Download Human Computer Interaction Specialty Report, November 1, 2002

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Theoretical computer science wikipedia , lookup

Animal communication wikipedia , lookup

Transcript
Attention Grounding: A New Approach to
IVIS Implementation
Emily Wiese
Cognitive Systems Lab
Department of Mechanical and Industrial Engineering
University of Iowa
In-Vehicle Information System (IVIS)

Computing Anytime, Anywhere
• Functions include: Cellular telephones, navigation
systems, collision avoidance systems



Est. $13 billion business by 2006
Improve productivity, satisfaction, and safety
Distraction Potential
• Traffic accidents cause 42,000 deaths and $150 billion in
costs each year
• 13% and 50% percent of crashes attributed to driver
distraction; 10,000 lives lost; $40 billion in damages each
year
Human Computer Interaction Specialty Report, November 1, 2002
Implementing IVIS Functions
Interference
Mitigation
Workload
Management
Attention
Grounding
Types of
communication
Links between
functions
Distraction
countermeasure
Direct
communication
No integration or
physical
integration only
Static interference
minimization
Direct
communication
Collaborative
grounding
Functional
integration with
direct
communication
IVIS-centered
workload
management
Functional
integration with
back-channel
communication
Driver-centered
attention
distribution
Human Computer Interaction Specialty Report, November 1, 2002
Interference Mitigation

Direct Communication
• Commands associated with specific tasks

Physical Integration Only
• IVIS functions may share the same interface or
location, but do not share information

Static Interference Minimization
• Distraction assessment focuses on structural
distraction, resulting in locked-out functions
• Does not consider how driving demands change over
time
Human Computer Interaction Specialty Report, November 1, 2002
Workload Management

Direct Communication
• Commands associated with specific tasks

Functional Integration (FI) With Direct Communication
• FI considers the information required and produced by each function to
support communication between the driver, the IVIS, and the roadway

IVIS-Centered Workload Management
• Dynamically predicts how workload will respond to changes in the driving task
and use of the IVIS
• IVIS actively adjusts functionality by estimating the demands of the current
roadway and IVIS states.
• Considers mental overload as the only source of distraction and overlooks the
issues of cognitive tunneling and attentional withdrawal
Attentional resource capacity
Attention to driving
Attention to IVIS
Time
Human Computer Interaction Specialty Report, November 1, 2002
Challenges facing IVIS Implementation

How can IVIS functionality be made safer and easier to use,
without succumbing to the usability paradox and resulting in
decreased overall roadway safety?
What factors affect a driver’s willingness to engage in IVIS
interactions?
2. What are the consequences of error recovery in speech
recognition systems on driver performance?
3. How do we guide drivers to avoid inappropriate reliance on speed
control and collision warnings functions?
4. How do we implement collision warning functions such that the
rate of false alarms will not undermine driver acceptance?
1.

The interference mitigation and workload manager approaches
have not addressed these issues
Human Computer Interaction Specialty Report, November 1, 2002
Attention Grounding

Collaborative Grounding

Functional Integration with Back-Channel
Communication

Driver-Centered Attention Distribution
Human Computer Interaction Specialty Report, November 1, 2002
Collaborative Grounding


Incorporates back-channel cues to allow us to
establish grounding without disrupting the flow of
conversation: creates a Shared Context
Collaborative Grounding supports:
•
•
•
•
•
•
Initiating communication
Delays in communication
Driver attention distribution
Information coordination
Understanding uncertainty in IVIS communication
Making Repairs
Human Computer Interaction Specialty Report, November 1, 2002
Functional Integration with Back-Channel
Communication

Allows for more complete driver monitoring

The shared context created by back-channel
communication and collaborative grounding can
extend the benefits of functional integration
Human Computer Interaction Specialty Report, November 1, 2002
Driver-Centered Attention Distribution


Uses shared context to consider the evolving driving
situation in distributing attention (vs. exceeding a pool of
resources)
Back-channel cues can be used to help distribute the
driver’s attention appropriately
Attention
Activation
Potential Field
Time
Driving
Task
IVIS
Task
Human Computer Interaction Specialty Report, November 1, 2002
Direct information from the roadway:
Collision situation ahead
Roadway
Back-channel
information from
the roadway:
The “feel” of the
surface
Back-channel
information from
the system:
Change in voice
intonation
Shared
Context
Back-channel
information from
the driver:
Pauses in
conversation
Driver
IVIS
Direct commands from the driver:
“Get directions to San Antonio”
Direct commands and information from IVIS:
nd
“Turn Right at 2 Street”
Collaborative communication: Direct and
back-channel communication in driving
Human Computer Interaction Specialty Report, November 1, 2002
Challenges facing IVIS Implementation
What factors affect a driver’s willingness to engage in IVIS interactions?
•
AG supports the driver as an active participant in choosing when and how IVIS
functions are used.
What are the consequences of error recovery in speech recognition systems on
driver performance?
•
Grounding in the state of IVIS can support error recovery and promote more fluent
speech interaction.
How do we guide drivers to avoid inappropriate reliance on speed control and
collision warnings functions?
•
Grounding in the state of automation may help avoid inappropriate reliance.
How do we implement collision warning functions such that the rate of false alarms
will not undermine driver acceptance?
•
Back-channel cues can provide drivers with continuous information that may be
less annoying than discrete alerts.
Human Computer Interaction Specialty Report, November 1, 2002