ROBUST APPROACH TOWARDS CONTEXT DEPENDANT

INFORMATION SHARING IN DISTRIBUTED ENVIRONMENTS

Jenny Lundberg and Rune Gustavsson

Department of Interaction and Systems Design, Blekinge Institute of Technology, 37225 Ronneby, Sweden

Keywords: Information sharing in teams, Semantics, Context, Abbreviation, Robustness.

Abstract: In the paper we propose a robust approach towards context dependant information modelling supporting

trustworthy information exchange. Shortcomings and challenges of present approaches of syntax-based

information modelling in dynamic context are identified. Basic principles are introduced and used to

provide a robust approach towards meeting some of those challenges. The approach has a main aim of

reducing brittleness of context dependant information and enabling intelligible information handling in

distributed environments. The application domain is Emergency Service Centres, where the distributed

handling of emergency calls in life critical situations of future change is in focus. The main contribution in

the paper is a principled approach of use of abbreviations in dynamic emergency situations. Points of

interaction for coordination are introduced as a tool supporting mappings of abbreviations between different

contexts.

1 INTRODUCTION

Design, implementation and maintenance of digital

support for handling of tasks in life-critical

situations are a challenge. We have addressed some

of those challenges even for distributed

organizations such as Emergency Service Centres

(Lundberg, 2007). The operators handling

emergency calls rely on proper ICT systems

supporting their everyday work. In this life-critical

context abbreviations are commonly used mainly as

a way of saving time, but also as a quality assurance

method by introducing structured action-types

related to calls in a semantically unambiguous way.

Abbreviation-based information exchange is

common in many life-critical situations such as

dealing with emergency calls (the main application

of this paper), Air Traffic Control (ATC), operator

control of critical infrastructures, and in several

medical applications. There are some clear benefits

of abbreviation-based information sharing in teams

but also some inherit, and potentially catastrophic,

limitations of this approach. In the following Section

2 Setting the scene, we illustrate those aspects and

identify some challenges towards ensuring

semantically correct context dependent information

exchange in teams. The reminder of this paper is

organized as follows. Section 3 present our robust

abbreviation approach based on changes of context

with a specific focus upon information modelling,

common ground, coordination, situations and work-

flows. Other approaches are shortly described in

Section 4. In Section 5 we revisit our challenges of

Section 2 and summarise our approach with some

pointers of future research. The paper ends with

references, presented in Section 6.

2 SETTING THE SCENE

To illustrate the challenges addressed in the paper,

we introduce a well-known example of using

abbreviation-based reasoning. That is reasoning

based on results done by handheld calculators as

depicted in the following Figure 1.

The result obtained by the calculator is the

depicted numeric value 1.41421356237095. This

value is calculated using the displayed equation

including ordinary numeric calculations and the

value of cos(pi) and sqrt(2).

The semantic information (value) can, however,

only be assessed by the user of the Calcul Engine in

the given context. Different contexts typically entail

different semantic information from the same

calculated value. In fact, this example illustrates the

power of algorithmic numerical calculation where

200

Lundberg J. and Gustavsson R. (2009).

ROBUST APPROACH TOWARDS CONTEXT DEPENDANT INFORMATION SHARING IN DISTRIBUTED ENVIRONMENTS.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages

200-205

DOI: 10.5220/0002004102000205

 SciTePress

Figure 1: Reasoning based on support by calculators.

the syntax determines the semantics of the

abbreviations used, i.e., numbers, numerical

operators, and numeric functions. The power of

algorithmic calculations is that they are context free

and the interpretation of the results, by the users, is

separated from the calculations of numbers. In fact,

the power of mathematical models (reuse in

modelling different applications) is due to their

independence of semantic meaning!

To further illustrate the strengths and

weaknesses of using abbreviation based reasoning,

we now take a closer look at workflow support in

Emergency Service Centers (ESC) and Air Traffic

Control Centers (ATC).

The workflow within individual ESC can be

described as a sequence of states.

• State 1: Classification of incoming calls

• State 2: Identification of appropriate

actions.

• State 3: Deployment of teams

• State 4: Debriefing and reporting

The tasks of each state are governed by a set of

abbreviations. The operators within the individual

centers understand how to faithfully code and

decode these abbreviations due to continuous

training and evaluation. Abbreviations support

efficient quality assured workflow support with

possibilities of parallelisms between states. That is

quality and efficiency.

The power of this type of abbreviations is also a

weakness. Inherent in the power of abbreviations is a

stable set of action-types corresponding to the

abbreviations and a closed group of users to enable

efficient training and other means of quality

assurance of the shared intended meaning of the

abbreviations. When connecting geographically

separated centers, and/or changing tasks to enable

more dynamic call centre handling, the mecha-nisms

of abbreviations becomes an obstacle against

changes and hence an issue to consider in depth

(Lundberg, 2007). In short, how could we, in a

principled way, handle context changes in

abbreviation based information exchange?

The following example illustrates the critical

dependency on a common understanding of context-

dependencies in abbreviation-based information

exchange. The example of misunderstand-ding in

coordination is due to the fact that the two main

actors involved didn’t succeed in handling two

implicit different contexts with overlapping

abbreviations.

In the example, an air traffic controller and a

pilot misunderstood the meaning of the abbreviation

‘holding’. In December 1995, the disaster of

American Airlines Flight 965 from Miami to

Columbia, resulted in a loss of 159 human lives as a

direct consequence to this misunderstanding. Part of

the conversation was as follows, where the air traffic

controller asked the pilot;

- Are you holding?

The pilot confirms with:

- Roger, we’re holding.

They misunderstood each other due to the

unsuccessful establishment of an agreed-upon

meaning of the word “holding”. Did it mean holding

latitude or holding rate of decent? The air traffic

controller and the pilot had different contexts and

interpretations of what ‘holding’ meant. As the air

traffic controller understood the holding as to

holding latitude, the pilot understood it as holding

rate of decent. A closer focus upon the situation, and

the fact that they had two different contexts could

most probably have avoided the loss of lives.

Those two examples illustrate the following two

inherent weaknesses with abbreviation based

information sharing:

• Problems associated with changes of

contexts

• Problems in understanding and

discovery of implicit sharing of

abbreviations in different contexts

The following Figure 2 captures the semantic

hurdles of mapping of intentions between a sender

and a receiver.

We have highlighted in the figure the challenges

related to introduction of an information-processing

artifact between the sender and receiver. The face-

to-face communication in natural language has thus

been disrupted with a processing unit with two

interfaces and a context model (CM) used by the

processing unit to perform syntax based (rule-based)

ROBUST APPROACH TOWARDS CONTEXT DEPENDANT INFORMATION SHARING IN DISTRIBUTED

ENVIRONMENTS

201

Context

Sender

Receiver

Information processing

artifact

Context

model

Interface

Intention!

Under-

standing?

Figure 2: Semantic hurdles in semantically correct

mappings between sender and receiver.

translation between the input format at the sender

interface and the output format at the receiver

interface. In Figure 1 the Context Model is the

numeric calculation algorithms of Calcul Engine.

In short, how can we model and trustworthy

convey meaning of artifact-mediated information in

distributed environments when change is the rule

and not the exception?

The essence of this challenge is addressing the

semantically correct mapping of the intention of the

information (by the sender) to the proper actions of

the receiver. This challenge has been in focus of

researchers in the fields of natural languages,

cognitive science, Artificial Intelligence, Cognitive

systems engineering, Knowledge intensive systems

engineering and HMI engineering since decades.

In the case of abbreviations the sender and

receiver agree about the semantic meaning of the

abbreviation and the tasks it should and could/could

not handle. Furthermore, the contextual meaning of

the information at hand has been reduced to the

syntax of the abbreviation. The remaining part of the

context is shared interpretation of the abbreviation

between sender and receiver as illustrated in using

hand held calculators (Figure 1).

As we have earlier noted; in certain contexts,

such as in numerical calculations, the Context model

(CM) of Figure 2 could indeed be purely syntax

based, i.e., the syntax of numerical calculations

fulfills the needs we have on the context model

whenever we need to do numerical calculations! In

short, numerical calculations in any context obey the

same syntax based rules. In fact, numerical

calculations are examples of algorithmic

applications where the reasoning power of the

computational artifact is syntax oriented and the

interpretation of the relevance of the computations in

a given context lies in the agreed upon modeling

principles and interpretation by the sender and

receiver at the two interfaces of Figure 2.

However, turning to knowledge-intensive

applications, such as support systems in real-world

decision-making, the situation is fundamentally

different. Let envelope, Env(CM), denote the set of

computations enabled by the Information pro-

cessing artifact in Figure 2, given the Context model

CM. Let Comp(C), denote the set of desirable

computations by the actors of Figure 2. Clearly we

always have that Env(CM) is a subset of Comp(C).

In the examples of abbreviations and algorithmic

application we can cope with this difference by

adding agreed-upon context dependant semantic

interpretation to the computational results. In general

this is not the case for knowledge-intensive

applications due to the (unintended) change of

context. The difference between Env(CM) and

Comp is sometime denoted the brittleness of the CM

model.

Despite this inherent shortcoming of machine

readable context models there have been

considerable efforts devoted to formal models and

ontologies of different domains. An ontology is a

syntactic specification of concepts and their relations

in a given domain. An ontology defines a formal

semantic of a domain. Large amounts of resources

have been spent on efforts creation of, for instance,

enterprise ontologies.

The purpose of having a shared or at least

intelligible enterprise wide semantic is to enable an

understanding of the business both within the

company as well as with customers and business-to-

business. However, this approach has not produced

the intended success (Hepp, 2007). One of the main

reasons for this back-lash is due to the lack of

understanding how changes in business processes

entails proper and controlled changes in the

corresponding CM as well as the specifications and

changes of the interfaces of Figure 2. For enterprises

in development and change, a control of ontological

framework and support for change of contexts and

interfaces is crucial for success. Our Robust

abbreviation-based approach suppor-ting change

(Section 3) gives some pointers to those ends.

ICEIS 2009 - International Conference on Enterprise Information Systems

202

3 ROBUST ABBREVIATION-

BASED CHANGES OF

CONTEXT

As illustrated in our examples of ESC and ATC in

Section 2 abbreviations can be seen as compiled

support of context dependent workflows. The

compiled version of workflows is sets of states. In

order to control and support changes of

abbreviation-based support we need to have a more

abstract model of states, that is situations, and

information flows to support robust changes of

Context Models, Interfaces, and Contexts (Figure 2).

The remaining part of this section is thus

addressing the following issues:

Section 3.1 Information modelling

Section 3.2 Common ground and coordination

Section 3.3 Situations and workflows

Section 3.4 Our approach of robust change

3.1 Information Modelling

Keith Devlin (Devlin, 2001) has provided a semantic

based logical framework supporting understanding

and structuring of information (InfoSense). The

logical framework has been influenced by the work

of Barwise and Perry (Barwise and Perry, 1999) at

Stanford in the 1980’s. They developed their

theories in order to understand human languages as

communication of meaning, semantics and

pragmatics. Suitable adaptations of the theories will

provide us with models and techniques to address

types of situations and hence workflows (Brandt,

2007; Östlund, 2007).

Devlin’s logical framework in structuring of

information can preferably be seen as a high level

description of information exchange. The connection

between Information (understood, or interpreted, by

a human agent) and its Repre-sentation is captured

by the following equation:

Information = Representation + Interpretation

The equation describe that information are

visible via a representation. The representation could

for example be a book, a computer system or

similar. The Interpretation describes the inter-

pretation capabilities of the receiving agent. As an

example, we have a situation of a fire, and a rescue

person sees smoke. The rescue person makes the

general assumption that there is a fire, since smoke

implies fire. Thus the constraint of the rescue

person’s knowledge about smoke and fire makes

him understand that this ‘type’ of seeing smoke, are

related to the ‘situation’ fire. One of Devlin’s basic

contributions in InfoSense is to clarify the relations

between representations and the proper

interpretations by users to identify the intended

situations (contexts).

The exchange of information between a sender

and receiver can be described as follows (Figure 2):

The sender, in figure 2, wants to inform the

receiver of a Situation S. The Representation of the

situation is described by a sequence of abb-

reviations A

that is fed into the Sender interface of

the artifact. The sequence A

is processed by the CM

and produces a output sequence of abbreviations A

The receiver interprets A

and can infer the Situation

S. If we assume that the syntax based processing is

correct and A

and A

have agreed upon semantics

then the sender has successfully informed the

receiver about the situation S and proper actions can

be taken. Agreed upon semantics of situations are

denoted common grounds (Devlin, 2001).

3.2 Common Ground and

Coordination

Common ground between stakeholders thus enables

correct abbreviation based semantic information

exchange related to situations. In abbreviation based

information exchange as in our examples ESC and

ATC the common ground is the agreed upon

interpretation of sets of abbreviations. Trusted

coordination in those teams can thus be assured by

proper training of skills mapping between situations

and sequences of abbreviations. Abbreviations can

thus be seen as coordination mechanisms in ESC,

ATC and similar applications.

3.3 Situations and Workflows

In Section 2 we identified that the workflow in a

ESC could be identified by 4 states. These states are

in fact compilations of corresponding four context

dependant Situations; S

, S

, and S

To enable a principled change of contexts in

abbreviation based coordination a first step is to

identify the corresponding set of situations that

underpin the workflows at hand. These are complex

tasks, not the least from a validation perspective. In

our ESC case we have identified and validated a

proper set of situations covering the relevant

workflows. Proper methods and tools to that end

include: work practise, ethnography, and situation

theories (Lundberg, 2007, Brandt, 2007, Östlund,

2007, Barwise-Perry, 1999, Devlin 1991, 2001).

ROBUST APPROACH TOWARDS CONTEXT DEPENDANT INFORMATION SHARING IN DISTRIBUTED

ENVIRONMENTS

203

3.4 Robust Change of Contexts

We propose a robust approach for context dependent

information modelling in critical information

infrastructures. A basic information process is

coordination (Chen, Sharman, Rao, Upadyaya,

2007). Coordination could be at different system

levels and between different system components. In

Figure 1 we model coordination at the highest

system level, that is, between system actors (agent or

users).

Figure 3: Interaction points between sender and receiver.

To be more specific, we introduce, in Figure 3,

interaction points in dialogues between agents to

enable support for different coordination aspects of

Figure 2.

Overlaying the interaction points with basic

information sharing of Figure 2 we recognise that

the lower abbreviation based interactions, of Figure

3, are facilitated by the Context Model processing of

data. The interpretation of those results is by the

sender and receiver in the given context and the

shared common ground supported by the

abbreviations.

The main reason for introducing interaction

points is that they give a natural structure of

coordination between sender and receiver. High-

level interaction points are focusing on the

contextual information sharing, whence low-level

interaction points are related to the information

processing system. Complex coordination can be

modelled using interaction points. Furthermore,

interaction points capture the critical coordination

challenges we have to address and maintain at the

levels of common ground and processing. We also

have to address the trustworthiness of translations of

sequences of abbreviations between those levels.

Our model supporting robust change of contexts

in abbreviation based information exchange is

founded on the following steps:

1. Identify the set of situations and

corresponding workflows that can be

inferred from the set of abbreviations.

2. Validate mapping from situations,

workflows to sequences of

abbreviations

3. Describe the new context and

workflows given the identified set of

situations

4. Make a mapping of the new workflows

on sequences of abbreviation

5. Validate mappings and introduce

training of mappings among teams.

Furthermore, our approach to abbreviation based

CM handling can be a basis for further

investigations on causes of brittleness, and

establishment of common grounds as well as off-

and on-line training of skills. Principled maintenance

of CM due to changes of contexts is also supported.

4 OTHER APPROACHES

Ongoing research and development on Web services

and Semantic web are focusing on ontologies and

schemas, i.e., on syntax-based structures (Hendler,

2008). Message passing between web services are

facilitated by SOAP messages encoded in XML. A

SOAP message between a seller and buyer could

have the message:

<orderstatus>confirmed</orderstatus>

The issue here is to have a consistent

interpretation of the abbreviation “confirmed”.

Again we have the problem of abbreviation-based

semantics! The works on syntax-based (ontology-

based) abbreviations in semantic web have the same

shortcomings as abbreviations discussed in Section

2. However, those ontology-based abbreviations

could be very helpful in defining the corresponding

Situation-based information flows, supporting

semantics in a given context, as outlined in Section

Approaches as (Veale, 2008) where the focus

area is limited have a clear and well defined

approach. However, most of the current approaches

are still based in the syntax area. Our top-down

approach to shared semantic based information takes

a supplementary view.

Interaction

oints

Abbreviation

based message

exchange

Context based

information

exchange

System

levels

Sender Recieve

ICEIS 2009 - International Conference on Enterprise Information Systems

204

5 CHALLENGES REVISITED

AND CONCLUSIONS

In Section 1 and 2 we identified two challenges in

abbreviation-based information sharing in teams:

• Change of context

• Misunderstandings

In Section 3 we outlined how we can identify the

abstract Situation types and information flows in

corresponding Context. A shift from localized

context to a distributed one is facilitated by

implementing the derived situation-based workflow

and defining a new set of abbreviations to support

the distributed information exchange as depicted in

Figure 2.

Dealing with misunderstandings due to identical

abbreviations can be solved by identifying those

ambiguities using the common ground. Resolving

ambiguities can be established in several ways. One

way is to distinguish the different contexts by

prefixes. For example; pilot-holding and tower-

holding in the given example presented in section 2.

The process outlined thus establishes a robust

model supporting abbreviation-based changes of

contexts. It can with advantage be used in training

situations as-well and to identify implicit brittleness

(Nardi, O’Day, 1999).

Future work includes more elaborated models

and tool supporting translation of abbreviations into

rule sets and information types. Taking into account

cognitive modelling and building rules from info

senses constraints would be an interesting approach

to consider.

REFERENCES

Barwise, J., Perry J., 1999, Situations and attitudes. CSLI

Publications

Brandt, P., 2007, Information in use – aspects on

information quality in workflows. BTH Dissertation

series, Blekinge Institute of Technology, ISSN:

1653:2090

Chen, R., Sharman, R., Rao, H., R., Upadhyaya, S., J.,

2007, Coordination in emergency response manage-

ment, Communications of the ACM, vol 51, nr 5

Devlin, K., 1991, Logic and information, Cambridge

University Press

Devlin, K., 2001, Info sense, W.H. Freeman and

Company, New York

Hendler, J., Shadbolt, N., Hall, W., Bernes-Lee, T., and

Weitzner, D., 2008, Web science: an interdisciplinary

approach to understanding the web. Communications

of the ACM, vol 51, no 7, 07/08

Hepp, M., 2007, Possible ontologies – how reality

constrains the development of relevant ontologies,

IEEE Internet comuting, published by IEEE Computer

Society

Lundberg, J., 2007, Principles of workflow support in life

critical situations, Blekinge Institute of Tech-nology,

ISBN/ISRN: 978-91-7295-105-1

Nardi, B. A., O’Day, V. L., 1999, Information ecologies-

using technology with heart, MIT Press

Veale, T. and Hao, Y., 2008, A context-sensitive frame-

work for lexical ontologies. The Knowledge

Engineering Review, Vol. 23:1, 101.115, Cambridge

University Press.

Östlund, L., 2007, Information in use – in and out

sourcing aspects on digital services. BTH Disserta-

tion series, School of engineering, Blekinge Institute

of Technology, ISSN 1653-2090, 2007

ROBUST APPROACH TOWARDS CONTEXT DEPENDANT INFORMATION SHARING IN DISTRIBUTED

ENVIRONMENTS

205