Affective Computing

1
Affective Computing

• Affective Computing A Rationale for Measuring
  Mood with Mouse and Keyboard
• 4/12/04
• Based on a fine paper byPhilippe Zimmermann

2
Outline

• Introduction
• Affect, Emotion, Mood
• Structure and Labeling of Affect
• Mood Measurement Methods
• Mood Induction Methods
• Discussion
• Method
• Analysis
• Conclusions

3
Introduction

• Measure User Affect
• Use non-invasive, unobtrusive measurement method
• Use standard computer devices (Mouse, Keyboard)
• Extract data from log files of mouse and keyboard
  movements to analyze correlations with affective
  state

4
Affect, Emotion, Mood

• Affect - refers to either an emotion or a mood
• Emotion A reaction to a specific cause or
  stimulus. An intense experience of short
  duration. Can cause moods
• Mood Subtle, longer lasting, more in the
  background affective state of the user
• An affective computer can take advantage of mood
  by presenting stimuli that sustain a certain
  mood, or counteract undesired moods.
• Frustrated users could be prompted to take a
  break
• E-Commerce website could take advantage of the
  study that said a low-risk purchase is more
  likely while in a good mood, and a high-risk
  purchase is more likely in a neutral or negative
  mood

5
Structure and Labeling of Affect

• Three basic emotional dimensions
• Affective Valence (pleasure)
• Positive (Pleasant) to Negative (unpleasant)
• Arousal
• Calm (low-arousal) to Excited (high-arousal)
• Dominance (control)
• Controlled to In Control

6
Mood Measurement Methods

• Physiological
• Heart Rate, Blood Pressure, Respiration, Muscle
  Action
• Psychological
• Self Reports - Verbal, Ckecklists, Questionaires
• Behavioral
• Facial expressions, Gestures, Movement
• Studies show relation of mouse click patterns to
  user frustration

7
Mood Induction Methods

• Mood changed by reading statements
• This method results in a short term mood change
• Mood changed by watching video with audio
• This results in a longer mood change

8
Discussion

• Physiological Signals will distract the user too
  much (You need wires connected to you to gather
  data such as heart rate and blood pressure)
• Self Reports of current mood can be inaccurate as
  the questions asked of the user to determine mood
  might invoke a mood.
• Using only a keyboard and a mouse will be
  non-invasive, and should not have the ability to
  change the users mood.

9
Method

• Film clips were used to induce moods.
• After watching the clip, the users were prompted
  with a brief survey on their computer to rate
  their current mood.
• Users were required to shop on an e-commerce
  website.
• Physiological (pulse, respiration, etc) data was
  taken.
• Mouse and keyboard movements were recorded in the
  background to a log file. Button Up/Down, mouse
  coordinates, and what key pressed were logged

10
Method (2)

• Users Filled out the mood survey, watched the
  clip, shopped the e-commerce site, and then
  filled out the survey again.
• The experiment lasted 1.5 to 2 hours

11
Analysis

• Using the physiological data and the surveys, the
  user's mood will be verified to be one of the 5
  different affective states
• The log files with the Keyboard and Mouse log
  files will be examined with the known mood
  changes to look for patterns.
• Number of mouse clicks per minute
• Duration of mouse clicks
• Distance of a single mouse movement
• Number and length of pauses in movement
• Minimum/Maximum/Average mouse speed
• Keystroke rate per second
• Average Duration of Keystroke

12
Conclusion

• If a system knows how a user feels it can
  appropriately react to those moods
• Questions
• Should a system try to change your mood?
• What emotions should Human Computer Interaction
  be concerned?
• How Accurate does the measurement have to be?