THE TECHNICAL OPPORTUNITIES AND SEMIOTIC

PITFALLS OF MULTI-ACTOR SYSTEMS

Support, Planning and Communication between Marketing, Dispatchers and

Passengers within the Netherlands Railways

Niels R. Faber and René J. Jorna

Department of Social Sciences, Fryske Akademy (KNAW), Doelestraat 8, Leeuwarden, The Netherlands

Faculty of Economics & Business, University of Groningen, P.O. Box 800, 9700AV Groningen, The Netherlands

Keywords: Multi-agent Systems, Planning, Problem Solving.

Abstract: In innovating its planning processes, passenger involvement in dispatching is one of the directions the

Netherlands Railways is exploring. Multi-agent systems provide a way to study organizational aspects of

such a change in the dispatching task that aims to bridge cultural differences between passenger and

dispatcher. In this study, the cognitive, coordination and semiotic implications are investigated. Two

versions of a multi-agent system have been constructed: NS-MAS 1 and NS-MAS 2. The involvement of

active passengers as is realized in NS-MAS 1 provides the organizational specifications of realizing

dispatcher-passenger communication. Furthermore, this implementation provides indications for bridging

the cultural differences between passengers and dispatchers. NS-MAS 2 operates with passive passengers,

simulated based on statistical data on passenger movements, and indicates the coordination possibilities

involved with passenger involvement in dispatching.

1 INTRODUCTION

The realization and support of planning and

dispatching in the Netherlands Railways

(Nederlandse Spoorwegen, NS) requires problem

solving, technical and communication skills of

various stakeholders. The stakeholders are diverse.

They include planners and dispatchers who plan, re-

plan and communicate, the marketing department

that wants to maximize customer satisfaction and the

passengers who want to travel from A to B as

convenient as possible. Essential in balancing the

various parties in planning is the determination and

valuation of constraints and goal functions. Planning

is a coordination task and “is the attunement or

assignment of multiple object types (such as

carriages, ticket collectors, engine drivers and arrival

and departure times), taking into account constraints

and goal functions” (van Wezel et al., 2006). As

long as only planners and dispatchers make the

plans, the choices and priorities of constraints and

goal functions are reasonably coherent and

unanimous. However, as soon as other parties enter

the problem solving space, conflict constraints and

contradictions may arise. For example, if reduction

of travelling time of a passenger in case of a detour

after an incident is taken into account by a

dispatcher - and this is not the normal situation -

other than the usual plans have to be generated and

other constraints become relevant. For planners, who

only think in terms of “replacing carriages over time

and place”, the concrete goal function of minimizing

detour time and communicating these goals with

non-planners, is new and therefore different. Then

besides technological and algorithmic skills, cultural

backgrounds affect the interaction between people in

the planning situation (Daft, 2001). A group’s

rituals, norms, and symbols result in a collection of

beliefs that is specific to that group. This collection

of beliefs governs the behaviours of the individuals

within the group. Because the rituals, norms, and

symbols are group specific, they are often a cause

for misunderstanding within group interactions.

The case that is presented here resides within the

context of the NS. The NS daily transports one

million passengers. Transportation takes place with

the help of 2,700 railroad carriages, which

approximately run 5,000 train services per day. The

173

Faber N. and Jorna R.

THE TECHNICAL OPPORTUNITIES AND SEMIOTIC PITFALLS OF MULTI-ACTOR SYSTEMS - Support, Planning and Communication between Marketing, Dispatchers and Passengers

within the Netherlands Railways.

DOI: 10.5220/0003267701730181

In Proceedings of the Twelfth International Conference on Informatics and Semiotics in Organisations (ICISO 2010), page

ISBN: 978-989-8425-26-3

trains run between 384 stations in the Netherlands.

Within the NS, five kinds of planning divisions

exist. The first concerns timetables and other plans.

The second concerns the partitioning in planning

rolling stock and planning rolling staff. The third

concerns the partitioning in local planning and

central planning (of stock and staff). The fourth

concerns the distinction in year plan (long term) and

day plan (short term), again of stock and staff and

the last is dispatching, meaning solving problems at

the day of execution because of accidents, delays

and detours. Overall approximately 400 planners are

continuously involved in making plans and

schedules.

Our study only concerns dispatchers. Until

recently, planning within the NS was only a matter

for planners and dispatchers. Because of the

increasing technological possibilities of planning

support, advanced telephone communications, AI

and agent software, and Internet, the NS is studying

the influence of the marketing department and of

preferences of active and passive passengers. We

will explain the details later. This changing

perspective and its complicated consequences is

studied with the help of Multi-Actor Systems (NS-

MAS). As the term MAS already indicates various

(kinds) of actors are then involved jointly solving

the re-planning or dispatching puzzle. Kinds of

actors are of course the human dispatchers, but also

software agents with different levels intelligence.

Apart from its technical implementation, a MAS

requires attention for and decisions about a) what

kind of information is relevant for whom at what

time and who understands this information, b) how

the coordination between actors is realized, who is

responsible for what and c) which minimal

requirements regarding signs and symbols are

relevant for meaningful communication. The first

question requires a cognitive answer or perspective,

the second an organizational answer and the third a

semiotic answer (Klos, 2000; van den Broek, 2001;

Helmhout; 2006). We will come back to this.

The dispatching task involves a series of actions

performed by a dispatcher that are intended to

recover from a disruption in the railways network,

with the objective to restore the original train

timetable as quickly as possible. Disruptions are

delays, train or railway breakdowns, or other causes

for the train service to deviate from the planning,

which are the cause for imbalance of available

material (i.e. trains and wagons) and personnel (i.e.

engine drivers and ticket collectors). The objective

of the dispatcher’s task is re-planning for certain

periods of time (e.g. 4 hours). This objective

narrows the dispatcher’s task down to a problem

solving activity (Newell & Simon, 1972; Simon,

1977; Laird et al., 1986), in which material,

personnel, and timetable comprise the problem

space. From this perspective, the dispatcher is

concerned solely with the components that relate to

the Netherlands Railways transport service. Until

now, dispatchers do not take into account actual

(individual or aggregated) passengers, their

preferences, or marketing goals.

For most transport situations of the Netherlands

Railways, the load consists of passengers who use

train service as a means to get from A to B. Unlike

cargo, passengers are actors that behave

intelligently. They are able to deal with delays in

their planning, by choosing the exact train they will

use taking into account some slack. Delays that

exceed this slack time result in passengers unable to

complete their journeys as planned. The specific

solution the dispatcher implements to overcome a

disruption does not explicitly include the desires of

affected passengers. The chosen solution might

benefit a particular passenger. However, because

neither passengers nor their desires are considered

when a dispatch action is devised, no certainty about

the effects of a dispatching action exists for

passengers.

This research wants to bridge the distinct

cultures of dispatchers, marketers and passengers,

constructing a multi-actor system that connects

dispatchers’ practices to passenger experiences and

preferences. Such inclusion of passengers in

dispatching changes the original problem space.

Where in the original situation the problem space

only contains non-cognitive elements (i.e. timetable

and trains) and few cognitive elements (i.e.

personnel), the new problem space will contain

many intelligent agents (i.e. passengers and AI

software). The purpose of the multi-actor system

(NS-MAS) is threefold.

First, the system’s objective is to provide insights

into the required communication and possible

coordination structures between dispatcher and

passengers. Currently, dispatcher and passenger only

communicate indirectly and one-way through the

measures taken by the dispatcher. In order to take

into consideration passenger preferences in

developing a dispatching measure, two-way

communication between dispatcher and passenger is

required. How such two-way communication should

be organized needs to be determined. Coordination

in dispatching is absent in the current situation: the

dispatcher decides which dispatching measure is

taken; passengers play no role. Actively involving

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

174

passengers in dispatching opens new coordination

possibilities, for then an additional deciding actor,

the passenger, becomes part of the dispatching task.

The second objective of the study is to

investigate the role of knowledge in passenger

involvement in the dispatching task. This objective

links to two complementary sorts of knowledge. The

first sort concerns the knowledge that needs to be

included about passengers in order to realize

passenger involvement. For perspectives are

different between dispatcher and passengers,

knowledge about passengers will be needed to

bridge the difference and facilitate knowledge

exchange. The second sort concerns the question

what knowledge a passenger should provide in case

s/he is assigned an active role in dispatching.

The third, more practice oriented objective of the

multi-actor system is to provide insights into the

effects of alternative dispatching measures on

passenger opinions. Alternative measures might

result in equally suitable solutions from the

traditional perspective (i.e. balancing material and

personnel, and restoring the original timetable as

quickly as possible) but render different responses

from affected passengers. This is interesting for the

marketing department of the NS

In this paper, we discuss the developed multi-

actor system in more detail, focusing specifically on

its functionality and the effect on the dispatching

task. The MAS has been developed in two versions

(1 and 2). Each version has its specific functionality.

The first version (NS-MAS 1) focuses on realizing

communication between dispatcher and real, active

passengers, through for instance a mobile device.

The second version (NS-MAS 2) concerns an

extension of the first version, enabling

communication between dispatcher and simulated,

passive passengers, and enabling the use of more

complex dispatching measures. In the second

version, statistical data was used to initialize these

passive passengers. We first provide an overview of

the functionality that was realised in NS-MAS 1.

Second, the functionality of the second version is

discussed in detail. Both versions of the multi-actor

system have been implemented in the Java Agent

DEvelopment Framework (Jade, 2007; JADE;

Bellifemine et al., 2007). The Prometheus Design

Tool and methodology (Prometheus, 2007; Padgham

& Winikoff, 2004) have been used to design the

multi-actor system.

2 NS-MAS WITH ACTIVE

PASSENGERS (NS-MAS 1)

In a MAS, humans and software agents collaborate

to solve the problem at hand and construct a solution

that combines logistics and passenger preferences

(Gazendam, 1990; 2003). Besides humans

(dispatchers and passengers), two types of software

agents are used. First, relatively autonomous agents

are part of a MAS. These agents ensure the internal

functioning of the MAS, primarily relaying

messages. Second, NS-MAS 1 consists of intelligent

assistants. Identified passengers are represented as

intelligent assistants. If a passenger identifies

himself to the MAS, an intelligent assistant is

constructed to represent him. In addition, a

dispatcher is assigned an intelligent assistant. The

agents differ in terms of cognitive complexity and

cognitive possibilities. The questions in the research

are threefold. First, to what extend should intelligent

software assistants have search and decision

authority. Second, what kind of coordination

mechanism is suitable to combine humans and

various kinds of software agents, and third what

does communication entail in terms of signs and

symbols.

The objective of the system has been to enable

the incorporation of passenger preferences in

dispatching, such that dispatchers would consider

passengers in the solutions they develop. In NS-

MAS 1, passengers register themselves with the

multi-actor system and specify their travel plans for

the journey(s) they will make at a specific time and

date. These passengers are called active passengers,

for they are able to actively communicate with the

dispatcher by means of intermediary software

agents.

Taking passenger preferences into consideration

demands that dispatchers and passengers are able to

communicate. Version one of the multi-actor system

realizes two-way communication between dispatcher

and active passengers; a dispatcher communicates

timetable information to passengers (i.e. the effect of

a disruption on the timetable, and of the dispatching

measure s/he suggests on the timetable), and

passengers are able to react and respond to these.

Dispatching measures, devised by the dispatcher, are

communicated to the multi-actor system, which

forwards these towards passengers’ mobile devices.

After being informed about a planned dispatching

measure, a passenger can respond by assigning it a

grade from 1 (undesirable) to 10 (very desirable).

All passenger responses are gathered and presented

back to the dispatcher. The dispatcher can then

THE TECHNICAL OPPORTUNITIES AND SEMIOTIC PITFALLS OF MULTI-ACTOR SYSTEMS - Support, Planning

and Communication between Marketing, Dispatchers and Passengers within the Netherlands Railways

175

decide whether the suggested dispatching measure is

implemented or an alternative solution needs to be

developed.

Though presented as a straightforward process,

communication between dispatcher and passengers

concerns a complex interaction involving

fundamental differences in culture and language.

Dispatchers are concerned with restoring timetable

functioning, and think in terms of timetable

information. The solution that restores the original

timetable as quickly as possible is regarded as the

best solution. In contrast, passengers only care about

the journey for which they use the train service. For

the NS-MAS passengers are modeled such that they

use the concepts travel time, train changes, and

comfort in relation to their journey. Any disruption

or delay that hinders them to finish their journey in

the planned time is undesirable. Organizing

communication between dispatcher and passengers

has been the main functionality that was

implemented the first version of the system.

Realizing this function required two main elements.

First, the communication channel needed to be built.

Second, the language difference that exists between

dispatcher and passengers needed to be bridged.

Communication between dispatcher and

passenger is realized through three main software

agents, namely the Planner, TravelManager, and

TravelCoach agents. The Planner and TravelCoach

agents are intelligent assistants and form interfaces

between human actors and the NS-MAS. The

Planner agent interfaces between (human) dispatcher

and NS-MAS. The Planner agent receives inputs

regarding delays and disruptions (see Figure 1), and

dispatching measures from the dispatcher, and

forwards these to the TravelManager agent.

Reversely, the Planner agent transmits passenger

response information it receives from the

TravelManager to the dispatcher. The TravelCoach

agent is the interface between the system and one

passenger, and interacts with the passenger

regarding travel information, i.e. delay and

disruption information, and dispatching measures.

The TravelManager agent connects the Planner and

TravelCoach agents, and facilitates internal

communication within the multi-actor system.

Figure 2 shows the graphical user interface of the

TravelManager agent, used to monitor and control

the agent’s behavior.

The content of the communication has been

conceptualized in one ontological model, integrating

dispatcher, passengers and various software agents.

The main distinction between concepts used by

dispatcher or passenger lies in the level of

Figure 1: Planner user interface.

Figure 2: TravelManager user interface.

aggregation and abstraction. Due to his task, a

dispatcher considers the railroad network as a whole,

especially the train services using the railroad

network in the region he is responsible for. A

dispatcher’s knowledge of the railroad network

consists of start and end station of the line,

intermediate stations, the line’s timetable, and other

parts of the railroad network the line passes.

Passengers consider the same railroad network from

the perspective of their individual journeys, which

translates to one specific route that is taken by the

passenger from his / her place of departure to his /

her destination, stations where s/he needs to change

trains and particular times involving the actions of

boarding and alighting.

Dispatchers formulate dispatching measures in

terms of adaptations of lines, speeding up or slowing

down a line at specific points of its path, removing

or adding stops to its path, shortening the line’s path

by making the line return before reaching its final

station, or taking the line out of service completely.

Such dispatching measures need to be translated to

the route that the individual passenger travels.

Within the NS-MAS, the first step is to identify

those passengers that are affected by the dispatching

measure. This is the responsibility of the

TravelManager agent. Because active passengers

have registered themselves and the travel plans of

the journeys they make, the TravelManager is able

to lookup the affected active passengers. After

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

176

having determined which passengers are affected,

the TravelManager determines the effect of the

dispatching measure on the travel plan of the

affected agents. This agent subsequently constructs

new travel plans for each individual active passenger

and communicates these plans to these passengers’

TravelCoach agents. The TravelCoach agents show

the new travel plans to their connected active

passenger and ask for a response. The passenger

responds by assigning a grade between 1 and 10 to

the new travel plan. All grades from affected

passengers are communicated back by their

TravelCoach agents to the TravelManager. Once all

responses have been gathered, the TravelManager

aggregates and categorizes the responses of the

affected active passengers into five categories,

reaching from “dissatisfied” to “satisfied”.

Frequencies per category are communicated to the

Planner agent, which in its turn informs the (human)

dispatcher about the outcome.

From a cognitive perspective, the characteristics

of the software agents are poor. They are

communication and ordering agents with the help of

simple algorithms. From an organizational view, the

leading coordination mechanism is authority and

hierarchy. The dispatcher is the boss and the

ultimate decision maker. The semiotically

interesting points are that dispatcher and passenger

use a different semantics whereas software agent can

only exchange meaningful information by means of

the human actors (Jorna, 2009; Helmhout et al.,

2009).

3 NS-MAS WITH PASSIVE

PASSENGERS (NS-MAS 2)

The second version of the NS-MAS extends the

initial version in two directions. First, the system is

extended to be able to use statistical data about

passenger movements through the railroad network

to construct (aggregated) simulated passengers. To

contrast them with active passengers, these

simulated passengers are labeled passive passengers.

Passenger movement data that has been gathered

consist of data about the station of departure and

destination, ticket type used for the journey, travel

motive, and the frequency a passenger travels by

train. These data are used to construct passive

passenger agents in NS-MAS 2. The second

extension concerns the handling of more complex

dispatching measures. The initial version only is

capable to process changes in times and stops of

train lines. In NS-MAS 2, dispatchers are able to

introduce detour scenarios to passengers. Whenever

a part of the railroad network is out of service due to

for instance a derailed train, a detour scenario

provides dispatchers the addition to relay passengers

around the blocked part of the network.

Additionally, passengers are provided a solution to

continue their journey with only a minimum delay.

The two extensions that have been realized in NS-

MAS 2 imply various changes to the initial

prototype. The required changes are discussed

subsequently, starting with the extension to

incorporate statistical data to simulate passive

passengers.

Three software agents have been added in NS-

MAS 2 to implement the handling of statistical data

to simulate passive passengers. First, the

StatisticalManager agent manages all statistical data.

Upon receiving a request to provide data about a

specific line, this agent responds with the amount of

passengers that make use of that particular line and

their travel plans in terms of station of departure and

destination. The CommunicationManager (Figure 3

shows its graphical user interface) is the

intermediate between the TravelManager and the

StatisticalManager and responsible for creating

passive passenger agents. Finally, a

StatisticalPassenger agent represents a passive

passenger. The StatisticalPassenger agent provides a

response to any dispatching measure it receives,

similar to the response an active passenger provides

(a grade between 1 and 10).

Communication between agents largely follows

the same pattern as in NS-MAS 1. The

TravelManager remains responsible for

communication between the dispatcher (Planner

agent) and the passengers (multiple

StatisticalPassenger agents). The main difference

between NS-MAS 1 and NS-MAS 2 relating to

communication is associated with the creation of

passenger agents. The communication that ensures

correct passenger agent creation is displayed in

Figure 4. Prior to starting communication with them,

passive passenger need to be instantiated. Upon

reception of delay or disruption information, the

TravelManager agent forwards this information to

the CommunicationManager, requesting the creation

of affected StatisticalPassenger agents.

When receiving a message from the

CommunicationManager that passenger creation has

finished, the TravelManager forwards delay and

disruption information to the StatisticalPassenger

agents.

THE TECHNICAL OPPORTUNITIES AND SEMIOTIC PITFALLS OF MULTI-ACTOR SYSTEMS - Support, Planning

and Communication between Marketing, Dispatchers and Passengers within the Netherlands Railways

177

Figure 3: CommunicationManager user interface.

Communication between TravelManager and

StatisticalPassenger agents about dispatching

measures and passenger responses follows the same

communication pattern that exists between

TravelManager and TravelCoach agents as described

in the previous section.

Enabling detour scenarios in the multi-actor

system required a change in the content of

communication between the TravelManager and

StatisticalPassenger agents. The content of

communication closely follows the implementation

of the initial version. Again, an ontological model is

used to bridge the different views of the railroad

network that exist between dispatcher and passenger.

Figure 5 shows the ontological model that has been

constructed for NS-MAS 2, in UML (OMG, 2009)

notation. However, in the initial version the Travel-

Manager agent was responsible for translating

dispatching measures to the travel plans of

individual passengers. In NS-MAS 2, the

TravelManager only communicates his dispatching

measures in terms of changes he has made to the

original timetable of lines to the StatisticalPassenger

agents. The latter translates these changes to their

own travel plans. Also, the TravelManager

communicates replacing lines to passengers. If for

instance, a passenger cannot complete his journey

because s/he arrives too late to get onto a connecting

train, the TravelManager provides the information

about an alternative train in the same direction. In a

similar fashion, the TravelManager agent is able to

communicate detours to passengers, replacing lines

within a passenger’s travel plan with an alternative

route. With NS-MAS 2, the TravelManager agent’s

routing and travel plan calculations have been

distributed to the individual StatisticalPassenger

agents. Additionally, StatisticalPassenger agents

have been made more ontologically rich, for these

agents are equipped with knowledge to understand

and process changes to lines, in addition to

knowledge about their own travel plan. The behavior

of StatisticalPassenger agents however, has been

implemented as a mathematical function that

calculates a comfort grade. The comfort grade is

calculated by comparing the original travel plan with

the new travel plan that incorporates the suggested

dispatching measure. The comfort grade is a

function of the travel time of the original travel plan

and of the new travel plan.

Figure 4: Communication specification NS-MAS 2.

From a cognitive point of view, the software agents

are much richer than in NS-MAS 1. They combine,

order, and integrate as if they were intelligent actors.

They do more than just handle and exchange

messages. From an organizational point of view, the

dispatcher still is the boss, but he is working with

intelligent software agents that take work out of his

hands and that he has to trust. It is therefore still

authority that is the coordination mechanism, but the

question can now easily be formulated at what

moment an in which circumstances can the software

agents be autonomous? From a semiotic point, the

communication is less rich than in NS-MAS 1.

Semantic and pragmatic considerations are left out

of NS-MAS 2.

4 THE OPPORTUNITIES

AND PITFALLS OF NS-MAS

Traditionally, dispatching is a problem-solving task

within a predefined problem space (Simon, 1977). A

dispatcher is required to as quickly and efficiently as

possible restore a train service according to the

original timetable. Within this space of timetable,

train lines, railroad network, personnel, and train

material the dispatcher is able to devise any solution

that meets a set of criteria. Passenger desires need

not be considered in these kinds of solutions.

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

178

Figure 5: NS-MAS 2 ontology.

Dispatching in a new NS-MAS sense, however,

develops towards a form of dispatching in which

passengers, i.e., the customers, take a much more

important position. No longer will dispatching

involve problem solving in a complex, but clearly

defined problem space. The incorporation of

passenger desires into dispatching requires

communication with these passengers at the moment

dispatching is required (i.e., at the moment a delay

or disruption occurs in train services). This

extension of the problem-solving task of the

dispatcher increases its complexity even further. The

dispatcher needs to combine known, stable

components from the original problem space with

unpredictable, intelligent passengers; unpredictable

regarding the specific journeys passengers make,

their travel motives, and their positions towards

changes in their travel plans. Summarizing, the

dispatcher’s task transforms from a reasonably well-

structured into an ill-structured problem-solving task

(Simon, 1973).

Prior to the start of this study for the NS, no clear

idea had been developed of how to organize

interaction between dispatcher and passenger, nor

was clear what knowledge is involved in such

interaction. The first version of our NS-MAS

provides two main insights. First, it indicates how

communication between dispatcher and passenger

needs to be structured. Essentially, communication

between dispatcher and passengers is a one to many,

two-way communication pattern. A dispatching

measure needs to be communicated to all affected

passengers. Reversely, a dispatcher needs to receive

one clear overview of passenger responses to a

dispatching measure s/he suggests. In our NS-MAS,

the TravelManager has been created as the pivot

between dispatcher and passengers. The

TravelManager sends a dispatching measure to all

affected passengers and receives and aggregates

passenger responses and sends this as one overview

back to the dispatcher. Summarizing, this NS-MAS

provides a structure to organize communication

between dispatchers and passengers.

The second insight by the first version of our

NS-MAS is the required knowledge that enables

dispatcher-passenger interaction. Plainly

communicating a dispatching measure to passengers

will not. In such communication, knowledge of

passengers and dispatcher regarding the railroad

network and train services is too different. During

the construction of NS-MAS 1, knowledge about the

travel plans of each individual passenger has been

identified as crucial in order to realize sensible

interaction between dispatcher and passenger.

Knowing the travel plans of individual passengers

enables the translation of dispatching measures to

the effect these measures have for passengers.

Receiving the effect of a dispatching measure on

their own travel plans enables passengers to

understand the measure, and enables them to

respond knowledgeably. Hence, incorporating

knowledge about passenger travel plans is a

THE TECHNICAL OPPORTUNITIES AND SEMIOTIC PITFALLS OF MULTI-ACTOR SYSTEMS - Support, Planning

and Communication between Marketing, Dispatchers and Passengers within the Netherlands Railways

179

Table 1: Overview cognitive skills / coordination in NS-MAS 2.

Agent Dispat-cher Planner Travel-

Manager

Active

passenger

Passive

passenger

Agent /

actor

actor agent agent actor Agent

Cognitive

skills

- Change

problem

space

- Apply

weights

- Search

within

problem

space

- Pass

through

information

- Search

problem

space

- Calcul-

ating

Coordina-

tion role

Boss Slave None /

combine

None None

necessary step towards enabling dispatcher-

passenger interaction.

NS-MAS 2 circumvents the largest disadvantage

of the initial version, namely the requirement that

passengers register their travel plans prior to starting

their journey, and that passengers actively respond

to suggested dispatching measures. We assume that

only a minority of passengers will register their

travel plan. This renders the NS-MAS of limited use

to dispatchers. Incorporating statistical data about

passenger movements, travel motives, travel

frequency, and ticket types to create passive

passengers removes the dependency on registration

of travel plans by passengers. In this way,

dispatchers are able to use the NS-MAS for

simulation purposes, exploring the responses of

passenger to dispatching measures.

In addition to removing the necessity to have

passengers registering their travel plans, NS-MAS 2

provides the possibility to specify more complex

dispatching measures, than are currently available to

dispatchers. Currently, dispatching measures are

formulated similar to dispatching in the initial

version of the system. Dispatchers only are able to

specify that a line is slowed down, or speeded up, or

that a line stops at more or less stations than its

normal service. Providing detours to passengers

currently only is provided at an individual basis by

ticket collectors, only in response to a passenger’s

request. NS-MAS 2 enables dispatchers to explore

the effects of detours, thus enlarging the portfolio of

dispatching measures they have at their disposal.

This initial version consists of agents that show

no cognitive skills, and cannot behave

autonomously. According to Wooldridge (2002), the

agents in version one therefore are not agents. NS-

MAS 2 in contrast, houses agents that are

cognitively richer, and are able to respond

autonomously to suggested dispatching measures

from the dispatcher, namely the StatisticalPassenger

agents. Furthermore, version two only deals with

one human actor: the dispatcher.

summarizes the various agents existing the in

the NS-MAS, their cognitive abilities, and their roles

in coordination. The presented NS-MAS is a hybrid

system; it connects human actors and software

agents. Together, actors and agents participate in the

problem-solving task of dispatching. However, the

two versions of the NS-MAS facilitate this

cooperative mode of problem solving in distinct

ways. In NS-MAS 1, software agents only relay and

transform messages between dispatcher and

passengers.

This initial version consists of agents that show no

cognitive skills, and cannot behave autonomously.

According to Wooldridge (2002), the agents in

version one therefore are not agents. NS-MAS 2 in

contrast, houses agents that are cognitively richer,

and are able to respond autonomously to suggested

dispatching measures from the dispatcher, namely

the StatisticalPassenger agents. Furthermore, version

two only deals with one human actor: the dispatcher.

From an agent perspective, versions NS-MAS 1

and 2 show similarities from a coordination

perspective. In both versions the initiative and

decision making power lie with the human actor(s).

Dispatchers take the initiative in communication.

Eventually, they also decide what dispatching

measure is brought into effect. Passengers,

irrespective of being active or passive passengers,

have the possibility to express their thoughts about a

suggested dispatching measure. Their decision

making space however is limited to assigning a

grade to the suggested measure. Both versions show

a hierarchical coordination mechanism, in which the

dispatcher is the authority and holds decision power;

passengers are subordinates without any authority or

power. Passengers only provide their opinion about

the devised dispatching measure to the dispatcher.

The dispatcher still has the choice to consider these

opinions when choosing what measure to

implement.

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

180

Coordination within the multi-actor system

however does not have to follow the pattern as

described above, in which authority and power

remain with the dispatcher. The provided description

closely follows the current organization of the

dispatching task within the Netherlands Railways.

As indicated, the NS-MAS primarily facilitates

communication between dispatcher and passengers.

This communication is a necessary element in

coordination between dispatcher and passenger,

enabling a broader spectrum of coordination

configurations than the mechanism just described. A

possible scenario could be to distribute authority

among the passengers that are affected by a

disruption in train service, and let passengers

together come up with a solution that i) restores the

balance in material and personnel, ii) ensures train

service to continue according to schedule as quickly

as possible, and iii) aids affected passengers in

continuing their journeys or relaying their journeys

as comfortable as possible. Additionally, passengers

could be granted the decision power to implement

the suggested dispatching solution. In such a

scenario the dispatcher’s task shifts from a problem

solving to a coordination and implementation task.

REFERENCES

Bellifemine, F.L., Caire, G., & Greenwood, D. 2007.

Developing multi-agent systems with JADE. John

Wiley & Sons.

Broek, J. van den. 2001. On Agent Cooperation: The

relevance of cognitive plausibility for multiagent

simulation models of organization, PhD Thesis SOM,

Groningen University, Labyrinth, Capelle a/d IJssel.

Daft, R.L 2001. Organization: Theory and Design.

Thomson Learning Inc., Cincinatti.

Fowler, M. 2003. Patterns of Enterprise Application

Architecture. Addison-Wesley, Boston, MA.

Gazendam, H.W.M. 1990. Expert Systems Supporting

Organization and Information Management. In:

Masuch, M. (Ed.). Organization, Management, and

Expert Systems: Models of Automated Reasoning. De

Gruyter, Berlin, pp. 123-153.

Gazendam, H.W.M. 2003. Models as coherent sign

structures In: Henk W. M. Gazendam, René J. Jorna,

& Ruben S. Cijsouw (Eds.). Dynamics and change in

organizations: Studies in Organizational Semiotics.

Kluwer Academic Publ., Dordrecht, pp183-213..

Helmhout, M.J. 2006. The Social Cognitive Actor: A

multi-actor simulation of organizations. PhD Thesis

SOM, Groningen University. Labyrint, Ridderkerk.

Helmhout, M., R.J. Jorna & W.W.M. Gazendam 2009.

The semiotic actor: From signs to socially constructed

meaning. Semiotica.

Jade 2007. Java Agent DEvelopment framework (Version

3.5 revision 5988) [software]. Available from

http://jade.tilab.com

Jorna, R.J. 2009. Introduction: Organizational semiotics

and social simulation. Semiotica.

Klos, T.B. 2000. Agent-Based Computational Transaction

Cost Economics. PhD Thesis SOM, Groningen

University. Labyrint , Capelle a/d Ijssel.

Laird, J., Rosenbloom, P. & Newell, A. 1986. Universal

Subgoaling and Chunking. Kluwer, Boston, MA.

Luck, M., Ashri, R., & D’Inverno, M. 2004. Agent-based

Software Development.. Artech House, Norwood, MA.

Newell, A., & Simon, H.A. 1972. Human problem solving.

Prentice-Hall.

OMG 2009. OMG Unified Modeling Language (OMG

UML) Superstructure. Retrieved from http://www.

omg.org/spec/UML/2.2/Superstructure/PDF/

Padgham, L, & Winikoff, M. 2004. Developing intelligent

agent systems: a practical guide. John Wiley & Sons.

Prometheus 2007. Prometheus & PDT (Version 3.3)

[software]. Available from http://www.cs.rmit.edu.au

/agents/pdt/pdt.shtml

Simon, H.A. 1973. The structure of ill-structures

problems. Artificial Intelligence, 4(3-4), 181-201.

Simon, H.A. 1977.

The new science of management

decision. Prentice-Hall.

Van Wezel, W., Jorna, R.J. & Meystel, A., (Eds.) 2006.

Planning in Intelligent Systems: Aspects, motivations,

and methods. Wiley, New York.

Wooldridge, M. 2002. An introduction to multi-agent

systems. John Wiley & Sons.

THE TECHNICAL OPPORTUNITIES AND SEMIOTIC PITFALLS OF MULTI-ACTOR SYSTEMS - Support, Planning

and Communication between Marketing, Dispatchers and Passengers within the Netherlands Railways

181