This is an online version of a graduate course at Georgia Tech, CS6750.
HCI is about humans interacting with tasks, mediated by a computer; or, alternatively, about humans and computers together interacting with tasks.
This idea seems related to the concept in Computers as Theatre.
Computers are everywhere: phones, cars, tablets, video games, microwaves, and more.
Depending on the design, an interface may be more or less visible. The ideal would be an interface that totally disappears, allowing the user to feel as though she interacts with the task directly. Games are an example of such an interface, granted the user is fluent with the controls. On the other hand, for some tasks the interface is very visible, as with controlling a home theater system with multiple remote controls.
HCI is a subfield of Human Factors Engineering, and includes its own subfields, such as UI Design, UX Design, and Interaction Design. HFE is concerned with both engineering and psychology.
Compared with Human Factors Engineering, HCI is interested in only computer aspects of interaction, where HFE is interested in broader interactions, such as physical layout of non-computer objects.
User Interface Design is primarily concerned with screen interfaces, while HCI covers interfaces of other kinds as well.
HCI is about understanding user interaction with computers, while UX Design is about dictating such user interaction.
HCI involves an iterative process of using research to create designs, and applying what is learned from the use of those designs to futher research.
By effective interactions is meant interactions that support the goal of the design. Typically this would be usability, but it may also be to support research--in which case a less usable design might be created to gain insights into its use--or to produce changed behavior in the user, such as a thermostat designed to cause the user to save energy.
HCI is used in many areas. This lesson gives a brief overview, to give you something to think about while going through the course. The divisions below are somewhat arbitrary.
HCI considers the user interacting with a task, mediated by the interface--not the user interacting with the interface per se.
Tips to identify a task:
An interface is useful if it allows a user to accomplish some task, but we wish to design interfaces that are usable, not merely useful.
In HCI, the user might be viewed in three ways: as a processor, as a predictor, or as a participant. Each of these suggests different requirements for the interface, and each may be evaluated using different methods.
As a processor, the user is like another computer participating in the actions, processing and responding to eevnts. The interface must be physically usable. Evaluate with quantitative studies. This view is influenced by the behaviorism school of philosophy, which is concerned only with observable behaviors and outcomes.
As a predictor, the user should understand what the results of their actions will be. The interface must make sense to the user. Evaluate with qualitative studies. This view is influenced by cognitive psychology, which is concerned with what happens inside the mind.
As a participant, the user's whole experience, including other tasks they are performing and tools they are using, should be considered. The interface should fit with the user's context. Evaluate with in situ studies. This view resembles functionalism or systems in psychology.
This processor model is helpful for optimizing existing interfaces, particularly for improved efficiency for expert users, but does not give insight into why novice users might fail.
The predictor model can target different levels of expertise, and by talking to the users, it is possible to learn about their thought processes and any misconceptions they may have. However, this kind of research is more expensive than quantitative research and is vulnerable to bias from the researchers. Additionally, it might miss complications due to context.
The participant model evaluates the interaction in a realistic context, so it gives a more complete picture of the interaction. Like the predictor model, this may be expensive and difficult to research and analyze, and its strength--the realistic context for interaction--is also a weakness, because it opens the study up to confounding factors. Additionally, since this model requires a complete interface to be available, it cannot be used during initial design.
The gulf of execution is the distance between a user's goals and the actions required to realize those goals. How does the user figure out what to do to accomplish their goals? What is the distance between what the user thinks they have to do and what they actually have to do?
Five tips about the gulf of execution:
The gulf of evaluation is the distance between the effects of those actions and the user's understanding of the results. Three components or phases of this are the interface output, the interpretation of that output, and evaluation, in light of the interpretation, of whether the goal was accomplished.
Five tips about the gulf of evaluation:
With a direct manipulation interface, the user should feel as though they are interacting directly with the object of their task. For example, when zooming an image on a phone, it seems like you are interacting with the image itself, not an intermediate interface.
The seminal paper on direct manipulation interfaces, from 1985, is:
It can be found here.
The example is given and revisited of moving files around. Following the desktop metaphor, we should expect moving files around on a computer to be similar to moving files around on a desktop. However, this insistence, I think, is itself something of a failing: there is no special reason why we should use this kind of interaction for dealing with our data. There are file systems that don't have folders, and there are systems that support tagging of files (even Windows supported this for many years, though the interface for it seems to have been de-emphasized, at least) so that they can be found without navigating a directory structure.
Actually, this is a good point: the 'files and folders' metaphor clearly begins to fail when you have folders nested in folders nested in folders ad infinitum. Sure, we could try to extend the metaphor by calling the outer folders file drawers, or cabinets, or whatever, and there's no reason why we can't understand a recursive structure here, but at that point we no longer have the interpretation "there is a folder on my desk with a group of related files in it", and instead what we seem to have is more like a directory with cross-references. It seems to me that the whole metaphor should be re-examined.
An invisible interface is one that seems to disappear--the user can forget that they are using an interface at all. This could be due to learning, such as turning the steering wheel when driving, or it could be by design.
Five tips about invisible interfaces:
Women are color blind at a rate of about 1 in 200, while men are color blind at a rate of about 1 in 12.
We will consider three kinds of memory: the working memory, the short-term memory, and the long-term memory.
The working memory or perceptual store holds the most recent experiences and lasts less that a second.
The visuospatial sketchpad holds visual images, the phonological loop holds audio, and the episodic buffer integrates these chronologically. The three are coordinated by a central executive.
Users can store four to five chunks of information at a time. These chunks can be understood as a single unit, such as a single word, as opposed to a collection of random letters.
This limit is the same one famously described in The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information, though the specific number given differs. The lecture claims that we have found that the 'true' number is really four or five. The text recommended in this course, Human-Computer Interaction: An Empirical Research Perspective, still says seven, however.
The example is given of a Leitner system.
There are two kinds of knowledge: procedural knowledge, which is how to do something; and declarative knowledge, which is knowledge about something.
Cognitive load is the amount of the brain's cognitive resources that are in use. Different tasks will require different amounts of a user's cognitive resources, and attempting to perform a task when insufficient resources are available may yield poor results. So it is important to understand the full context in which an interface will be used, in order to estimate the user's cognitive load, and to minimize the cognitive load caused by the interface.
Five tips for reducing cognitive load:
We must be aware of how quick and precise a user can be, and we should be tolerant of errors.
We will consider four sets of principles for design:
For this lesson, these are merged into fifteen principles:
There are other principles in a variety of works:
Dix, Finlay, Abowd, and Beale proposed three categories of principles: learnability for how easily a new user can grasp an interface, flexibility for how many ways an interface can be used, and robustness for how well an interface gives feedback and recovers from errors.
In Human-Computer Interaction.
Jill Gerhardt-Powals has a list of principles for cognitive engineering, aimed especially at reducing cognitive load. Her list is in particularly useful applications for data processing and visualization.
In "Cognitive Engineering Principles for Enhancing Human-Computer Performance".
In "The Humane Interface", Jef Raskin outline some additional revolutionary design rules. I wouldn't necessarily advocate following them, but they're interesting to see a very different approach to things.
In "Computer Graphics Principles and Practice", Jim Foley and others give some principles that apply specifically to 2D and 3D computer graphics.
Susan Weinschenk and Dean Barker have a set of guidelines that provide an even more holistic view of interface design, including things like linguistic and cultural sensitivity, tempo and pace, and domain clarity.
In Designing Effective Speech Interfaces.