1. CSC 341 Human-Computer Interaction
Prepared by
Dr. Mai Elshehaly
Main reference for this lecture:
MacKenzie, I Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher:
Morgan Kaufmann. ISBN: ,

2. Outline
Tasks (again and again!)
Modeling interaction

3. How can you tell when your tasks are good?

4. Good Tasks
Are good at:
Abstraction
Generalization
Evaluation

5. Good Tasks Are good at: Abstraction  no domain-specific details
Generalization  apply to different usages by different people
Evaluation  ????

6. Tasks are good for evaluation if…
They represent:
A task that is similar to actual or expected usage will improve the external validity of the research–the ability to generalize results to other people and other situations.
They discriminate:
A task should elicit different behavioral responses that expose benefits or problems among the test conditions. This should surface as a difference in the measured responses across the test conditions

7. Example from midterm
Sameh is a manager in a bank. He is comfortable using smartphones and finishes much of his work using his phone. He is focused and goal-oriented, but he has many customers which makes his job sometimes challenging. He needs to keep track of the interests of each customer (e.g. real-estate, stock market, etc.). He wants to send them customized news and updates. He also needs to stay up to date with news about the stock market, currency rates and other daily information. Additionally, he wants to be reminded of meetings, conferences, and social events that are scheduled with his co- workers and his customers.

8. Example from midterm
Sameh is a manager in a bank. He is comfortable using smartphones and finishes much of his work using his phone. He is focused and goal-oriented, but he has many customers which makes his job sometimes challenging. He needs to keep track of the interests of each customer (e.g. real-estate, stock market, etc.). He wants to send them customized news and updates. He also needs to stay up to date with news about the stock market, currency rates and other daily information. Additionally, he wants to be reminded of meetings, conferences, and social events that are scheduled with his co- workers and his customers.

9. Bad Task Specification: simple copy-paste
T1: Keep track of the interests of each customer (e.g. real-estate, stock market.)
Does it represent? (What about other interests? Other customer information?)
Does it discriminate? (How do you measure the error and time it takes someone to “keep track” of something?)
We need a better task specification

10. Andrienko’s Functional View of Data
Data is structured into two main components:
Referential components: R
Represent contexts in which measurements and observations are made
Characteristics: C
Represent observations or measurements taken in context of the referential components
Source: Andrienko, N., Andrienko G. (2006). Exploratory Analysis of Spatial and Temporal Data: A Systematic Approach. Publisher:Springer-Verlag Berlin Heidelberg. ISBN

11. Back to our example
T1: Keep track of the interests of each customer (e.g. real-estate, stock market.)
Referential components: R
Customers
Characteristics: C
Interests, age group, gender, geographical area, profession, etc.

12. Andrienko’s Functional View of Tasks
Any task (question) implies two parts:
a target, i.e. what information needs to be obtained, and
the constraints, i.e. what conditions this information needs to fulfil.
The target and constraints can also be viewed as unknown and known (specified) information, respectively;
The goal is to find the initially unknown information corresponding to the specified information.

13. Formulating a Task
User objective: Keep track of the interests of each customer (e.g. real-estate, stock market.)
User Task (as a question):
T1: What are the (interests/communication preferences/other information) pertaining to customer X?
User Task (as a process):
T1: Overview the (interests/communication preferences/other information) pertaining to customer X

14. Other possibilities… Filter… Arrange… Change… Select… Aggregate…
Overview…
Lookup…
Identify…
Summarize…
Explore…
Navigate…
Browse…
Compare…
Locate…
Filter…
Arrange…
Change…
Select…
Aggregate…
Annotate…
Import…
Derive…
Record…

15. Modeling Interaction

16. What is interaction?
Interaction happens when a human performs a task using computing technology

17. What is interaction?
At a stripped-down level, interaction involves humans using their sensors and responders to monitor and control devices, machines, and systems that include computing technology.

18. What is Modeling?

19. What is Modeling? Descriptive Models Predictive Models
Models using “loose verbal analogy and metaphor” to describe phenomena
Ex.: An architect’s model of a building
Models using “closed-form mathematical equations” to describe phenomena
Ex.: A physicist’s equation of a trajectory of a tossed ball

20. What is a Descriptive Model?
A design space
A framework
A taxonomy
A classification
An organization of the subdivisions of a problem space

21. Example of a Descriptive Model
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

22. Some Descriptive Models of HCI
The Quadrant Model of Groupware
Key-Action Model (KAM)
Model of Bimanual Control
Three-state Model for Graphical Input

23. Quadrant Model of Groupware
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

24. Key-Action Model (KAM)
Have you ever thought about the operation and organization of keys on a keyboard?

25. Keyboards Contemporary keyboard layouts: QWERTY Dvorak
– layout: frequently used letter pairs far
apart (increased finger travel distances)
– used by all English-language keyboards
– trained users: up to 150 words per minute
Dvorak
– layout: vowels on the left, most common
consonants ('D','H','T','N','S') on the right
– decreased finger travel distances
– trained users: up to 200 words per minute
– ± 1 week needed to get used to the layout

26. Descriptive Model for Keyboard Layout
KAM categorizes keys on a keyboard into one of three types:
Symbol Keys: deliver graphic symbols to an application such as a text editor (e.g. letters, numbers, punctuation)
Executive Keys: invoke actions in the application or at the system level (e.g. ENTER, F1, etc.)
Modifier keys: set up a condition that modifies the effect of a subsequently pressed key (e.g. SHIFT, ALT)

27. Key-Action Model (KAM)
Why do we care?
With a 3:18 left-right ratio of executive keys, the desktop keyboard is clearly entrenched with a right-side bias  the right hand is kept busy
What if the right hand is gripping the mouse and there is a need to press a right-side executive key?
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

28. Model of Bimanual Control
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

29. Model of Bimanual Control
Why do we care?
Scrolling is traditionally accomplished by dragging the elevator of the scrollbar positioned along the right-hand side of an application’s window.
Acquiring the elevator is a target acquisition task taking up to two seconds per trial.
Users should not be required to divert their attention from the primary task (reading, editing, drawing, etc.) to acquire and manipulate user interface widgets.

30. Model of Bimanual Control
Why do we care?
Desktop affordances for scrolling changed dramatically in 1996 with the introduction of Microsoft’s IntelliMouse, which included a scrolling wheel between the mouse buttons.
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

31. Notice something? Design Decision:
Should scrolling be performed by the right hand?
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

32. Scrolling on the left
Image Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

33. Pointing devices
Pointing devices are used to point at and select items.
• Direct-manipulation approach – faster, fewer errors, easy to learn
• Pointing devices’ tasks:
selecting an item
dragging and positioning an item
orienting (rotating) an item
defining a path / curvature
text writing / editing
• Pointing devices can have:
direct control on screen surface
indirect control away from screen surface

34. Pointing devices Direct-control pointing devices: Light pens
– can be used for any pointing device task
– obscure the screen, cause arm fatigue
Touch screens
early designs (imprecise): physical pressure, interruption of a grid of infrared beams
recent designs (high precision): interruption of ultrasonic waves, optical imaging (touch shows as a shadow), calculating mechanical pressure on the glass
Stylus (Buttons)
– Widespread in PDAs (personal digital assistants)

35. Pointing devices

36. Three-State Model of Graphical Input
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

37. Three-State Model of Graphical Input
Why do we care?
Can guide the design of interaction techniques based on sequences of state transitions of pointing devices.
Ex.: One of the interaction techniques supported by touchpads is “lift-and-tap,” where primitive operations like clicking, double-clicking, and dragging are implemented without a button.

38. Advantages of a Descriptive Model
Has a name (makes it easier to talk about interaction)
Delineates a problem space
Exposes design problems and suggest opportunities
Ability to identify common elements for interactions located in the same area of the problem space

39. Predictive Models
An equation that predicts an outcome based on an independent variable

40. Example: Is there a relationship between stylus tapping speed and touch typing speed?
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

41. Yes there is
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

42. Some Predictive Models in HCI
Fitts Law
Choice Reaction Model
Leystroke-level model (KLM)

43. Fitts’ Law
Fitts was an experimental psychologists interested in applying information theory on human performance.
Argued that:
Amplitude of an aimed movement ~ electronic signal
Spatial accuracy of the move ~ noise

44. Fitts’ Law Experiment 1: (1954)
a serial, or reciprocal, target acquisition task where participants alternately tap on targets of width W separated by amplitude A
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

45. Fitts’ Law Experiment 2: (1964)
subjects selected one of two targets in response to a stimulus light
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

46. Effective Target Width
Fitts’ Law
ID, the index of difficulty
Effective Target Width
If using Fitts’ law as a predictive model is the goal, then the movement time (MT) to complete a task is predicted using a simple linear equation:

47. The slope and intercept coefficients in the prediction equation are determined through empirical tests, typically using linear regression.
The tests are undertaken in a controlled experiment using a group of participants and one or more input devices and task conditions

48. Fitts’ Law Why do we care? Has three uses in HCI:
To learn if a device or interaction technique conforms to the model by building the prediction equation and examining the correlation for “goodness of fit,”
to use the prediction equation in analyzing design alternatives, or
to use Fitts’ index of performance (now throughput) as a dependent variable in a comparative evaluation.

49. Experiment using Fitts’ Law
Twelve participants performed a series of reciprocal point-select tasks across nine target conditions, using two different pointing devices (RemotePoint and a mouse).
A = 40, 80, 160 pixels crossed with W = 10, 20, 40 pixels.
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

50. Experimental Results
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

51. the standard deviation in a participant’s selection coordinates
The effective target width (We) adjusts W to reflect a 4% error rate
the standard deviation in a participant’s selection coordinates
Fitts’ index of performance, now called throughput (TP, in bits/s), is calculated by dividing ID (bits) by the mean movement time, MT (seconds), computed over a block of trials:

52. Experimental Results
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

53. Fitts’ Law Why do we care? Has three uses in HCI:
To learn if a device or interaction technique conforms to the model by building the prediction equation and examining the correlation for “goodness of fit,”
 We learned that both devices conform to Fitts’ Law (high R2)

54. Fitts’ Law Why do we care? Has three uses in HCI:
To learn if a device or interaction technique conforms to the model by building the prediction equation and examining the correlation for “goodness of fit,”
to use the prediction equation in analyzing design alternatives, or
 We learned that RemotePoint performed poorly compared to the mouse

55. Choice Reaction Time (CRT) Model
Stimuli
Source: MacKenzie, I. Scott (2013). Human-Computer Interaction: An Empirical Research Perspective. Publisher: Morgan Kaufmann. ISBN: ,

56. The Keystroke Level Model

57. Announcements Project Phase II: deadline extended to November 24th
Read Ch. 7 in Mackenzie’s book (link on website)
Assignment (deadline November 27th)
Implement a simplified version of Fitts’ first experiment using D3.