COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES

Peter Bøgh Andersen

Department of Information and Media Studies, University of Aarhus

Helsingforsgade 14, 8200 Aarhus N, Denmark

Keywords: Activities, business processes, mobile computing, context-aware computing, semantic roles, semiotics.

Abstract: Mobile and context-aware technology enables new activity-centred ways of using digital technology that

require systematic methods for representing actions and activities computationally. The paper uses findings

from ethnography, linguistics and philosophy to paint a generic portrait of activities, and suggests ways of

representing it in an object-oriented framework. The paper makes a sharp distinction between the represen-

tation and the represented. The representations are not activities, they only represent them.

1 INTRODUCTION

This paper discusses the problem of representing

activities by means of computational media.

It is motivated by a new emerging use of tech-

nology, enabled by mobile and context-aware tech-

nology. From a situation centred on documents and

applications used from a stationary device, we are

moving into a dynamic situation where many docu-

ments and services are used intermittently in the

same activity, where activities are intertwined, and

where use of computers no longer takes place at a

certain place, but follows our daily activities as we

move around in the world (Bardram 2006). The fo-

cus is shifting from documents and applications to

the activities in which they are used.

Bardram (2006) suggests the name activity-

based computing to denote an architecture where

activities are the fundamental building blocks and

documents and services can be requested as needed

from the resources that happen to be present in the

environment.

A similar argument is made in Kristensen (2002,

2003). He criticizes object-centric methods and ad-

vocates the association as a good tool for represent-

ing the new kind of computer use. Associations al-

low dynamic patterns of cooperation and specify the

roles its participants may play. The notion of roles

recurs in Moran (2005) that sketches an architecture

for activity-based software.

Whereas Bardram mainly represents his activity

concept through software architecture, Kristensen

views associations as a first class citizen that moti-

vates new programming language concepts and ex-

plicit formalisms.

In this paper I address the question of how ac-

tivities can be represented computationally: 1. what

is to be represented, i.e. what are the characteristics

of activities, and 2. how may the digital representa-

tion look like. To answer the first question, we

should listen to ethnographers, sociologists, psy-

chologists, and linguists, whereas the second ques-

tion belongs to computer science.

Since it is a question of devising signs of some-

thing, I shall base myself on semiotics. The repre-

sentamen of the sign (in the following: representa-

tion) is a computational representation that can be

executed on a computer, its object is activities in

some problem domain, and its interpretant consist in

using these representations for conducting everyday

activities, guided by rules of interpretation.

Semiotics reminds us that the representation is

not identical to its object. Although the two can be

related through similarily (icons), they can also be

held together through causal relations (indexes, e.g.

via sensors and actuators), and they can in fact be

totally unrelated except via conventions (symbols).

Computer science has a strong tendency for pre-

ferring icons – there must be some kind of similarity

between the computational model and its object. In

the past, this has lead to an (often criticised) export

of good concepts for understanding the behaviour of

information systems to also cover an understanding

of the object referred to by the systems. For exam-

ple, since the class-diagrams or state-machines of

the Unified Modelling Language are excellent for

understanding the structure and behaviour of infor-

mation systems, they are also assumed to be good

Bøgh Andersen P. (2007).

COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - ISAS, pages 95-104

DOI: 10.5220/0002347600950104

 SciTePress

for understanding the part of the world represented

by the systems.

This has created conflicts between domain ex-

perts and computer scientists: domain experts wrin-

kle their nose at the simplistic nature of the compu-

tational representations, whereas computer scientists

quickly loose interest in the subtle distinctions of the

domain experts.

If we realize that the representation needs not to

be similar to the represented, the situation becomes

more tractable. There is one set of requirements that

should be fulfilled by the computational representa-

tion. Apart from being useful for representing their

object, given a suitable set of conventions, they must

connect to current programming wisdom, since oth-

erwise no system would be produced in the end.

Another set of requirements are valid for de-

scriptions of the object: they must account for what

is currently known about the object and preferably in

the simplest possible way.

The task is now to make a representation that is

in fact programmable and which, given a suitable set

of rules of interpretation, will be accepted as repre-

senting its object by its users.

Therefore: reproaching the programmer for

working with a simplistic view on human activities

is just as absurd as reproaching the filmmaker for

viewing reality as consisting of cuts, sequences, and

scenes, in spite of the fact that the film is accepted

by the audience as an exciting thriller. In both cases,

the acceptability of the product is filtered through

the users’ willing suspension of disbelief.

In the rest of the paper, the object – human ac-

tivities – is mainly described based on linguistic

evidence. The reason for this is that language is our

main empirical access to the way activities are seg-

mented and classified. When a field-work is started,

the events form a confusing flow where patterns and

boundaries are hard to see. The only method of

bringing order in the confusion is to ask people to

describe what they do and why.

If we want to capture the generic recurrent fea-

tures of activities, the best evidence is the grammati-

calized features of language, such as case inflexion,

word order, affixes, tenses, etc. The reason is that

these features mark distinctions that have been used

frequently during a long time. Therefore they proba-

bly represent basic distinctions in our understanding

of reality.

As for the computational representations, I shall

use two popular representations from the UML stan-

dard, namely class diagrams and state-machines.

The following preliminary requirements con-

cerning the nature of activities are based on field-

work projects in the maritime domain and hospitals

(Andersen 2001, 2004a, 2006, Bardram & Bossen

2005, Bødker & Andersen 2005).

Activities must encompass both material and

communicative actions, since they are intermingled

in practice. They must deal explicitly with errors

and failure of equipment, and provide countermea-

sures for such failures. Many activities contain stan-

dard countermeasures against past accidents. For

example, all pedestrians will look to the right and

left for approaching cars before crossing a road (see

also Andersen 2001 for a similar analysis of mari-

time manoeuvres).

Actors must be able to suspend and resume ac-

tivities, and cooperative activities with several peo-

ple involved must be possible. Most activities are

performed simultaneously with and intertwined with

other activities.

Activities must contain roles. In both reference

domains, there are formal modal hierarchies of roles

requiring particular abilities, knowledge, intentions,

rights and obligations. For example, the helmsman

must be able to turn the wheel but has no right to

determine the course, and certain physical abilities

(hearing, sight) are required to work on a ship in the

first place. In the hospital, a doctor must be present

when resuscitating cardiac arrests, nurses are not

allowed to do it alone.

The rest of the paper elaborates these character-

istics using findings from linguistics, psychology,

and philosophy.

2 ACTIONS AND ACTIVITIES

This section presents a selective overview of re-

search into the structure of actions and activities.

2.1 Roles

2.1.1 The Object of the Computational Rep-

resentation

Activity theory teaches us (Vygotsky 1962) that hu-

man actions are composed of (at least) three differ-

ent elements that play very different roles: the sub-

ject intending the action, the object towards which

the action is directed, and the mediator that mediates

between subject and object. Linguistics adds a larger

number of roles to these three, and connects the role

to specific process types (on the relationship be-

tween semiotics and activity theory, see Bødker &

Andersen 2005).

Roles in linguistics are called thematic or seman-

tic roles and denote relations between a process and

its participants. The theory (e.g. Fillmore 1968,

1977) claims that

ICEIS 2007 - International Conference on Enterprise Information Systems

– there are a limited number of thematic roles

(normally between ten and twenty),

– each process type can be characterized by requir-

ing a small number of obligatory roles, and

– there are important regularities in the way roles

are expressed in sentences. Specifically, roles are

systematically marked by case-inflexion, prepo-

sitions and/or word order.

According to M. A. K. Halliday, the founder of

functional-systemic linguistics, the rough rules relat-

ing object to representations such as Birds are flying

in the air are: the process is expressed by means of

the verbal group (are flying), the participants by

nominal groups (birds), and circumstances by ad-

verbial groups or prepositional phrases (in the sky).

Halliday 1994 presents a list of a dozen process

types, each characterized by a specific frame of

roles. A more traditional list of roles is given by Ju-

rafsky & Martin 2000. Table 1 shows an adapted

version. I have illustrated the table with authentic

examples from real life.

The notion of roles is rather common in information

systems research. Consider for example the follow-

ing definition of workpractice:

A workpractice means that some actors make

something in favour of some actors, and some-

times against some actors; this acting is initiated

by assignments from some actors, and is per-

formed at some time and place and in some

manner, and is based on material, immaterial and

financial conditions of transactional and infra-

structural character and a workpractice capability

which is established and can continuously be

changed. Goldkuhl & Röstlinger 2006: 53.

Here we meet roles like Agent (‘actor’), Beneficiary

(‘in favor of some actors’), Theme (‘something’),

Time, Place (‘some time and place’) and Manner

(‘in some manner’). Parunak 1995 uses roles as a

tool for specifying agents, the ORM database meth-

odology is based on explicit use of roles (Halpin,

1996, 1998), and Sowa 2000 uses it in his concep-

tual graphs.

In activities as well as and their linguistic repre-

sentations, there are restrictions as to which partici-

pants can fill which roles. For example, abstract

concepts liken sincerity cannot be the grammatical

subject of predicates like “smile” or “have colour”,

nor is the 2nd officer allowed to plan the voyage of

the ship. In fact, the role concept itself implies that

the holder of a role has certain rights and obligations

and must possess certain abilities.

Table 1: A list of thematic roles. Authentic examples from

the maritime domain.

Role Definition and examples

Agent The participant initiating and controlling

an event: Can we berth her without a

tug?

As she goes full speed at shallow water,

then she creates a water wave

Experi-

encer

The participant that senses an event:

Maybe we can see the ‘Gudrun’ from

here.

Theme The participant most directly affected by

an event: Can we berth her without a

tug?

Material The participant that changes identity in

an event: Isn’t that the only place where

we get a copy of those receipts?

Result The final identity of the material: We

make a three sixty (manoeuvre)

Content An event or a state of affairs: I said to

him that as soon as you were finished

steering, you would come down so that

we could get it in

Instrument The mediator of an event: Can we berth

her without a tug?

Addressee The intended recipient of a communica-

tive action: A, have you talked to the

pilot?

Benefici-

ary

The participant that benefits from an

event: And Sir have you received our

pilot chart we have it ready for you here

Comita-

tive

The participant playing a role similar to

the Agent: Is there a chance that he will

come with the helicopter, over there, the

pilot?

Source The start location of the Theme of a

transfer event: I really thought he came

from Rotterdam

Destina-

tion

The end location of the Theme of a trans-

fer event: ‘Gudrun’ must sail before we

can get in.

Purpose The intention of the Agent of the event:

Well, down to about 7.5 meters draught,

you need that in order to run properly

with the top of the tunnel.

Time The time of the event: he will not sail

until two o’clock.

Location The place of the event: we are still lying

here waiting

Manner The manner in which the event is per-

formed: Shall we start turning slowly

now?

In addition, there may be conflicts between these

modalities. According to Ryan 1991, activities in

COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES

narratives must contain conflicts between two or

more of the following modalities: knowledge, inten-

tion, obligation, and desires, and between these and

the actual world, in order for them to be tellable. In

melodramas the hero is for example torn between

obligations and desires.

But conflicts are also important in non-fiction for

diagnosing organisational conflicts: a secretary is

obliged to finish a report at a certain time, but is

unable to do so because the information was not

delivered to her.

In the technical domain, displays in process con-

trol have the important purpose of telling the opera-

tor when components are no longer able to function

in the way they ought to according to their specifica-

tion.

Business processes too can be characterised by

the modalities of ability and obligation. In the work-

practice definition above, assignments refer to obli-

gations and capability to ability.

Business transactions (Haraldson & Lind 2006)

can in fact be analysed as the creation, fulfilment

and removal of obligations: the supplier sends a quo-

tation to the customer and they agree on the transac-

tion, meaning that the supplier is obligated to deliver

the product and the customer obligated to pay the

agreed price. Then the goods are delivered, the cus-

tomer sign for the product, thereby removing the

supplier’s obligations, sends the money, and re-

ceives a receipt that frees the customer from his ob-

ligation to further payments. The Normbase system

in Liu 2000 represents such processes.

To sum up: there is good empirical reason to

posit dynamic bindings between participants and

roles consisting of obligations, rights, desires and

abilities.

2.1.2 The Computational Representation

The basic computational entity (Figure 1) in the sys-

tem is simply one that has a name, an identity, a set

of sensors and actuators, and belongs to one or more

categories. Sensors and actuators are motivated by

our emphasis on context aware technology. The

categories are necessary, since there are restrictions

on who can fill which roles. Entities have two sub-

classes, things and processes. Processes are charac-

teristic in being associated with one or more roles

that bind other entities to the process. A role is rep-

resented by a role-name, plus a binding that contains

a number of dimensions representing the partici-

pant’s modal relations to the activity. A possible

representation is depicted in Figure 1. Note that

processes as well as things can participate in proc-

esses. For example, the participants of the Purpose

and Content roles may be processes, as illustrated in

Table 1. Note that a participant can be bound to

more roles, and that one role may have more partici-

pants bound to it. The methods add and remove par-

ticipant allow us to manipulate participants during

the execution of the activity.

Figure 1: Basic relationships between Things, Processes,

Roles and role Bindings.

2.2 Process Types

Process types are useful when we want to create an

overview of activities in a certain domain.

Bækgård 2006 proposed five general action

types for information systems and in the domain of

process control, Lind 1994 proposed a framework,

Multilevel Flow Modelling, for modelling mass and

energy flows in process plants, and suggested six

basic process types involving mass or energy:

Sources provide it, transports transport it, barriers

block transports, storages store it, balances distribute

it, and sinks consume it.

2.2.1 The Object of the Computational

Representation

But there is evidence that humans in general system-

atically distinguish between a limited number of

process types. In cognitive linguistics these are

called image schemata (Talmy 1988, Johnson 1992:

2).

Table 2: Vendler’s four verb types.

Static Telic Punctual

Activity - - -

Accomplish-

ment

- + -

Achievement - + +

State + - -

Before cognitive linguistics, Zeno Vendler (1957,

1967) classified verb meanings according to the dy-

namic structure of their referents. According to him,

there are four major types of processes, namely Ac-

tivities (play football – Vendler’s activity should not

ICEIS 2007 - International Conference on Enterprise Information Systems

be confused with activity in activity theory), Accom-

plishments (drive to the parking lot), Achievements

(win a race; in computer science we talk about state

changes) and States (sleep, be a plumber). The proc-

esses differ in the terms of three oppositions:

static/dynamic, telic/non-telic, and punctual/non-

punctual (Van Valin & LaPolla, 1997), as shown in

Table 2.

In activities an action is repeated an indefinite

number of times. In accomplishments there is also

action but it stops when a certain limit has been

reached, e.g. when I have arrived at the parking lot.

Achievements denote a momentary state-change,

and state terms denote the continuation of a state of

affairs. Vendler proves the linguistic existence of

these types by showing that grammatical features,

such as ing-forms and adverbials of time and dura-

tion, and, I may add, in Scandinavian languages, the

auxiliary of the past participle, depend upon them.

In fact, the observation is much older; the idea

that language structures processes into a few types,

was described more than a hundred years ago under

the heading of Aktionsart (e.g. Noreen 1903-1923).

Suffixes and prefixes are frequent markers of Ak-

tionsart: -en as in blacken (ingressives), -de as in

decolourize (cessatives), as well as aspectual verbs

like begin, keep, stop (keep running, durative).

Lind (1994) proposes an analysis of basic proc-

esses in industrial process control that is reminiscent

of the old Aktionsart-theory, but was in fact inspired

by Von Wright’s behavior analysis for defining four

basic kinds of actions (Table 3).

2.2.2 The Computational Representation

In the following Vendler’s analysis is used to de-

scribe the dynamic morphology of processes,

whereas Lind’s classification is used to classify the

effect of one action on another. Processes are di-

vided into four subprocesses representing Vendler’s

four classes (Figure 2).

Figure 2: Representation of Vendler’s four process types.

The class of terminatives – accomplishments and

achievements/state-changes – share the feature of

having success criteria that define when the process

has ended with success. Accomplishments and ac-

tivities have other processes as components (Figure

2). All classes can be in an active or suspended state

(Figure 3), and each class is differentiated by the

state-changes it undergoes inside its active state.

Figure 3: The shared structure of all processes.

Accomplishments and states are exemplified in Fig-

ures 4 and 5.

Table 3: Four basic control processes. T = change, d = doing, acting. Condition: what happens if no action is taken.

Condition Explanation Action Explanation Result of action Explanation

pT¬p p exists but vanishes unless maintained d(pTp) p is maintained pTp p remains

(duratives)

¬pT¬p p does not exist and does not happen

unless produced

d(¬pT p) p is produced ¬pTp p happens

(ingressives)

pTp p exists and remains unless destroyed d(pT¬p) p is destroyed pT¬p p vanishes

(cessatives)

¬pTp p does not exist but happens unless sup-

pressed

d(¬pT¬p) p is suppressed ¬pT¬p p remains ab-

sent

COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES

Figure 4: The internal structure of the active state of ac-

complishments.

disabled

executing

not executable

effects

Figure 5: The internal structure of the active state of states.

Both can shift between a disabled state and an executing

state when they receive the messages executable/not ex-

ecutable from their participants (Section 3). For example,

the 10 cm radar announces to the officer that it has be-

come less able to participate in steering the ship; the steer-

ing activity is therefore not optimally executable before

the 10 cm radar is replaced by its 4 cm colleague. Thus,

participants can desert from and become enrolled in the

activity at execution time as part of normal operating pro-

cedures.

The accomplishment is meant to represent a

goal-directed action performed by an Agent. When

the Agent executes an accomplishment, he executes

the processes represented by its components; the

process ends if it receives the messages success or

failure defined by the success criteria. If the process

is in the disabled state, the Agent can try to find

remedies for repairing the defects.

A state is only defined by the effects it has on

other processes, both in its executing and disabled

state. It has no defined ending point, and does not

contain component actions. Examples: a patient may

be conscious which enables a certain set of behav-

iors (talking, eating, etc.), or he may be in a coma,

which disables these behaviors. A ship may be mov-

ing, which enables steering, or it may lie still, which

disables steering.

Vendler’s activity is represented as a state with

component actions added. For example, a running

pump may be driven by a combustion engine that

performs an activity where pistons move back and

forth regularly. But note that we are talking about

representations: a real process may be represented

by a state or an activity, depending upon the aspects

we are focusing on. If we disregard the movements

of the piston and are only interesting in the flow

produced by the pump, we would represent it as a

state, not an activity.

Vendler’s achievement is an accomplishment

with component processes removed. Whereas ac-

complishments have duration – they repeatedly per-

form their component actions – achievements are

momentary.

3 EFFECTS AND FAILURES

3.1 The Object of the Computational

Representation

We now need to specify what failure and effect

means. We propose the following definitions:

An action executes if and only if.

1. All participants are able to fulfil their roles to

some degree.

2. The Agent filler is strongly obligated and/or

strongly desires to fulfil his role.

This means that an action is disabled if one or more

participants are no longer able to fulfil their roles,

and/or the Agent is no longer obligated or no longer

desires to fulfil his role. In accordance with these

definitions, the effect of actions are defined as

3. a change of the participants’ role-bindings to

other actions, possibly to newly generated ac-

tions.

Note the phrase to some degree in (1). It is not re-

quired that all participants perform faultlessly, since

in this case, we would be blind to the parts of reality

that is full of degraded pumps, corroded pipelines,

inattentive operators, and worn tools. This would be

inappropriate if we used the theory to design MIMIC

diagrams for operators of process plants, since a

ICEIS 2007 - International Conference on Enterprise Information Systems

100

main purpose of such diagrams is to alert the opera-

tor of suboptimal equipment.

In the definitions so far, there is nothing hinder-

ing machines in participating as Agents in activities,

quite in line with Actor Network Theory. This does

not imply that humans and non-humans are equally

qualified as Agents in all activities. A clock is per-

fectly suited for being an Agent in the physical proc-

ess The clock ticks (which is also the way language

treats it), but is absolutely unqualified to participate

in the business process the clock sends a quote to its

customer. As ANT rightly claims, agency is a func-

tion relating a participant to an network, not an in-

herent property of a participant.

3.1.1 Effects

Lind’s four process control types can now be rede-

fined as shown in Table 4.

The difference between these material actions and

communicative actions is that material actions influ-

ence the ability of participants, whereas communica-

tive actions influence their desires, rights and obliga-

tions.

Thus, A requesting B to do C presupposes that B

would not have done it by himself, and has the in-

tended effect that B assumes an obligation to do C

(cf. Searle 1994: 65 ff: A attempts to get B to do C);

in other words: the request has the intended effect to

increase the role-binding associating A to the role of

Agent in the action C. It is thus an example of pro-

duce, whereas the act of reminding B to do C is a

case of maintenance. Communicatively destroying

something is exemplified by a moderator removing a

topic from the agenda, and suppressing is to prevent

embarrassing topics from popping up.

Since communicative actions involve abilities,

instrumental actions can influence communicative

ones. For example, a VHF-radio is not able to par-

ticipate in the communication “The captain ordered

the first officer to let the lines go on the VHF”,

unless the captain is close to the VHF. This can be

achieved by the captain taking hold of it. Con-

versely, the first officer is not entitled to participate

in the instrumental action of letting the lines go be-

fore ordered by the captain.

Above I have assumed that machines as well as

humans can fill the Agent role; taken literally this

means that to switch on the ignition is to obligate the

Figure 6: The door-opener as part of an activity system. Dest = destination. Ag = agent. Path = path.

Table 4: Four types of effects.

Process Definition

D maintains P P is executing and D further increases the role-bindings between an able participant and P. Example:

D = the captain turns the wheel, P = the captain steers the course by means of the rudder. The rud-

der’s ability to participate in keeping the course is increased to counteract the effect of wide-wind.

D produces P P is disabled and D increases the role-bindings between a defect participant and P. Example: D = the

chief starts the engine, P = the engine produces rotational energy. The engine’s ability to function as

the Agent in energy production is increased.

D destroys P P is executing and D decreases the role-bindings between an able participant and P. Example: D =

turning off the fuel supply, P = an engine is running by means of fuel. The fuel is disabled to func-

tion as an instrument for the process.

D suppresses P P is disabled and D further decreases the role-bindings between a defect participant and P. Example:

D = a cooler moves water through the walls of a burner, P = the burner melts its walls. The walls’

ability to participate in the melting process is further decreased.

COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES

101

machine or make it strongly desire to run. This is

clearly not a good description of how machines be-

have. Here is a better solution.

The point of departure is that all modal dimen-

sions are organized in a weak and a strong operator.

In classical modal logic, they are called N = neces-

sity and M = Möglichkeit. They are related as shown

in (4).

N P = ¬ M ¬P

(4)

When we describe human actions, we need to distin-

guish between deontic (obligations) and axiological

(desires) variants.

obligated P = ¬ allowed ¬P

(5)

desire P = ¬ inclined ¬P

(6)

The reason is that for humans to be obligated to do

something can be quite different from desiring it, so

two different modalities are clearly necessary here –

an inner and an outer necessity that may conflict.

However, for machines, the two are merged into

one, so here ‘desires’ and ‘is obliged to’ are syn-

onymous with ‘is forced to’ (7). The difference be-

tween human and machine agents is thus that hu-

mans are bound to activities by a variety of modal

ties which coalesce into one in machines.

forced P = ¬able ¬P

(7)

(7) can be found in descriptions of business proc-

esses where a process can trigger another process

(force it to execute) or it can enable it (Rittgen

2006).

In the following, I shall only use necessity

(forced to) and possibility (able to) where machines

are concerned.

The effects of action A on action B can be dia-

grammed as shown in Figure 6 that describes an

automatic door-opener. We draw an arrow from A to

B and annotate it with the modal change <Role

→

Role

, change, dimension>, meaning that if someone

has participated successfully in action A as Role

his likelihood to participate in action B as Role

changed along the specified dimension. For exam-

ple, when a person walks towards the door, the door

participates as Destination in the action; this forces it

to participate as Agent in the opening action. When

it opens, it looses the ability to participate in open-

ing, but is enabled to participate in the action ‘heat

evaporates through the door’ and ‘persons pass

through the door’ in the role of Path.

Note that the actions are half-baked, as is the

case with most actions (Andersen 2004a, 2006). The

#door participant (# symbolizes an instantiated par-

ticipant) is filled in and fixed, but the person partici-

pant is only specified by its category. If a car tried to

participate, it would fail because it is of the wrong

category. The idea is that we seldom construct ac-

tivities from scratch because the environment al-

ready offers us half-baked actions to use.

Note also that the behaviour of the door-opener

is formulated as part of a network of actions involv-

ing humans and machines. The basic construct is an

activity comprising humans and machines, and the

machine is described as a participant in this activity

(see Bødker & Andersen 2005 for maritime exam-

ples of this analysis). This makes it possible to gen-

erate understandable descriptions of its behaviour

since it is described in terms of activities humans

participate in. In the concrete case of the door

opener, understandability presents no problem, but

ease of verbalization and visualization becomes an

advantage if we are faced with complicated danger-

ous machinery, such as nuclear power plants where

particularly abnormal situations present a problem.

3.1.2 A Classification of Modal Changes

Some types of modal changes are very frequent and

represent phenomena known from the literature.

Theme

→

Instrument. Suppose that my lawn-

mower fails to fill the Instrument role in I am mow-

ing the lawn with my lawn-mower. Then my focus

shifts from my lawn to my mower (on focus shifts,

see Bødker 1996), I repair the mower: <Instr

→

Theme>. When the mower is fixed, the reverse

process happens and the instrument’s ability to par-

ticipate in mowing is increased, <Theme

→

Instr

+abil>.

Destination

→

Source. Similarly, if I know

there is a flight connection from Copenhagen to

London, and I myself am located in Stockholm, then

travelling from Stockholm to Copenhagen will en-

able me to later catch the plan to London. In this

case, the Destination of the subordinate action is

identical to the Source of the super-ordinate action,

and participation in the subordinate action increases

the airport’s ability to participate in the super-

ordinate one: <Dest

→

Source +abil>. This modal

change underlies flow-diagrams, such as the ones in

Lind 1994.

Instrument

→

Agent. A third example of role-

shifts is in automatic systems. They often consist of

components with responsibilities that can be de-

scribed by means of roles. A higher-level component

uses a lower-level component as Instrument and

delegates certain tasks to it in which it is forced to

act as the Agent: <Instr

→

Agent +nec>. For exam-

ple, the autopilot maintains the course by means of

the rudder servo system and delegates the task of

maintaining the angle of the rudder to the servo sys-

tem (Bødker & Andersen 2005).

ICEIS 2007 - International Conference on Enterprise Information Systems

102

Addressee

→

Agent. Transfer of control where

A orders B to do C causes B does C to execute can

be described as the role-shift <Addressee

→

Agent

+obl > where the addressee of the first action is ob-

ligated to become the Agent of the second one.

Bækgård 2006 suggests that such control patterns

should be among the basic building blocks in infor-

mation systems.

In addition to modal changes, we also regularly

meet participant changes. For example, Automation

→

Operator, where automation replaces a human

operator in the role af Agent.

3.2 The Computational Representation

Digitally, the schema <Role

→

Role

, change, di-

mension> relating action A and B can be described

by an Effect object that specifies the modal changes

and is associated to the two role objects, Role

and

Role

, that are in turn associated to A and B (cf. Fig-

ure 1).

In principle, all participants get qualified by par-

ticipating in other processes, but what about time

and place? The action I board the train at the plat-

form now requires that the train and I are located at

the platform of course, but the action can still not

execute at 4 pm if the timetable says 5 pm. Time can

be treated just as the other participants, since the

now too participates in a process, namely the passing

of the time, which is measured by the clock sensor.

Only when the clock points to 5 pm, is the now able

to participate in the train departure.

In many activities, participants are only quali-

fied if they are assembled at the same location. This

can be measured by location aware technology.

Thing

-sensors

-actuators

Operator

-displays

-controls

+planning()

+execution()

Automation

Figure 7: Classes representing the human operator and an

automatic system.

The shift Automation

→

Operator can be supported

computationally as shown in Figure 7. We make the

Operator a subclass of Thing that inherits sensors

and actuators. The Operator class is for manual op-

eration, so it adds displays and controls that let hu-

mans use sensors and actuators. Automation is a

subclass of Operator and inherits the displays and

controls, but adds methods for planning and execu-

tion. The switch automation

→

operator can now be

represented as exchanges of instances of the Auto-

mation class and the Operator class. The benefits of

letting Automation inherit the displays and controls

of the Operator class is that the human can see what

the automation is doing, and is able to override the

automation by using the controls. The former allows

the operator to understand what the automatic sys-

tem does (Bødker & Andersen 2005) and the latter is

used in cars and ships with automatic cruise control

to allow manual intervention for safety reasons.

4 SUMMARY

Motivated by new activity-centric uses of IT, I have

drawn a generic sketch of activities and suggested

ways of representing it computationally. An activity

is associated to one or more roles to which partici-

pants are bound by modal ties of various strengths.

Activities can be classified in a few morphological

types. They execute when all participants are able to

fulfil their roles and the Agent is sufficiently obli-

gated or motivated. The effect of executing an activ-

ity is to change the modal ties of its participants to

other activities. During execution participants can

desert and new participants enrol.

REFERENCES

Andersen, P. Bøgh, 2001. Maritime Work and Communi-

cation. Australian Journal of Information Systems,

8(2): 83-102

Andersen, P. Bøgh, 2004. Anticipated Activities in Mari-

time Work, Process Control, and Business Processes.

In K. Liu (ed.), Virtual, Distributed and Flexible Or-

ganizations. Studies in Organizational Semiotics.

Dordrecht: Kluwer Academic Publishers, 35-60.

Andersen, P. Bøgh. 2006. Activity based design. Euro-

pean Journal of Information Systems 15(2), 9–25

Bækgård, L., 2006. Interaction in Information Systems –

beyond human computer interaction. Proceedings of

ALOIS 2006. 257-272. Borås, Sweden: University col-

lege of Borås. http://www.vits.org/?pageId=317.

Bardram, J. E. & C. Bossen, 2005. Mobility Work -- The

Spatial Dimension of Collaboration at a Hospital.

Journal of Computer Supported Cooperative Work,

14(2), 131-160.

Bardram, J. E., 2006. From Desktop Task Management to

Ubiquitous Activity-Based Computing. In Victor Kap-

telinin and Mary Czerwinski (eds.), Integrated Digital

Work Environments: Beyond the Desktop Metaphor.

Cambr., Mass: MIT Press.

Bødker, S. & P. Bøgh Andersen, 2005. Complex Media-

tion. Human-Computer Interaction. (20), 353–402.

Bødker, S., 1996. Understanding computer applications in

use. In: Holmquist et al., (eds), Signs of Work. Berlin:

de Gruyter, 325 -348.

Fillmore, Ch. J., 1968. The case for case. In: E. Bach &

R.T. Harms (eds.), Universals in Linguistic Theory.

COMPUTATIONAL REPRESENTATIONS OF ACTIVITIES

103

London, New York, Sydney, Toronto: Holt, Rinehart

and Winston, 1-90.

Fillmore, Ch. J., 1977. The case for case reopened. In P.

Cole and G. M. Sadock (eds.), Syntax and Semantics:

8. Grammatical Relations. New York: Academic

Press, 59-81.

Goldkuhl, G. & A. Röstlinger, 2006. Context in focus:

transaction and infrastructure in workpractices. Pro-

ceedings of ALOIS 2006, 41-57. University college of

Borås: Borås, Sweden. http://www.vits.org/?

pageId=317.

Halliday, M. A. K., 1994. An introduction to functional

grammar. London: Edward Arnold.

Halpin, T., 1996. Business Rules and Object Role Model-

ing. Database Programming & Design, vol. 9, no. 10,

66-72. Retrieved March 1 2007 from

http://www.orm.net/ overview.html.

Halpin, T., 1998. Object Role Modeling (ORM/NIAM).

In: Bernus, P., Mertins, K. & Schmidt, G. (eds.),

Handbook on Architectures of Information Systems,

Springer. Retrieved March 1 2007 from http://www.

orm.net/overview.html

Johnson, M., 1992. The body in the mind. Chicago and

London: The University of Chicago Press.

Jurafsky, D. & J. H. Martin, 2000. Speech and Language

Processing. New Jersey: Prentice-Hall.

Kristensen, B. B., 2002. Associative Modeling and Pro-

gramming. Proceedings of the 8th International Con-

ference on Object-Oriented Information Systems

(OOIS'2002), Montpellier, France, 2002

Kristensen, B. B., 2003. Associations: abstractions over

collaboration. Proceedings of the 2003 IEEE Interna-

tional Conference on Systems, Man & Cybernetics,

Washington, D.C., USA, 2003.

Lind, M., 1994. Modeling Goals and Functions of Com-

plex Industrial Plants. Journal of Applied Artificial In-

telligence, 8, 259-283

Liu, K., 2000. Semiotics in Information Systems Engineer-

ing. Cambridge: Cambridge University Press.

Moran, T., 2005. Unified Activity Management: Explicitly

Representing Activity in Work-Support Systems.

ECSCW 2005 (European Conference on Computer

Supported Cooperative Work) Workshop on Activity:

From a Theoretical to a Computational Construct.

Paris, September 20, 2005

Noreen, A, 1903-1923. Vårt Språk, vol. 1-7. Lund:

Gleerup.

Parunak, H. Van Dyke, 1995. Case grammar: a linguistic

tool for engineering agent-based systems. Retreived

March 1 2007 from http://www.newvectors.net/staff/

parunakv/casegram.pdf

Rittgen, P. 2006. Towards a language for business action

theory. Proceedings of ALOIS 2006, 3-16. University

college of Borås: Borås, Sweden. http://www.vits.org/

?pageId=317.

Ryan, M-L., 1991. Possible Worlds, Artificial Intelligence

and Narrative Theory. Bloomington & Indianapolis:

Indiana University Press.

Searle, J., 1994.

Speech Acts. Cambridge: Cambridge Uni-

versity Press.

Sowa, J. F., 2000. Knowledge Representation: Logical,

Philosophical, and Computational Foundations. Pa-

cific Grove, CA: Brooks Cole Publishing Co.

Valin, Van R. D. & R. J. Lapolla, 1997. Syntax. Cam-

bridge: Cambridge University Press.

Vendler, Z., 1957. Verbs and times. The Philosophical

Review. 143-160.

Vendler, Z., 1967. Linguistics in Philosophy. Ithaca, N.Y:

Cornell University Press.

Vygotsky, L.S., 1962. Thought and Language, Cambridge,

MA: The MIT Press.

ICEIS 2007 - International Conference on Enterprise Information Systems

104