CMSC434: Human Factors in Computer and Information Systems

CMSC434: Human Factors in Computer and Information Systems

University of Maryland, Department of Computer Science

Prof. Ben Shneiderman, Spring 1998

Exam #1, Solutions

Part I:
Q1. List the three principles of direct manipulation and discuss each in the context of OAI model (15)

The three principles of direct manipulation are:

Continuous representation of the objects and actions of interest with meaningful visual metaphors.

Physical actions of presses of labeled buttons, instead of complex syntax.

Rapid incremental reversible operations whose effect on the objet of interest is visible immediately.

Continuous representation of the objects allows the user to see the effects of user actions on the object immediately (100 msec). An effective direct manipulation design would directly map objects/actions in the task domain to corresponding objects/actions in the interface domain. Meaningful metaphors when applied in the interface domain allows users to make associations between the interface actions and objects and the high-level task domain. Physical actions (clicking or dragging) are employed to interact with objects in the interface domain, instead of written complex syntax, which gives users the visual feedback to directly manipulate objects in the task domain. Rapid, incremental, and reversible actions with immediate feedback follow the physical model of the real world (i.e. task domain).

Q2. Compare command language interfaces to direct manipulation interfaces with respect to compactness, speed of performance and learnability (15)
Compactness:

Demands of textual vs visual objects:

Typically textual objects (e.g. ls, file1, prog1) takes less screen space compared to visual objects (icons, etc.)

Speed of performance:

Task complexity, repetition:

Typically simple tasks (e.g. copy, move) are performed faster in direct manipulation interfaces, where complex tasks (e.g. ps aux | cut –1,4 –d"x") are performed faster in command languages. Repetitive tasks tend to have simplifications in command languages (e.g. ls –la *ks?.txt).

User expertise:

In general novice users perform tasks faster in direct manipulation interfaces as opposed to command language interfaces. As users get more training the situation might change for certain tasks.

Error frequency:

Direct manipulation interfaces typically eliminate situations in which users can make errors. Command language users make a lot of errors, such as typos or wrong commands, because the syntactic complexity is difficult to retain without frequent rehearsal. Errors increase the time to perform tasks.

Mouse keyboard transition:

In direct manipulation interfaces users typically switch between the keyboard and the mouse frequently, which may decrease user performance.

Learnability:

Syntax:

There is a lot of syntax in command language interfaces, which make such interfaces hard to learn.

Visual Language, metaphor:

Visual language (dragging, selection, icon images) may also require some learning by the user, and depends on cultural characteristics as well. Use of metpahors may also need some learning.

Q3. Compare and contrast controlled psychological experiments and usability tests in the evaluation process of user interfaces with respect to three points of your choice (15)

Multiple vs Single application/interface/idea

User experiments compare two or more applications, whereas usability studies focus on a single application

Proof of hypothesis vs. Refinements to design

Experiments try to prove a hypothesis, whereas the purpose of usability studies is to improve an existing design.

Quantity of subjects

Typically more subjects are needed for psychological experiments.

Controlled, specific vs. Random nature

Experiments are controlled and measure very specific variables, whereas valuable improvements in usability studies are typically discovered in a somewhat random manner.

Time to complete evaluation

Experiments typically take longer to complete.

Q4. List four of the "Eight golden rules in interface design" and briefly describe each in one or two sentences, with an example. (20)

Consistency

Shortcuts for frequent users

Informative feedback

Design dialog to yield closure

Error prevention and simple error handling

Easy reversal of actions

Locus of control

Reduce short-term memory load

Part II:
An image browsing tool is very useful for manually browsing large collection of digital images. It allows users to review large numbers of images simultaneously in a rapid manner while searching for a particular image or set of 500 images. When designing image browsers typically decisions have to be made as to how to use the available screen area most effectively to achieve an acceptable level of user performance: speed and accuracy. Some say many small thumbnails are better than a few larger ones. Three sizes have been proposed: small, medium, and large.
Assuming substantial resources, design a controlled experiment to resolve this issue.

Q1. Describe the hypothesis of the experiment. (5)

Null hypothesis: In an image browser, users image-retrieval task performance time and accuracy does depend on the number and size of images

Hypothesis: As the number of images on the screen increases (many small images) users performs image-retrieval tasks faster than (few large images). Accuracy in image-retrieval tasks increases as the size of images increase.

Q2. Describe the independent and dependent variables and their treatments. Be very specific about user tasks, and controls. (15)

IV: Screen layout (size and quantity) with three treatments

Many small images (50x50 in 8 rows and 8 columns)
Medium quantity of medium size images ( 100x100 in 4 rows and 4 columns)
Few large images ( 200x200 in 2 rows and 2 columns )

DV: User task completion time (in ms), accuracy (percent of correct images found)

User tasks:

Given an image shown in resolution 300x300, find and click on the same image from the set of image samples shown. You may use previous and next buttons to change the current image set.

Repeat the above task for a randomized sequence of images 40 times in all three treatments. Counterbalance the order of treatments.

Controls:

Window size, user expertise, resolution, image content (similarity among images, user familarity, color, etc.), image sequence, target image sequence, spacing and alignment of images, etc.

Q3. Briefly describe which statistics should be used and why. (5)

A one-way ANOVA (three treatments) statistics, along with average and standard deviation can be used to interpret the results of the experiments. The analysis involves only a single variable with three treatments.

[Note: If your design involves more than one variable with more than two treatments, then a T-Test will not work, you should use ANOVA]

Q4. Predict the results of the experiment and discuss them in a graph. (10)

My prediction is that as the number of images on a single screen increase users task completion time wil increase due to the need to use previous and next buttons. However, I also predict that it will take more time to recognize the image when the image size is small. So, in terms of task completion time, there will be a global minimum at the 100x100,4x4 screen design.

My prediction is that as the size of images increase users’ accuracy will increase, yielding a higher percentage of correctly retrieved images. However, this effect is very likely to level off after a certain image size.