Cognitive Work Analysis in an Information Systems Design Context

1
Cognitive Work Analysis in an Information Systems
Design Context

• Ann M. Bisantz
• University at Buffalo
• Department of Industrial Engineering
• Center for Multi-source Information Fusion

2
Overview

• One perspective
• CWA (esp WDA) as framework for interface design
  particularly, system control interface
• Thoughts on expanding the perspective
• Enlarging the system boundary to include
  components such as sensors and automation
• Fitting WDA within a information fusion system
  design

3
Collaborators

• Natalia Mazaeva
• Emilie Roth Catherine Burns
• Galina Rogova and Eric Little

4
WDA and System Boundaries

• Where does the interface end, and the work domain
  begin?
• How do sensors, automation fit in? Can they?
• What about parts of the system that you cant
  control?
• Traditionally, sensors have not been represented
• Sensors have been viewed as the means to view or
  understand the state of system components, rather
  than an interactive system component
• Similarly, automation has been seen as a part of
  the how question

5
Why is this important?

• Want to make sure to understand what needs to be
  sensed or controlled
• Not just what IS sensed
• Not HOW control is happening (task)
• BUT
• Sensors need to be controlled, use can affect
  higher level goals
• Automation is itself a system which must be
  controlled

6
Example Interactions from a Command and Control
Environment
Defense of Systems and Personnel
Achieve Assigned Missions
Signature Maintenance
Battlespace Awareness
Use of sensors necessary for offensive functions
Use of some sensors makes ship detectable
Environmental Sensors
Battlespace Sensors
Maneuvering Systems
Moving may disambiguate sensor data
Knowledge of environmental conditions can impact
choice of sensor settings
7
Modeling With Sensors

• Focus is on the interaction between sensing
  systems, other systems, and higher level goals
• Input to displays which highlight these
  interactions and potential conflicts
• Highlights need for new sensor designs which
  minimize these conflicts

8
Recommendation A Dual Approach
Physical Environment

• Utilize models of system, environment, and other
  agents as required by domain
• Include sensing systems in entity model, with
  links to other models highlighting what sensors
  can, and cannot reveal

Agent
System
9
Similar Approach for Automation?
Physical Environment

• Model an automation layer, in addition to
  system/environment/entity
• Links indicate relationships among automation and
  system nodes to be controlled, other
  affected nodes

Agent
Entity
Automation System
10
AH of Automated Camera

• Camera Model
• Automation Model
• Purposes, components of automation different than
  that of camera
• Connections across models to indicate which parts
  of the automation controlled functions, parts of
  the camera

11
Film Exposure Film Movement Light Focus
Exposure Control
Camera components which are actually controlled
Specific Components
12
Contrasted to Decision Ladder Models

• More typical approach to the representation of
  automation
• Three part user-automation-designer decision
  ladder for one control task setting the aperture
  
• Highlighted task sequence for user in setting the
  aperture, responding to an exposure warning light
• Input to procedures/training

13
User
Task Setting the Aperture
Designer
Automation
14
Conclusions

• Combined camera-automation model highlighted
  connections that could be important in training
  or fault diagnosis
• What automated function are available to control
  camera functions
• Why an automated function is not being
  accomplished
• Differing degrees of automated control over
  different functions

15
Conclusions

• AH-DL relationships
• Interleaved human/automation activities act on,
  and change the state of components in the AH
• AH, DL provide different types of design guidance
  related to automated systems

16
WDA of an Emergency Management System

• Took the system-agent-environment approach
• Explicit interactions
• Included sensors or situation assessment
  functions

17
Work Analysis of SEMS-Structured EOC
18
Ambulance
Dispatch
Ambulance
Hospital
Center/
System
Medical
Dispatchers
Personnel
Civilian
Transportation
Civil
Land
Responders
Infrastructure
Infrastructure
Atmosphere
Communications
Population
Infrastructure
Situation assessment function relies on EMS
personnel as well as sensor systems and
algorithms Note assessment system environment
agent systems
19

20
Plan-Human Interaction

• Primary Display Implications
• Available resources from other jurisdictions and
  their availability based on mutual aid and other
  planning agreements (e.g., Display a fire
  company's location, along with its availability
  potential availability to a selected
  jurisdiction) Jurisdictional boundaries
  dynamic/"virtual" jurisdictions which change as
  mutual aid plans are activated
• Secondary Display Implications
• Procedures/plans detailing mutual aid and
  state/federal assistance why a resource is/isn't
  available procedures for requesting mutual aid

21
(No Transcript)
22
Building-Casualty Interaction

• Primary Display
• Predicted numbers/types of casualties based on
  building types time of day, amount of damage,
  updated with real time reports (LINK WITH SENSOR
  SYSTEM)
• Secondary Display
• Specifics of report/input to fusion algorithms

23
Information Fusion Project Goals

• Working as part of a multi-year, multi-site AFOSR
  sponsored program to develop methodologies for
  design of higher level, information fusion
  technologies
• Post-disaster management (specifically,
  earthquake response) chosen as the test bed
  environment

24
Approach

• Application of Cognitive Systems Engineering/Work
  Domain Models
• Extract information requirements which can be
  fulfilled using information fusion technologies
• Explore relationships between WD, Ontological
  models
• For information fusion design
• To support information fusion computations

25
Information Fusion
Information fusion is a way to deal with all
kinds of information imperfection by exploiting
multiple source data to obtain information of
better quality and provide better understanding
of an observed phenomenon
Databases of historical information
Multiple sensor information
Domain knowledge
Process refinement
S1

• Sensor information processing
• spatial registration
• temporal registration
• sensor fusion

• Reasoning about entities IDs and relations
  between entities for
• situation interpretation
• prediction of consequences

Feature extraction and fusion for single entity
processing and identification
Lower Level Fusion
Higher Level Fusion
26
Ontological Modeling

• Ontological models can be used to categorize
  anything that exists, including
• objects, their properties and attributes,
  temporal spans, and the hosts of relations which
  exist between any of these items.

27
The Structure of an Ontology

• Upper-Level (Formal)
• Most general categories of existence (e.g.,
  existent item, spatial region, dependent part).
• This Level of the ontology is rationally driven,
  meaning it is the product of philosophical
  reasoning.
• Relies on a sound metaphysical description of the
  world (e.g., realism).

28
The Structure of an Ontology

• Domain-Specific Level
• Contains categories that are specific to a
  particular domain of interest (disaster,
  military/defense, medicine).
• This level of the ontology is empirically driven,
  meaning it is produced by gathering expert
  knowledge about a given domain of interest.
• The expert knowledge is used to create a
  consistent and comprehensive lexicon of terms.

29
Synthesized Ontology Model
30
FORMALLY STRUCTURED DOMAIN-SPECIFIC CATEGORIES
Existent Item
Dependent Items States, Qualities, Powers
31
Work Domain Modeling Methodology

• Visits to CA Emergency Operations
• FEMA, State, County, and LA City operations
  centers 911 dispatch (seeing post-wildfire
  activities)
• Primarily EMS management/coordination personnel
  as well as dispatch/911 operators
• Observed Northridge Exercise in LA City EOC
• Overall WDA framework development from
  literature/planning documents meetings
  interviews
• Annotated with sources and quotes
• Information Display requirements generated
  within this model

32
Example Information Requirements

• Number/injury level of casualties locations
• A priori Prediction based on building materials,
  geographic features, time of day,
  type/characteristics of earthquake event
• Updated as new information comes in
• Areas of risk of secondary hazard (e.g.,fire)
• A priori prediction based on previous weather
  (rainfall), types of vegetation building
  materials update with info re. Gas line
  ruptures, water main breaks (areas with low
  pressure)
• Resource balance assessments
• Is there enough of any particular resource to
  meet projected demand?

33
Ontological Model

• Situational Hypotheses with Confidence levels
• current situation
• a priori predicted risk
• consequences (risk)

Goals
Situational objects
Combination of elementary situations (COS)
Abstract objects
Physical objects (PO)
Static
Dynamic
Elementary situations (ES) Situation of interest
for a decision maker/group of DM

• Bridge
• Attributes
• Location
• Probability of damage

• Road
• Attributes
• Location
• Probability of damage
• Capacity
• Traffic flow

• Casualty
• Attributes
• Location
• Level of injury
• Type of injury

Casualty situation Attributes (discrete and
cont.) .
Aggregates
Hospital situation Attributes ..

• Ambulance
• Attributes
• Location
• Level of injury
• Type of injury

Temporal
Spatial
Tr. Network Situation

• Cluster of casualties
• Attributes
• location
• Exp. level of injury
• Area

Cluster of ambulances
Resource situations

Secondary threat situations fire, chem. spills
Cluster of ambulances
Ambulance situation


34
Links with Ontology

• Node descriptions/definitions should match (be
  informed by) the ontology
• WDA could be constructed from the formal
  (upper-level) ontology since it represents
  logical connections between the most basic
  concepts of interest
• In tandem, the domain ontology could learn
  domain-specific concepts of interest from the WDA
  (since it is based on expert knowledge)

35
Mapping WD model to the Ontology

• Direct comparison showed similar concepts at the
  physical form, physical function level of the WDA
• Different organization an entities existence is
  represented separately from the multiple states
  that it can have (e.g., location, operability)
• Current disaster-response ontology (Dis-ReO) does
  not represent higher level functions, abstract
  functions, or system purposes, but could be
  augmented

36
CSE, Ontology, Fusion (Design)
37
CSE, Ontology, Fusion (Processing)
Medical Operations processes ---Fixed Location
Resource-- Hospitals
Work Domain Model
Do I need to request additional medical resources?
Ontology
Entities of interest
Work driven information requirements
Relevant states, attributes, and relations with
other entities
States of Damage, Available capacities
Fusion Processing
Fused estimates
Databases (a-priori dynamic)
Observations
Environment