38
Modelling the effect of stress on Human Behaviour May 12 1999 LTSS51 Orlando Andy Belyavin CHS DERA

Modelling the effect of stress on Human Behaviour May 12 1999 LTSS51 Orlando Andy Belyavin CHS DERA

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

Modelling the effect of stress on Human Behaviour

May 12 1999

LTSS51 Orlando

Andy BelyavinCHS DERA

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

Role of constructive simulation

Entities involved in man-in-the-loop virtual simulation for training In future analysis of military systems it can be anticipated that

there will be more use of man-in-the-loop virtual simulation This will be effective for managing the burden of the analysis of

tactics and outline questions on crewing and systems definition It will not support the analysis of system performance in all

contexts There will be a large role for the constructive simulation of

human behaviour under stress

Constructive modelling of human performance

Based on a structure of what the crew has to do

Task analysis leading to task networks

– IMPRINT

– MicroSAINT

Task frames in SAFs

– ModSAF

Rule bases in command agents coupled to SAFs

Classical approach to stress representation

Define task taxonomy

– Cognitive task

– Perceptual task

– Physical task etc.

Map environmental stress to task types

– ‘Arousal’ affects cognitive performance etc.

Model effect as a crude degradation

– Adjust task time and precision

Long term strategy for stress description

Three things have to be achieved: Define the phenomenon we are trying to represent

– Define the stressors we need to consider

– Define nature of best scientific knowledge

Review current approaches

– How is it done in current tools?

– SAFs, IMPRINT, IPME

Project how these methods should develop

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

Environmental stressors

A 1994 review at DERA identified more than 40 stressors

These include both regular environmental factors and social effects

Suggest that even a concise list of the most important is 10 long

Environmental stressors (2)

Sleep loss fatigue / circadian effects and time on task Physical fatigue Thermal effects (Thermal strain / dehydration / discomfort) Visual environment Fear / Anxiety / Morale Task demand - workload Noise (continuous and impulse) Vibration Hypoxia (Loss of oxygen in high flying fast jets) High G (Fast jets only)

Metrics of “behaviour”

What is the crew / operator going to do? Generally domain of cognitive analysis – possibly open

ended

– How good is Situation Awareness?

– What course of action is selected?

Given what the crew /operator does, how well do they do it? Generally domain of task analysis and task performance

– How fast is the task completed?

– Is the task performed accurately?

Relationship between Environment and Performance

Environment State Change

Operator/ Crew State Change

Operator / Crew Performance Change

Effect of sleep loss / Time of dayon performance

Sleep loss and time of day affect operator state

State variable is “Mental Alertness”

Mental Alertness affects performance

Different effects for different tasks

Current analysis covers “Vigilance” and “Cognitive” tasks

Alertness Model

-15

-10

-5

0

5

10

15

7 9 11 13 15 17 19 21 23 25

0

10

20

30

40

50

60

70

7 9 11 13 15 17 19 21 23 25

Circadian Effects (time of day)

‘S’ Effects (time since sleep)

t timeof day

current time

y t

tod

tod tod

13 4 2 24. cos( ( ) / )

t time ce sleep

y t t

tss

tss tss tss

sin

. exp . / . 865 0 317 0 0612

Resultant Alertness

0

10

20

30

40

50

60

70

80

7 9 11 13 15 17 19 21 23 25

A y ytss tod 13 4.

Alertness effectVigilance Misses

0.4

0.5

0.6

0.7

0.8

0.9

1

0 10 20 30 40 50 60 70 80 90 100

Alertness

Su

stai

ned

Att

enti

on

Pro

po

rtio

n C

orr

ect

Proportion correct adjusted for constant

Predicted proportion correct

TG5 WP 1997

CHS Whole body thermal model

Solves diffusion equation for linked cylinders Represents blood flow inside the body in moderate detail Handles radiation / evaporation / conduction at surface Handles active controllers:

– Sweating

– Shivering

– Blood flow modification

Handles sweat evaporation through dry clothing Coupled to IPME through socket interface

Thermal strain and performance

Preliminary indications

Dehydration affects error rate on cognitive tasks Dehydration affects physical performance

High temperature speeds performance Discomfort slows performance Dehydration slows performance

Nature of states

Candidate examples:

Anxiety

– Possibly influences whether the Operator / Crew may or may not participate

– Possibly influences nature of Situational Awareness etc.

Motivation

– Possibly influences participation / course of action

Alertness / Arousal

– Influences performance and errors

– Influences decision to act

– In extreme case leads to falling asleep

Task demand

Military operations frequently involve high task demand reflected by the need to do more than one thing at once

Classically represented by a “state” – workload Workload then determines allocation of priorities and

performance of the task and / or choice of action

Two models do not involve state DERA Prediction of Operator Performance (POP) model Canadian Information Processing / Perceptual Control

Theory (IP / PCT) model Both based on interference effects

Relation between Environment and Performance / behaviour

Original proposed simple model: Environment to State to Performance / Behaviour Incomplete

More complex model needed Add interference between tasks and its effects Multiple states have to be considered Initial evidence is that interaction effects can be ignored

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

System information

Background environment information

– Scenario details (Threats)

– Conditions (Temperature, Duty pattern etc.)

Team characteristics

– Fatigue state

– Training etc.

Performance modifiers

– Fatigue degaradations etc.

– Determined by task taxonomy etc.

Task data required

Time distribution

Probability of failure

Consequences of failure

Who is doing the task

Nature of the task according to the taxonomy

Associated task demand (optional)

OperatorTrait

EnvironmentState

OperatorState

TaskExecution

OperatorPerformance

Feedback (Workload)

Performance shaping model

Areas covered and under studyunder IPME project

Effects of circadian / sleep loss cycle (CHS alertness model)

Effects of heat / dehydration / discomfort on task performance (Cognitive and physical)

Effects of visual environment on performance Effects of terrain on movement speed Effect of task demand (workload) on task performance

(POP model) Alternative model of stressor degradation (interference

hypothesis - applied to anxiety)

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

Possible structure

Environment / state Model 1

Environment / state Model 2

Crew / OperatorStates

Task demandInterference model

Cognition / PerceptionModel

Performance / ActionModel

Five classes of model identified Model of cognition / perception

– Situational Awareness

– Perception of environmental information (Sensory models)

Model of course of action / performance

– Decision making (NDM / Rule base / task network)

Model of task interference effects (“Workload”) Model of influence of state on first two models

– Performance degradation

– Choice of action modification

Model of influence of environment on state

Environment to state

Models of relationship between environment and state can be complex

Full CHS Alertness model taking account of shift work / time zone shift involves solution of differential equations

Wide range of thermal models with varying degrees of complexity

Interpolation formulae to full systems of differential equations

Different applications demand different levels of detail and complexity

Argues for a modular solution to this component

Task demand

Range of solutions of varying degrees of complexity Simple compounding models based on task characteristics

(VACP) More complex models handling interference effects (DERA

POP) Yet more complex models handle prioritisation and

modifications to courses of action (IP / PCT) Again the level of complexity dictated by the application

arguing for a modular approach

Effects of state

Less well developed topic Some simple interpolation formulae available for task

performance Some more complex models of impact of state on

perception Few well developed models of effects of state on course of

action Last point important to overall effectiveness Non-participation / suppression a very important effect

Crew as system of systems

Many highly developed models of aspects of human behaviour

Varying levels of complexity and applicability Re-use and long term development argues strongly for a

modular design with a standard interface between the models

HLA architecture can be applied below the level of the system to the crew

Aims of the presentation

Outline the scope of the problem from a modelling perspective

Sketch a structure in which the problem might be solved

Outline implementation in IPME

Generalise the approach to broader class of architectures

Man as system of systems - possible solutions

Strategic solution

Modular replaceable blocks: Perceptual engine – take account of state Cognitive engine – take account of state Possibly use NDM pattern recogniser and ignore state State predictors from environment can be simple or

complex Task demand managers can be simple or complex Re-use of existing models implied

Major issues for future

Definition of modular architecture

Defining set of states which we need to recognise

Defining how state interacts with cognition and perception

Defining relationship between environment and state

Defining relationship between traits and state