HCI

HCI Notes

Interaction techniques: Data files (punch cards). Command lines. Line editor (rather than replacing whole punch card), Modeless interface (in different contexts commands have same effect). WSIWYG. Menus (easier to recognize than recall). Pointing devices. Graphic displays (rather than teletype).Windows and Icons.

Direct Manipulation Properties (for WIMP rather than command line):Object of interest visible graphically. Operations involve physical actions not commands. Actions should be rapid , incremental and reversible. The effect of actions immediately visible. Small set of novice commands, expanded expert commands.

Style guidelines: Microsoft Style guidelines similar to DMP above. Used by apple for branding, familiar interfaces.

Heuristic evaluation: Experts evaluate a system from a list of usability heuristics.

Psychological User Models:

Marrs theory of vision: Marr proposed a series of black boxes (input,output) to process visual input. Retinal image -> (nerve cells find boundaries) ->Primal sketch -> (Analyse, find which objects infront of each other) -> 3d sketch.

Gestalt laws of perception: Proximity (Objects close to each other are grouped). Similarity (Visually similar objects are grouped). Good continuation (Dotted lines appear continuous). Closure (Prefer to see common shapes).

Visual Tasks: Depth perception. Face recognition. Visual search, takes T = k log_2 (n+1) to find one item among a number of similar items.

Fitts law: Time taken to point at a given location greater with distance moved by mouse, less with size of target.

Models of memory: We can remember about 7 +- 2 chunks (eg words, letters). We remember things by creating connections, so producing rich associations aids retrieval.

Generalised Problem Solver: Searches a state space for possible intermediate states between some initial and goal state. At each stage either select and intermediate goal that gets you closer to the goal, or if no such goal decompose into subgoals.

System Models:

Model human processor: Summate all different events for an action.Perceptual events=100ms. Cognitive events=70ms. Motion events=70ms.

Keystroke level model: Similar to above. Key press: 0.28s. Home (on mouse or keyboard): 0.4 seconds. Point with mouse: 1.1 seconds. System response time. Mentally prepare for an action: 1.35 seconds.

GOMS (Goal, Operators, Methods and Selection): Extends KLM with GPS idea. Users select a method which consists of operation to complete a goal.

Mental models: Users simulate what effect they think actions will have. If users can't explain something by their mental model, they use analogy. Eg people think that electricity is like water, and will fall out of the socket.

User Orientated Design Methods:

Prototyping: The waterfall model requires strict and correct specifications.Rapid prototyping accepts users requirements change and evolve. With the spiral model the prototype is refined iteratively. In deep prototyping one aspect of the system functionality is fully implemented before developing the rest of the interface. You can use rapid development tools such as Macromedia Director, or even low fidelity prototyping (drawings).Functionality in low fidelity prototypes can by provided through the Wizard of Oz technique (a person pretends to be the machine).

Evaluation techniques:

Controlled Experiments: An experiment is based on observations (measurements made when someone uses an experimental interface). It should be performed over many subjects and results are usually found to have a normal distribution. There are normally experimental treatments, eg a good UI vs. a bad UI. To find the difference we need a significance test such as the t-test. The null hypothesis is "What is the probability that the observed difference is due to random variation?". It is normally required the chance is less than 5%. There are variations due to different tasks, IQ, hints etc. The experiment should be made under a realistic environment.

Surveys: Normally composed of closed (yes/now, 1-5) or open questions.Questionnaires are a type of survey, normally used for large populations.

Think aloud studies: Subjects are asked to carry out some task while talking as continuously as possible. Data is transcribed from a tape into a verbal protocol.

Bad techniques: Subjective reports. One user (the programmer). Hypothesis without analysis.

Cognitive Walkthrough: Under the user model a user sets a goal to be accomplished. The user searches the interface for possible actions, selects the action that seems likely to make progress towards the goal, then performs and evaluates the systems feedback. Users aren't involved; evaluators describe a description of users, tasks to be completed and the correct actions for them. Present are the interface designer, a group of peers, a scribe and a facilitator. They consider the users goal, accessibility of controls, the quality of the controls label and goal, and the feedback of the system. Cognitive walk-through deals with inexperienced users performing possibly new actions.

Heuristic Evaluation: Nielsen suggested that the usability of a system should be evaluated by a number of experts, each working from a list of usability heuristics. Separate evaluators, if possible from different backgrounds, each evaluate the system independently perhaps by running through scenarios.All evaluators use the same heuristics. Nielsen gave a list of sample heuristics including “Visibility of system status”,”Match between system and the real world”, “User control and freedom”, “Consistency and standards”, “Error prevention”, “Recognition rather than recall”, “Flexibility and efficiency of use”, “Aesthetic and minimalist design”, “Help users recognize, diagnose and recover from errors”, “Help and documentation”. Heuristic evaluation is used often as it is cheap, simple and easily understood. It doesn't however touch deeper issues or offer answers for them, and doesn't have concepts of different users. Unlike with walkthrough, evaluators work separately.

Task Orientated Analysis:

Structured Interviews: Normally find requirements from representative. Also conduct studies such as interviews with the users.

Observational Studies: Video recordings of the subject, which are transcribed into a video protocol. Can also do diary studies.

Ethnographic field studies: Observe subjects in a wide range of contexts, over a period of time. Useful for requirements capture. Microsoft write personas from market research (with photographs, fictional biographies etc.) then the designer tries to accommodate this user.

Field tests: Such as the follow me home program by Quicken, where users are given the software free in exchange for designers following them home and observing them install and use it.

Use case design: Analyse user activities in terms of use cases - specific scenarios for interactions with the system. The systems behavior is described from the point of view of actors- representations of roles that users will take when interacting with the system. Useful in later stages of design, to trace the specification and validation of different subsystems.

Cognitive Dimensions of notations: The Cognitive Dimensions of Notations is an approach to analyzing the usability of information artefact's (eg software systems, physical machines etc.) It is a small vocabulary of about 12 terms which describe aspects that are cognitively-relevant. Such as: Premature commitment (can you perform actions in any order?), Hidden dependencies (eg HTML links), Viscosity (easy to change?), Visibility (Easy to see what you're editing?), Closeness of mapping (between notation and the real world), Consistency (similar semantics?), Diffuseness (brief notation?), Error-proneness (some mistakes easy to make?), Hard mental operations, Progressive evaluation, Provisionality (can you sketch?), Role expressiveness. There are also sub-devices: Helper devices (eg post-it notes), redefinition devices (eg short cut codes). There are notational activities: search, incrementation (adding new data), modification (changing structure), transcription, exploratory design. To evaluate, identify the main notation and environment in which it is manipulated. Then identify sub devices and describe their notations (there may be separate layers to be analysed separately). Consider each notation in terms of dimensions.Identify problems, and consider design decisions. Useful for complex software with notations such as spreadsheets.

Research trends:

Computer Supported Cooperative Work: Users normally work with other users. Design discussions, for example, can be broken down into questions, options for addressing them, and criteria by which an option should be selected. This provides a design rationale (reasons why design decisions were taken). Examples of what is looked into include shared whiteboards and social networking sites.

New interaction devices: Immersive virtual reality, augmented reality (superimpose virtual onto real). Audio responses to actions.

End user programming: Psychology of programming. Useful for things such as laboratory automation language. Programming by example is where a computer infers what the user wants. Also programming languages to teach programming.