abilities? One reason is that they offer you rules of thumbs or common sense of good design. We have been learning empirical ways of studying what is good design (e.g., observing how people use existing system, conducting usability testing). While it is critical to engage with the users, running observational studies and experiments takes time, effort, and money. If we can tap into prior knowledge of “good design”, we may be able to create good design more efficiently.
A large body of cognitive psychology literature teaches us about our abilities like attention, perception, memory, learning, and motor skills. We seem to combine all these abilities to do complex tasks like problem solving and decision making in our daily lives, including interactions with software systems. Such knowledge could be useful in drawing insights for “good design”, but the sheer size of the body of knowledge make it hard to apply the accumulated insights into the design of software systems (Card, 1986). What would be useful is approximated models for design and engineering that allow us to approximate people’s abilities and behaviors, that are simple enough and useful for designing.
Models we learn are an abstraction of how a user interacts with a computer. It helps us get a rough idea of how a user performs a series of tasks. Here, a task means something that a user has to do in order to achieve his/her goal. A top-level task can be broad, and we can divide it into smaller sub-tasks. For example, if the top-level task is a “find information on a web site”, then you can break it down into into “finding a right page” and “reading the page”, and these subtasks could further be subdivided into things like “enter a search keyword into an input area and hit enter”, “look for a direct link by scanning a page, then use a mouse to move a cursor and click a link”, and so on (Dix pp.519). A model gives an estimate of how much effort, measured in metrics like task completion time, is needed to perform such sub-tasks. And by aggregating all the required effort for all the subtasks, you could estimate effort needed to perform the top-level task. Such a model of human-computer interaction allows us to give us a rough idea to questions like “how much complexity is acceptable in a given user interface” and “how can we train/instruct the user so that they can learn to effectively use our system”.
The human processor model. (Card, Moran, and Newell, 1983)
The model human processor developed by Card et al. (Card et al., 1986) describes how long a user takes to perform a cycle of a task using a computing device. It aims to estimate time needed to perform a series of input/output interaction between a user and computer.
Using literature from cognitive psychology, the model estimates time needed for a user to “plan a motor control”, “perform input”, “perceive output”, and “process the output”. Card argued that such a model is useful for estimating “how fast people can consume written information on a document” and “how long it takes for a user to perform a decision making task”. See (Card et al., 1986) for more information.
Model human processor was a nice abstraction for human-computer interaction for motor- and perception-heavy tasks. But it was not designed to approximate effort need to perform a task taxed cognition. For example, if a task requires a lot of time for a user to think and remember some information, effort needed for perception (e.g., reading) and motor control (e.g., typing) becomes negligible compared to time needed to think.
People were curious about modeling people’s cognitive ability. Prominent research on human cognition describes how people’s reaction time evolves as they get exposed to tasks (power law of practice) (Newell and Rosenbloom, 1981) and how people’s memory work (Chase and Simon, 1973).
Newell, A., & Rosenbloom, P. S.
The implication of the power law, for example, was simple yet interesting; “Almost always, practice brings improvement, and more practice brings more improvement”, and “plotting the logarithm of the time to perform a task against the logarithm of the trial number always yields a straight line, more or less” (Newell and Rosenbloom, 1982).
This gives an insight for the design process. While we persuade ourselves to come up with innovative design ideas, we should also encourage ourselves to strive for (potentially boring) consistent design and follow design standards that other people follow. The power law of practice suggest that the user gets better at performing a given task as they practice it more often. And the rate at which you get better follows the power law distribution. Following the design standard that everyone else follows allows the users to learn the design convention in using other interfaces, and the learning would transfer when they perform a task using your user interface. You should come up with a new design only if you have a strong reason to deviate from a common design.
The model human processor suggested that we are much better at retrieving information from the working memory than accessing the long term memory. But the working memory has smaller capacity compared to the long-term memory. Then it would be a good idea to reduce the cognitive load for the user by designing a task so that people don’t have to memorize so much things. But just how much information can we store in our working memory?
Early research suggests we can store in short term memory is 7 objects or so. But what is the unit of the "7 objects"? Chase and Simon hypothesized it is a "chunk." Chunks are any stimuli that have become familiar through repeated exposure and hence can be recognized as a single unit (Chase and Simon, 1973). As it is discussed in the video, this implies that we can pack more information by organizing and group information in a way that is easier for people to memorize and process.
These models have invited user interface design considerations (Preece et al., 2015), including: