Fitts Law and Tactile Feedback

1
Fitts Law and Tactile Feedback

• Maureen Duffy
• October 29, 2003

2
Overview

• INTRODUCTION
• Whats this all about?
• JUSTIFICATION
• Why do this? How will it contribute?

3
Overview

• NITTY-GRITTY
• Fitts Law
• Types of Tasks
• Exisiting Research
• How Tactile Effects Affect Acquisition Time and
  Error
• How Friction, Texture, and Friction Texture
  Affect Acquisition Time
• How Magnitude Affects Acquisition Time
• Psychological Theories on the Haptic Visual
  Modalities
• People in this Research Area
• Issues for Consideration

4
Introduction

• What is Fitts Law?
• A formula for target acquisition time based on
  target width and the distance to target.

5
Introduction

• How might tactile feedback devices effect target
  acquisition time?
• Variables of the feedback that we might consider
  include
• the magnitude
• the texture
• the frequency (pulses per second)
• the area ( length x width) of the screen that
  its attached to
• the friction (how the effect slows down
  pointer/cursor movement)
• the recess dimensions
• the gravity effect (how the pointer is pulled
  towards the target)
• the style (is it a ridge around the target? Does
  it completely cover the target?)

6
Introduction

• The Goal
• Come up with a modified version of Fitts law to
  describe and predict target acquisition time in
  bimodal visual-haptic interfaces. This version of
  the law would take into account at least some of
  the variables unique to tactile-feedback.

7
Justification

• Why Do This?!
• One Law to Rule Them All we have so many
  different types of tactile feedback devices the
  PHANToM, the Pantograph specialized mice,
  trackballs, joysticks, pens, and palm PCs
  virtual reality gloves, chairs, suits game
  console controllers and steering wheelsThe law
  may show us how we can improve target acquisition
  times when using any of these devices. It will
  also give us a basis for performance comparison
  between the different devices.

8
Just For Fun
The PHANToM (Sensable Technologies Developed at
MIT)
9
Just For Fun
The Pantograph (McGill University)
10
Just For Fun
CyberGrasp Glove (Virtual Technologies, Inc.)
11
Justification

• Why Do This?!
• Screen Space is A Precious Resource When we size
  down our targets, we increase target acquisition
  time. What if we could nullify this effect by
  introducing tactile feedback to the target? We
  could further minimize our use of screen real
  estate without a time penalty!An added benefit
  to minimizing the use of screen space is a
  reduction in visual information overload

12
Justification

• Why Do This?!
• Tactile Feedback is Cheap! Why Not Use It? While
  until recently youd have to fork over at least
  15k to get your hands on a PHANToM or build your
  own tactile feedback device, these days tactile
  feedback devices are cheap - 30 cheap. So price
  is no longer an impediment. A law such as this
  one could inform tactile feedback design in order
  to make it useful, instead of a gimmick that may
  even be less efficient, and hopefully would
  encourage the use of these now-affordable
  devices.

13
Justification

• Why Do This?!
• Benefit the Disabled Numerous studies have
  suggested that visual interfaces augmented with
  tactile feedback decrease target acquisition
  time. One would assume that this is because of a
  decreased reliance on vision and an increased
  reliance on the sense of touch. This decreased
  need for visual perception may be of help to the
  visually impaired in navigating these bimodal
  interfaces.

14
Fitts Law Movement Time

• MT a b x ID
• Where
• MT movement time (time to acquire target)
• ID Index of Difficulty (Shannons information
  channel capacity)
• a, b slope and intercept coefficients
  (determined by linear regression on data)

15
Fitts Law Index of Difficulty

• IDFitts log2 (2A / W)
• IDMacK log2 (A / W1)
• Where
• ID Index of Difficulty, measured in bits
  (Shannons information channel capacity)
• A movement amplitude, measured in pixels
  (distance to target, was power of signal in
  Shannon)
• W target width, measured in pixels (error,
  was noise in Shannon)

16
Fitts Law All Together Now

• MTFitts a b log2 (2A / W)
• MTMacK a b log2 (A / W1)
• Where
• ID Index of Difficulty, measured in bits
  (Shannons information channel capacity)
• A movement amplitude, measured in pixels
  (distance to target, was power of signal in
  Shannon)
• W target width, measured in pixels (error,
  was noise in Shannon)
• a, b slope and intercept coefficients
  (determined by linear regression on data)

17
Fitts Law Fitts vs. MacKenzie

• Why would we pick one equation over the other?
• According to Dosher, Fitts original equation is
  more appropriate where there is a large A (target
  distance) to W (target width) ratio, while
  MacKenzies equation is more appropriate for a
  small A-to-W ratio.

18
Fitts Law Throughput

• TP ID / MT
• Where
• TP Throughput, measured in bits/second
• ID Index of Difficulty, measured in bits
  (Shannons information channel capacity)
• MT movement time, measured in seconds (time to
  acquire target)

19
Fitts Law Experiment

• Choose a particular task to study.
• Create many instances of this task with varying
  Indexes of Difficulty by manipulating A (distance
  to target) and W (width of target) in each
  instance of the task.
• Conduct multiple trials for each instance of the
  task.
• Record MT (target acquisition time) and accuracy
  for each trial.

20
Fitts Law Experiment Analysis

• Use linear regression techniques to discover a
  b values for Fitts equation.
• We can use these equations to predict compare
  user performance Here device A wins.

4 -- 3 -- 2 -- 1 -- 0
Device A MT 2.3 0.5 ID
Movement Time (s)
Device B MT 0.5 0.25 ID
Index of Difficulty (bits)
1 2 3 4 5
21
Types of Tasks

• Target Acquisition
• Pursuit Tracking following the target around
  the screen, trying to catch it flyswatting
• Freehand Inking drawing or signing your name
• Constrained Linear Motion the steering task
  for example, selecting from drop down menus
• Constrained Circular Motion for example,
  rotating objects in a 3D interface.

22
Types of Tasks

• Advantages of studying Target Acquisition
• Its simple
• Its studied widely and well-documented
• Its very applicable to everyday desktop usage

23
How Does TF Improve Target Acquisition Time?
Movementbegins
Cursor Enters Target
CursorStops
Mouse buttonClick
Stopping Time
Clicking Time
Approach Time
Selection Time
Target Acquisition Time (MT)
TIME
(This diagram is from Akamatsu MacKenzie)
24
Effects of Feels on MT

• Putting the Feel in Look and Feel by Oakley,
  McGee, Brewster, Gray, CHI 2000.
• Studied four different haptic effects Texture,
  Friction, Recess, Gravity
• Task Simple target acquisition Fitts Task
• Device Used Immersion Corporations PHANToM

25
Effects of Feels on MT

• The experimenters looked to see how the four
  effects affected
• Slide-over errors the user misses the target
  slides right over it
• Slip-off errors the user acquires the target,
  but falls off before they can select it
• Target Acquisition Time

26
Effects of Feels on MT

• The results
• Slide-over errors
• Texture ? significantly more error than no TF.
• Friction ? about the same error as with no TF
• Recess Gravity ? sig. less error than no TF.
• Slip-off errors
• Texture ? significantly more error than no TF.
• Friction Recess ? about the same error as no
  TF.
• Gravity ? significantly less error than no TF
• Target Acquisition Time
• No Significant Differences!

27
Effects of Friction and Texture

• Movement Characteristics Using a Mouse with
  Tactile Feedback Akamatsu, MacKenzie 1996
• Studied Friction (what they call force
  feedback), and Texture (what they call tactile
  feedback), and Friction Texture
• Task Simple Target Acquisition
• Device Tactile/Force Feedback Mouse

28
Effects of Friction and Texture

• The Results
• All three feedback types produced statistically
  significant effects
• Texture performed the best,
• Texture Force performed the 2nd best,
• Force alone performed the worstIt was worse than
  no feedback.
• Texture had more error than no feedback.

29
Effects of Magnitude

• Preliminary Two-Dimensional Haptic thresholds
  and Task Performance Enchancements Lee
  Hannaford (2001?)
• Studied Magnitude of Haptic Feedback
• Task Target Acquisition
• Device Pen-Based Haptic Display

30
Effects of Magnitude

• Results
• Two tasks acquire indicated target from amongst
  5 different targets when only desired target has
  feedback and acquire indicated target when all 5
  targets have feedback.
• Haptic forces as low as 50 milliNewtons (weight
  of 2 US dimes) showed small performance
  improvements.
• As the magnitude of the force went up, the time
  to acquire the target decreased for both tasks.

31
Other Research Results

• Dennerlein, et. al. found that haptic feedback
  improved both the error and time in task
  completion of steering tasks.
• Balakrishman, et. al. developed a custom tactile
  feedback joystick device to work with an
  on-screen virtual hand-held tool. The device
  shows a 44 increase in accuracy but a 64
  increase in time to completion.
• Sallnas, et. al. had subjects pass cubic virtual
  objects to each other, one group receiving haptic
  feedback from the objects and one group not...
  the haptic feedback group did not have a
  significantly reduced time of completion but had
  a significantly lower error rate.
• Wall, et. al. used the PHANToM to have subjects
  complete Fitts' tapping tasks. The force
  feedback was found to improve subject's movement
  times.
• Dosher found that the amplitude of haptic
  stimulus and improvement in time were positively
  correlated.
• Arsenault found a 20 improvement in target
  acquisition time with tactile feedback using a
  PHANToM in a 3D virtual environment.

32
Psychology Of Touch Theories

• Touch Teaches Vision Berkeley
• What our eyes tell us is meaningless without
  tactile experience to calibrate it. (Not
  supported very well by research)
• Developmental Integration Piaget (1953)
• Early in life, we experience both visual and
  tactual sensations, but as we mature these
  separate experiences integrate into single
  objects.

33
Psychology Of Touch Theories

• Visual Dominance - Sometime between 1-2 years of
  age, people tend to have integrated senses of
  vision and touch with vision dominating.
  Conflicts in perception are resolved by vision
  dominating.
• However, touch is as accurate as vision at
  perceiving texture (Lederman Abbot)
• Auditory information better than vision at
  informing us of time-based events (Myers, Cotton,
  Hilp)

34
Psychology Of Touch Theories

• Modality Specialization Freides
  (1974)Different senses are better for different
  tasks as the task becomes more complicated,
  appropriate modalities kick in even more.

35
Psychology Of Touch Theories

• Warren Rossano suggest three kinds of tasks
  that introduce different relationships between
  touch and vision
• Texture vision and touch perceive about equally
• Shape more accurately perceived by vision when
  theres conflict, vision wins
• Spatial Location more accurately perceived by
  vision however, when theres conflict, vision
  does not always win

36
Celebrities

• I. Scott MacKenzie Dept. of Comp. Sci. at York
  University in Ontario Fitts Law Guru
  Extradordinare, also interested in tactile
  feedback.
• Bill Buxton Dept. of Comp. Sci. at University
  of Toronto Fitts Law and Touch-Sensitive
  Tablets

37
Celebrities

• Shumin Zhai IBM Almaden Research Center
  University of Toronto Researches Fitts Law
  Steering Law tasks
• Johnny Accot (no picture)
• Julie Jacko Industrial Systems Engineering,
  Georgia Tech Multimodal Feedback, Designing for
  the Partially Sighted

38
Celebrities

• Blake Hannaford Dept. of Electrical
  Engineering, University of Washington
  Researches haptic thresholds/magnitude
• Jesse Dosher
• Gregory S. Lee

39
Celebrities

• William Schiff Psychology Dept., NYU The
  Psychology of Touch and Haptics
• Morton Heller Psychology Dept., Eastern
  Illinois University The Psychology of Touch and
  Haptics

40
Issues to Consider

• In a desktop application, how will the different
  haptically charged UI elements affect each
  other? Is it better to give everything a tactile
  feedback, selectively give UI elements on the
  tactile feedback, or only give the target tactile
  feedback?
• If were only giving the target tactile
  feedbackhow do we know what the next target is?
  (Stochastic classificationcollect user data over
  time to predictMunch Dillman) Is this
  realistic?

41
Issues to Consider

• How does error fit into Fitts Law? Some
  researchers reduce error to a nominal variable
  (either present or not present) and they count
  the number of times that it occurs, or they
  determine how much error by recording how many
  pixels off the selection was.
• Can a separate law be created to predict error?

42
Issues to Consider

• How do all of these haptic variables effect
  performance in tasks other than
  target-acquisition?
• What expectations do we have of different tactile
  sensations (hard, soft, smooth, textured, etc.)
  and how will those expectations affect our
  perception of different UI elements? (for
  example, would someone find themselves less
  motivated to use a certain option in a program
  because it had a jagged, possibly disconcerting
  feeling?)