developed with the aim of providing a good coordination support system that pairs the best computer-based tool a group may have in any situation (dispersed ...
A Prototype of an Integrated Coordination Support System Alessandra Agostini, Giorgio De Michelis, Stefano Patriarca, Renata Tinini Cooperation Technologies Laboratory Dipartimento di Scienze dell'Informazione Università di Milano - ITALIA
Abstract. UTUCS is a system for supporting a group of people (an office, a team, etc.) interconnected through a communication network in handling conversations carried on through different communication media. It has been developed with the aim of providing a good coordination support system that pairs the best computer-based tool a group may have in any situation (dispersed versus non dispersed, synchronous versus non synchronous) with the ability to switch from one to another, maintaining integrated and linked the information it creates. As UTUCS is a general system devoted to integrating conversations independently of the communication media exploited, it has been designed in such a way that it can be enhanced by developing a module for any communication medium that can be effectively supported by a computer network. Up to now the Electronic Mail module, the Face to Face Couple Colloquies module, and the Face to Face Group Meetings module have been implemented.
Key words. Conversation, coordination support system, electronic meeting system, speech act.
1
1. Introduction
In a recent paper in which they give a general framework for an Interdisciplinary Theory of Coordination, Thomas Malone and Kevin Crowston (1991) propose the following definition for coordination: "Coordination is the act of managing interdependences between activities"; and indicate, among others, the following objective for the research in this area: "Designing new technologies for supporting human coordination". We agree that a lot of work remains to be done to design effective coordination support systems. The existing tools and prototypes (e.g. The Coordinator, Information Lens, CHAOS, Contract, Polymer, Visual Calendar), while offering deep insights on some aspects of coordinated processes, do not meet all the requirements an effective coordination support system should satisfy. Let us discuss the above point, in order to make clearer what is still lacking in existing tools. Christine Bullen and John Bennett (1990) analyzed 25 enterprises where coordination tools were used, and made, among others, the following observations: "Message linking is a key improvement provided by electronic communications. (...) Isolated tools hinder productive use of groupware systems. (...) People report most value from tools that parallel their non-electronic activities". We can derive from the above observations some further considerations, generalizing them as follows: i) computer-based tools improve the effectiveness of work processes as they silently integrate and link, messages, messages and related objects, and communication and work environments; ii) it is a good choice to start
2
trying to parallel, with these integrated support systems the way people do their non-electronic activities. These observations can be made more precise if we reflect on some findings Stephen Reder and Roger Schwab (1990) report from the analysis of the temporal structure of cooperative activity within a large American company. They observe that more than 70% of communicative chains (i.e. "sequences of distinct interactions between the same individuals on a given task") with more than two communicative events involve at least one channel switch (e.g. from telephone to face to face meetings, from telephone to fax). The analysis of Reder and Schwab shows that the first observation of Bullen and Bennett can be transformed into the following: "linking communicative events can be a key improvement provided by computer-based tools". So our first consideration can be reformulated as follows: i') computer-based tools improve the effectiveness of work processes as they integrate the communicative chains within activities, silently linking the outcomes of their communicative events (no matter the channel through which it has been performed) together with information created and/or used during them. It is also useful to recall that human communicative behaviour depends significantly upon the medium through which people are interacting. If they are meeting face to face, they make broad use of mimics and indications (even if not Italians!); if their encounter occurs in public (in a group setting) what they say is heavily influenced by the presence of others; if they write an e-mail message or a letter, they choose the words accurately in order to avoid misunderstandings that can slow down the communication. Moreover, on the one hand, people choose the medium they will use from those
3
available on the basis of the interaction they need; on the other hand, they choose what they say on the basis of the medium they are using. It is therefore important that the computer-based communication support system is capable of leaving maximum freedom both in regard to human behaviour with any supported medium and in selecting the best medium. Our second consideration can therefore be reformulated as follows: ii') it is a good choice to start trying to parallel with an integrated coordination support system the way people do their non-electronic activities, in particular the way people behave when interacting through any of their communication media. The important conclusion we can derive is that, in real settings, we generally do not find dispersed and non-dispersed groups; nor do we find groups interacting synchronously and groups interacting asynchronously. Rather, we find groups that are sometimes dispersed and sometimes meet together, as well as groups switching back and forth from synchronous to asynchronous communication. A good coordination support system, therefore, cannot fall into any special category; rather, it must pair the best functionalities that a group may have in any of the above mentioned situations (dispersed versus non dispersed, synchronous versus non synchronous) with the ability to switch from one communication module to another while maintaining the information created and the integration and linking of that information. Taking into account that real human behaviour loses effectiveness whenever it is constrained by organizational rules and/or by tool interfaces, both if these constraints reduce the choice the people have of communication media and if they restrict their behaviour possibilities when using any communication medium, we
4
can conclude that a good coordination support system must pair the ability to maintain integrated and linked information created by a group in its mutual interactions with that of paralleling the way people do their (non-electronic) activities. A careful observation of human behaviour when people use new communication media (as video-teleconference) as well as any long-distance synchronous interaction, can of course indicate new behavioural patterns that have to be supported and enhanced by computer-based tools. As a matter of fact, the modifications computer-based tools provoke on non-electronic interactions have always to be in the sense of broadening the possibilities people have of playing them, if what is gained in efficiency is not to be lost in effectiveness. The UTUCS (User-To-User Communication Support) system is the prototype we have developed with the aim of moving in the direction indicated by the two above conclusions. We do not claim that it meets the above requirements fully, but rather that it opens a new field of possibilities for the designer of CSCW applications. UTUCS has been designed as a component of an office environment, within an Esprit-2 project, ITHACA, where it provides communication services as well as a general exception handling mechanism for a distributed procedure system. UTUCS supports in an integrated way communication through electronic mail and face to face private and group meetings, and it is open to extension to other media. UTUCS, on the one hand, is based on the Language/Action perspective proposed by Terry Winograd and Fernando Flores (Winograd and Flores, 1986; Winograd, 1988); on the other, it represents a new proposal within it. A deep discussion of the Language/Action perspective (Suchman, 1993) is out of
5
the scope of this paper which is devoted to describing a multimedia conversation manager, nevertheless, some comments about our position with respect to it seem necessary in order to explain the design choices we made. CHAOS (De Cindio et al., 1986), the first CSCW prototype developed in Milan was inspired by The Coordinator (Action Technologies Inc., 1988), designed and built by Fernando Flores at ATI. The design of The Coordinator was based on the Language/Action perspective, proposed by Winograd and Flores (1986). We agree with this perspective which states that communication has a pragmatic nature relating human beings in terms of past, present and future actions as well as past, present and future possibilities for action. The mutual commitment of two human beings to an action or to a possibility for action is carried on through the exchange of a sequence of utterances (messages), i.e. through a conversation. In our opinion, it is one of the major achievements of the Language/Action perspective that, ontologically, the basic unity of communication is not the message but the conversation. Human beings act and inter-act immersed in conversations, giving sense to what they do. Utterances within a conversation are related to actions (or to possibilities for action) through their illocutionary point. Speech Act Theory, (Austin, 1962; Searle, 1969), offers some useful categories to exploit the interpretation of the utterances human beings exchange within conversations. On the one hand, in accordance with Winograd and Flores we consider utterances as speech acts (Austin, 1962; Searle, 1969); on the other, we do not think that the illocutionary point of a speech act within a conversation can be univocally defined, in accordance with the well known Searle taxonomy (Searle, 1975). It is common knowledge that
6
human beings do not always agree about the sense of an action, that they interpret actions differently and that they do not always understand each other. The interdependence between the complexity of human communication and human interaction is based on the problems of interpretation that can arise with respect to the illocutionary point of an utterance. If an observer looks at a conversation from the viewpoint of a given commitment, then she can recognize the illocutionary point of each utterance of the conversation (De Michelis et al., 1989). The WinogradFlores model of a conversation for action is ontological, but for each conversation there are as many conversations for action as commitments related to it. Our Wittgensteinian inspiration, see also (De Michelis et al., 1989), lead us to adopt Searle's taxonomy and Flores-Winograd models of conversations to characterize the images of a conversation, without reducing the latter to them. Avoiding the reduction of a conversation to any of its possible images is also the main motivation for not using Gricean Conversational Logic (Grice, 1975): Grice's theory analyzes how a conversation allows two persons to understand each other, under a cooperation principle, assuming, similarly to Searle and, Winograd and Flores, that any conversation has a unique underlying structure of meanings and pragmatic relations. Formalizing a commitment within a conversation entails more than making it explicit: it implies the adoption of a common point of view between the two conversing persons during the whole conversation. Whenever the possibility for acting effectively by two persons depends upon the satisfaction of a commitment between them, then formalizing it can be useful. Otherwise, it constrains their communication without any benefit.
7
It has to be underlined that, if the conversation for action model can become a constraint for human communication (a weak constraint if, as we could see in a real environment, the users of The Coordinator (Action Technologies Inc., 1988) learn very fast to overcome its constraints, whenever they want), allowing the members of a group to design their own conversation model(s) is not a solution. The Winograd-Flores conversation for action model is, in fact, able to make a commitment explicit, while tailorized conversation models are by definition arbitrary behaviour patterns.
UTUCS is a communication support system based on the following design specifications: 1. It is a multimedia conversation manager, supporting up to now email, group and two-person face-to-face meetings. Its architecture allows us to extend the system to any other communication medium. 2. Its conversations follow weakly the Winograd-Flores conversation for action model, offering to its users all the possibilities of conversing freely. Any conversation type (conversation for action, for possibility) can be weakened to become a free conversation with an unlimited number of participants.
While with respect to electronic mail UTUCS provides the communication channel as well as the facility to let the context of the conversation be visible, with respect to face to face meetings UTUCS is only providing services for supporting its users in making visible the context of their interaction. In any of the above cases, the aim is to let the computer-based tool be as invisible as possible, whenever the service
8
provided is not directly useful in the interaction. For group meetings, we have developed a system for supporting a meeting coordinator i.e. facilitator, (more information on meeting support systems is given in Section 3.3) based on the ideas of John Whiteside and Dennis Wixon (1988). The services provided by UTUCS are manifold. It supports calling a meeting (e.g. sending invitations, preparing the agenda, etc.), running a meeting (e.g. displaying the minutes of previous meetings and other useful information, composing and displaying the minutes of the current meeting etc.). It assists the participants in maintaining the context of the meeting for future interactions, by updating their information basis with the minutes. The information base of any UTUCS user is composed of three interlinked parts: the first one is a record of all conversations, while the other two record all private and group meetings. The paper is organized as follows: section 2 introduces the conversation types of UTUCS; section 3 briefly describes the architecture of UTUCS and its main modules; section 4 shows an example of a working session with UTUCS; section 5 indicates the development lines of the research around UTUCS and finally an appendix illustrates the implementation issues.
2. The conversations
As we have mentioned above, UTUCS is based upon the Language/Action perspective, stressing the communication process, or more precisely its pragmatic
9
dimension, as the primary aspect of office work. Utterances from one person to another are grouped into interaction sequences, in such a way that all the subsequent 'replies' are linked to the related sentences. Together with Winograd and Flores (1986), we call 'conversations' these interaction sequences and not the single utterances, and regard the conversation as the atomic element of communication. Within a conversation the focus is therefore on the mutual pragmatical relations the participants are establishing. These relationships -the commitments a partner is making with the other to do something or to be someone- are considered the conversation topics. Since 'real' conversations possibly involve various topics, to avoid any confusion we abstract within conversations all those about a particular single topic. This is done without loss of generality, as we aim to support the commitment negotiated within a conversation and not the conversation itself. As a matter of fact, we are not creating a communication medium, but only a communication facility, and only the electronic mail module integrates both medium and tool. UTUCS helps its users to augment the effectiveness of their conversations without disciplining the way they communicate. UTUCS therefore supports communication depending on the medium the users are exploiting.
In UTUCS each conversation involves the person who opens the conversation, 'actor', the person with whom the conversation is going on, 'partner', and, optionally, a set of people the actor can designate to have visibility on the conversation, 'carbon copy'. In the following moreover, we will nominate 'active' the person directly involved in the accomplishment of the commitment, and
10
'passive' the other one. The conversations, classified according to topics, are subdivided into four types:
• 'commitment to do' conversations: in order to perform an action; • 'commitment to be' conversations: in order to modify role distribution; • 'information providing' conversations: either to give or to require information; • 'information handling' conversations: either to delegate to someone or assume directly the task of managing information.
We can immediately see that while 'information providing' conversations are subcases of 'commitment to do' conversations, 'information handling' conversations are sub-cases of 'commitment to be' conversations. They have been distinguished because both of them constitute classes of conversations that can often happen within office work. The proposed typologies, therefore, offer only an example of a classification of conversations: different choices may be useful in specific contexts.
Since UTUCS focuses its attention on the unique topic of any conversation, each conversation is characterized by the possible states of its topic and by the possible transitions that can be performed (i.e. the speech acts exchanged among the partners) from one state to reach another. Formally, according to Winograd (1988) and De Michelis (1989) we can describe each kind of conversation by the automaton:
11
M = where:
S
is a finite and not empty set of states;
I
is a finite set of transitions;
Ω
is the function that receives as inputs the current state and a transition (speech act), and calculates as output the next state;
F⊆S is the not empty set of terminal states.
2.1. 'COMMITMENT TO DO' CONVERSATIONS
The 'commitment to do' conversations represent the main structure joining the cooperative work; they deal with all communications where two persons speak to negotiate the terms of a commitment to do an action that one must accomplish for the other. This kind of conversation can be initiated by a person making a request to a partner for doing an action or making an offer to her. The set of all possible states for this conversation (Sc-t-d) is:
S c-t-d = {Requested, Offered, Taken, Deleted, Rejected, Concluded, Satisfied, Not Satisfied}
where the set of the terminal states (Fc-t-d) is:
12
F c-t-d = {Satisfied, Not Satisfied, Deleted, Rejected}
In figure 1 the state-transition diagram representing a 'commitment to do' conversation is depicted. In each state the possible speech acts are represented by the oriented arcs. In the figure only the transitions that change the state of the conversation are shown; the dashed arcs are the possible active's actions and the remaining ones the passive's.
If a person (Passive, P) makes a request to another one (Active, A): for example "Could you test my program by the end of the week?", then the commitment under discussion reaches the "Requested", from which it can reach other possible states, depending on the response of A.
13
Requested
Offered
Deleted
Taken
Rejected
Concluded
Satisfied
Not Satisfied Transition of Passive Transition of Active
Figure 1
If A agrees to perform the action on the proposed terms, e.g. within that deadline, the state becomes "Taken"; otherwise "Rejected", if A refuses to undertake the commitment. In the case in which A is willing to test the program, but she would like to negotiate the conditions of the action to carry out -for example she wants to postpone the time of accomplishment- the conversation reaches the "Offered" state. It might happen that A has already done the requested action, then her response will lead inevitably to the state "Concluded". Also P can determine a transformation of the commitment to the state "Deleted", if
14
she changes her opinion for some reason after the request and withdraws. In the state "Concluded" the Passive evaluates whether the commitment has been brought successfully to a conclusion, and then goes to either of the alternative terminal states "Satisfied" or "Not Satisfied". The state-transition diagram in figure 1 also contains all the information regarding a conversation initiated by A making an offer to P to do something. The diagram in figure 1, reproduces the Winograd Flores model with minor changes. In fact there is a different approach to handling the anticipated conclusion of a conversation; although the Winograd Flores automaton and our diagram both have three terminal states managing the non-satisfaction cases, they are slightly different. Moreover the conversation starting with an offer and the one starting with a request are unified in our model. In table I a summing up of speech acts that modify the state of commitment is shown.
15
Present State
Speech Act
Next State
()
A P
Offers Requests
Offered Requested
Requested
A A A A P
Counter-Offers Refuses Accepts Concludes Withdraws
Offered Rejected* Taken Concluded Deleted*
Offered
P P P A
Counter-Requests Refuses Accepts Withdraws
Requested Rejected* Taken Deleted*
Taken
P P A A
Counter-Requests Cancels Counter-Offers Concludes
Requested Deleted* Offered Concluded
Concluded
P P P
Refuses Evaluates Evaluates
Taken Satisfied* Not Satisfied*
* Terminal States; A = Active, P = Passive
Table I At any time both active and passive are able to send notes or solicitations to the partner; for example, if A delays in answering P, the latter is likely to press her for a quick response. This kind of speech acts does not cause any changes in the actual state of the conversation. People belonging to the carbon copy, namely those who have visibility on the conversation, can send comments or notes to the two interlocutors, and these messages do not change the state of the commitment.
16
2.2. 'COMMITMENT TO BE' CONVERSATIONS
This category consists of conversations for possibility. Their effect is to transform the organizational setting, i.e. they modify role distribution within a group and open new spaces of possibility for its members. The set of reachable states for this conversation (Sc-t-b) is:
S c-t-b = {Requested, Offered, Accepted, Rejected, Deleted}
where the set of the terminal states (Fc-t-b) is:
F c-t-b = {Accepted, Rejected, Deleted}
Using the same conventions as before, figure 2 shows the state-transition diagram describing this class of conversations.
17
Requested
Offered
Rejected
Deleted Accepted Transition of Passive Transition of Active
Figure 2
A 'commitment to be' conversation can be started either by the Active or by the Passive with an offer or a request (to assume a role, to open a new activity field, and so on). For example, if A offers to fill a position herself, saying, for example, "May I assume the direction of The Cooperation Technologies area?", the conversation reaches the state "Offered". If P accepts that A will fill that position, the commitment becomes "Accepted"; meanwhile, if P has some perplexities and intends to change the terms of commitment, she is able to counter-request (going to the "Requested" state). Let us suppose that it is P who makes a request to A, such as "Would you like to be the head of The Cooperation Technologies Lab?", getting the commitment to be "Requested". Now various choices are at A's disposal: she can answer with a drastic refusal
18
("Rejected"), accept unconditionally ("Accepted"), or lastly make a counter-offer ("Offered"). It is also possible that the actor closes the conversation by withdrawing the commitment and causing it to enter the "Deleted" state. The development of this kind of conversation extends Winograd Flores typologies. In table II the complete list of possible state transitions is given.
Present State
Speech Act
Next State
()
A P
Offers Requests
Offered* Requested*
Requested
A A A P
Negotiates Accepts Refuses Deletes
Offered* Accepted Rejected Deleted
Offered
P P P A
Negotiates Accepts Refuses Deletes
Requested* Accepted Rejected Deleted
* Initiating states; the other ones are all terminals
Table II
Even in this kind of conversation, at any state it is possible to send notes, solicitations or comments without changing the actual state of the conversation. These types of speech acts are not represented in the diagram of figure 2 or in table II, not because they are less important but because they are non-essential to the diagram. It has to be remarked, moreover, that 'commitment to be' conversation in UTUCS
19
allow users to also perform a teleconference, extending the services offered by The Coordinator. A person can in fact open a conversation with a declaration such as "I declare open the teleconference on ..." and the carbon copy list is filled in with the participants in the teleconference. In the current version of UTUCS some restrictions are imposed on the user. In fact, the user must select a person with whom to open the conversation, and it is not possible to modify the list of participants in the conversation after it has been initiated. These drawbacks will be solved in the next release of UTUCS.
2.3. 'INFORMATION HANDLING' AND 'INFORMATION PROVIDING' CONVERSATIONS
Even if the Language/Action theory places emphasis on the concept of a language as an activity and not only as a transfer of information or an expression of thought, we do not forget that within organizations people also communicate for the purpose of exchanging information. Indeed, 'commitment to do' and 'to be' conversations allow participants to manage information concerning the topic under discussion, both by inserting it in ordinary messages and by sending specific notes. In addition to that, an office member is able to open a conversation to delegate the handling or managing an information basis to her partner ('information handling' conversations) or, more simply, for giving to or requiring from her some information ('information providing' conversations).
20
The state-transition diagrams modelling conversations about information have a structure hardly dependent on the organizational context that we want to model. This implies that up until now the design of these automata has only been sketched. Let us underline that this limited possibility of tailorizing conversation types to specific organizations does not allow users to create any conversation pattern - as e.g. in Information Lens, (Malone et al., 1986). Any new conversation type must be a specialization of one of the two basic conversation types. As we have said above, when conversing, human beings do not associate an utterance with a unique topic; rather, they usually conduct multi-subject conversations. UTUCS has been conceived in such a way that it preserves to a high degree, the natural quality of human communication, and does not impose any unnecessary discipline on the communicative behaviour of its users. Only the Electronic Mail module (see section 3.1), which is both a communication medium and a communication support system, reproduces the single topic conversation pattern shown above: the constraints it imposes are anyhow syntactical and can be overcome without difficulties. All the other modules (see below) allow the users to speak with the greatest freedom and help them to fix the state reached by any topic of conversation at the end of the communication session.
3. The UTUCS overall architecture
UTUCS is a system for supporting a group of people (an office, a team, etc.)
21
interconnected through a communication network in handling conversations carried on through different communication media. As UTUCS is a general system devoted to integrating conversations independently of the communication media exploited, it has been designed in such a way that it can be enhanced by developing a module for any communication medium that can be effectively supported by a computer network. Up to now the Electronic Mail module, the Face to Face Couple Colloquies module, and the Face to Face Group Meetings module have been implemented as a platform to which other bricks can be added. The next enrichment will be a module for supporting Telephone Colloquies; afterwards, new communication devices such as Fax, Video-Teleconference and so on will become part of the system. Figure 3 depicts the global architecture of UTUCS.
Telephone
User Interface
Face to Face Couple Colloquies
Electronic Mail
Face to Face Group Meetings
Message Switching System
Mail Box
Talks
CRRs
Information Base
Figure 3
22
Minutes
By means of the 'User Interface', users have access to the above-mentioned modules. The 'Information Base' contains data structures which can be reached by every module. The Current Relation Record (CRR) is the central nucleus, a file, shared by all the modules, where the information about the objects and the related conversations are stored. While Messages exchanged via electronic mail are only stored in the CRR, the Minutes of group meetings and the Talks produced in private colloquies are also recorded in specific files. In these cases, in fact, the tool is not even the medium of communication, so additional information is necessary in order to reconstruct the communication context. Dynamic links between CRR, Minutes and Talks allow the user to switch from one module to another, in an easy and natural way, using data as gates between different environments. This means that at any moment during a conversation, no matter which medium is used, we may have the whole conversation at hand in the CRR and, while glancing through our diary, we may always have the possibility of taking up a conversation. Besides, while looking at a particular step of a conversation, the user is able to review the entire minutes of a meeting or of a private colloquy to which the step belongs. Because of the high number of active conversations and for privacy, the Information Base is distributed and replicated. This means that each user of the UTUCS system has at her own workstation a personal copy of the conversations in
23
which she is involved, as a participant or as a member of the carbon copy list. The 'Message Switching System' is in charge of all the physical exchange of Messages, Minutes and Talks. Mailboxes are used to temporarily store messages before they are read or actually dispatched.
3.1. THE ELECTRONIC MAIL MODULE
The electronic mail module allows the user to start and resume conversations in an asynchronous way; it supports writing, reading, dispatching and storing semistructured messages. When the user wants to compose a message, which belongs to either a new conversation or an already existing conversation, the system makes available, in a pull-down menu, only the speech acts the user can perform on the ground of the present state of the conversation. Each user has two electronic mailboxes containing, respectively, the incoming messages (before they have been read) and the outgoing ones (before their actual dispatching in the network). As mentioned above, users have their private copy of CRRs of all conversations in which they are involved, and only Active and Passive can change their states; consequently, there are two relevant copies of a CRR related to a conversation to be kept mutually consistent. Consistency problems can arise because of the asynchronism of electronic mail communication. In fact, after a message is sent, it moves through a communication channel, then it is stored in the mailbox at the
24
receiver's workstation. The receiver can read it, maybe after some days, and during this period the two users get a different image of the conversation to which that message belongs. In order to solve this problem, the system has been equipped with an 'acknowledge mechanism'. At the sending of a message the state of the related commitment is flagged, and from now on the sender can only perform a limited set of speech acts (soliciting, commenting, closing, ...) as part of that conversation, until the message has been read by the partner. When the message is received, that is, when it arrives at the addressee's workstation, it is put in the private user's mailbox, and she cannot perform speech acts in the conversation before she has read it. To sum up, the system constrains speech acts related to conversations with pending messages. UTUCS signals the reading of the message to the other partner's system by sending a particular kind of message called 'acknowledgement', which is transparent to the interlocutors. This conversational protocol can appear too rigid, but it is too lazy as well in that it is insufficient in preventing the users from doing some actions that could generate inconsistencies in the Information Bases. If actors A and P send each other 'contemporaneously' a message within the same conversation (contemporaneity has in UTUCS a duration in time, as messages run on the communication network in non-zero time), then the acknowledgment mechanism is unable to avoid an inconsistency. This mechanism is therefore
25
integrated by a 'consistency mechanism', allowing us to solve the impasse as we will see in the next paragraph.
Users not directly involved in a conversation, but having it visibile, can only send messages that do not modify the state of the commitment; so in sending and receiving this kind of message there is no need for a synchronization mechanism. UTUCS does not allow any user to change her role (Active, Passive, Carbon Copy) within an ongoing conversation: if necessary, a new conversation must be opened.
User A Process
Monitor A
Monitor B
_mss_rcv
_mss_snd
_mss_snd
#:in-mailbox #:out-mailbox
IN
System Mailbox
Virtual Transmission
OUT
User B Process
_mss_rcv
#:out-mailbox #:in-mailbox
IN
OUT
Figure 4
Figure 4 shows the architecture of the electronic mail module as it was developed for the first release of the system: a virtual transmission between users' mailboxes corresponds to a physical transmission implemented by means of the MSS primitives. We have preferred to keep the concept of a UTUCS user distinct from
26
that of a MSS user, even though the system is able to switch from the MSS identifier to the UTUCS identifier of each user and vice versa, in a univocally determined way. That gives us the possibility of easily integrating UTUCS with any other message handling systems.
3.1.1. The mechanism of recovering from inconsistencies
As we said in the previous paragraph, to ensure the prevention of inconsistency, it is necessary to define a behaviour protocol for the user that keeps her from modifying her copy of the CRR without modifying the interlocutor's copy too. The introduction of the acknowledge mechanism avoids inconsistencies in the time slice between the instant in which the sender dispatches her own message to the moment in which she receives the related acknowledgment, as it prevents her from sending a new message. If the messages of two interlocutors are sent 'in the same moment', the consistency mechanism is activated restoring the consistency between the two Information Bases. This latter mechanism, as well as the former one, is based on the states -empty or full- of the two mailboxes of the users involved. The mailboxes of a user can be in any of the following states:
a) 0,0
The mailboxes are empty, the user can do what she wants.
b) 0,1
The user has sent a message to the partner, but she has not yet received
27
the acknowledge indicating that this message has been read. The user cannot send any message (because of the acknowledge mechanism). c) 1,0
The user has received a message but she has not yet read it. The user cannot send a new message before reading the received message (because of the acknowledge mechanism).
d) 1,1
An inconsistency has occurred. The consistency mechanism restores a consistent situation.
The transitions in the graph of figure 5 describe the possible moves that the user (Elabelled transitions) and/or the system (D-labelled transitions) can do in any of the mailboxes' states.
{0 0} E1
E2 E3
D1
{1 0}
{0 1} D3
D2 E3 {1 1}
Figure 5
- E1
Sending a message.
- E2
Reading a message.
- E3
Receiving a message.
28
- D1 Receiving an acknowledge. - D2 Cancelling the received message. - D3 Cancelling the sent message.
A function is defined that calculates the present state of mailboxes on the grounds of the move made. The messages are automatically processed upon their arrival at the user's workstation: immediately if the user is connected to the system, at the login otherwise. Since this is the moment in which an inconsistency can occur, the consistency mechanism checks if any of the CRR mailboxes are in the (1,1) state. If this is the case, then the consistency mechanism restores normality by eliminating one of the two conflicting messages, either the received (D2) or the sent (D3) message, informing its author. The criteria for managing messages -in their application order- are that messages changing the state are preferred to those not changing the state, and that among messages changing the state those not closing the conversation are preferred.
3.2. THE FACE TO FACE COUPLE COLLOQUIES MODULE
In carrying out usual working activities, people spend much of their time meeting other people either in their offices or in some specifically designated meeting room. Whereas electronic systems designed to support more effective group meetings
29
already represent a growing trend, tools that facilitate drafting reports concerning colloquies between only two people are nearly non-existent. The main motivation for this lack is that this kind of communication does not seem to need any specific tool. In fact, the presence of only two persons avoids problems of floor management as well as context sharing and awareness problems. Nevertheless, UTUCS is interested in handling face to face colloquies in order to support the whole conversation to which those colloquies may belong, integrating their outcomes in the related CRRs. Therefore the design has been inspired by the following principle: introduce as little stiffness as possible into the way people converse. We look for effectiveness with respect to a couple of interlocutors, both involved actively in the positive outcome of the colloquy, disregarding the case when they seek to avoid direct commitment about the topics of the conversations. What they need in this case is that both make a consistent recording of undertaken commitments and agreements. The efficacy of the colloquy seems to be negatively influenced mainly by bad retranscriptions, different interpretations and forgetfulness. The overcoming of these mistakes looks very easy; in fact a computer tool that allows us to transcribe immediately and by common consent the results of the discussion, by updating the interlocutors' datebook with these data, is sufficient in dissolving the problem. Precisely through this UTUCS module, the user is able to record in a formalized way the decisions taken during private colloquies with another person. For privacy reasons a work session can only be opened by both interlocutors using their passwords. During the colloquy no restriction is imposed on users at the level
30
of the transitions of a commitment, thereby avoiding the reproduction of the whole conversation carried out in a verbal way. So they can both open new conversations in non-initial states and, for commitments formerly established, move immediately in to states that can be reached through more linguistic steps. For example, the users can store any new 'commitment to do' conversations in the "Taken" state or, for an already existing conversation in a "Taken" state, they can jump directly to a "Satisfied" state. By means of purpose built menus, the two partners are able to select from the Information Base any topic discussed in an already open conversation and continue the latter, maintaining some previous information and avoiding filling in again all the fields. Among the many available functionalities of this module, the opportunity given to the users at any time to retrieve all the steps of the already open conversations, no matter through which communicative medium they were done, must be stressed. At the closure of the work session, beside updating the CRRs of both the interlocutors and the observers -people in the carbon copy- the information of the colloquy is stored in a specifically designated file within the knowledge base (as mentioned before). This makes subsequent retrieval of the terms of the discussion more flexible; in fact the user will be able to recover conversations about which she cannot remember anything except the occasion during which they were updated.
This new way of handling the private colloquies is in our opinion likely to reduce users' resistance with respect to using UTUCS, as it does not condition the behaviour of the users in a forceful way.
31
At a first glance, taking notes with pen and paper, relying only upon one's own fallible memory to recover the context of the interaction, could be regarded as a non onerous waste of time. By all means it is less effective than a mutually consented rewriting of the commitments undertaken within a colloquy using an electronic tool that automatically updates the personal datebook.
3.3. THE FACE TO FACE GROUP MEETINGS MODULE
Meetings play a fundamental role in any socially organized group. Coordination, collaboration and co-decision cannot leave a mutual exchange of ideas out of consideration. Despite the growing attitude in favour of long-distance meetings, it seems that people are not willing to give up simultaneous physical co-presence; after all, face to face context is irreplaceable, as the following quote from (Whiteside and Wixon, 1988) stresses: "Body language, eye-contact, physical contact, ... speaks to something very deep in the human condition that electronic communication will never replace. This physical being-there is essential to the building of consensus and trust in groups." But if a team is not located in the same physical space the barriers of physical separation must be overcome and in this case very interesting research and systems are available. Between the computer-controlled teleconferencing or videoconferencing system developed, we want to emphasize the one denominated "media spaces".
32
Heath and Luff (1991) argue that media spaces may provide a foundation for new forms of sociality in the workplace that have characteristics very different from those which occur in shared physical space. Yet there is an advantage, in that people may be able to achieve a new sense of copresence which is not as obtrusive as sharing a physical office. Every system that enhances the meetings of a number of people is equally important in cooperative work; in our case the starting point was a module supporting people physically present in the same room, favoring computer tools with low technological costs, even if we want to underline that the UTUCS architecture is open to any medium. Also, in designing the meeting module of UTUCS, we took the viewpoint of the Language/Action perspective, finding a great inspiration in the approach that Whiteside and Wixon (1988) call the contextualist paradigm. The fundamental concept of contextualism is the uniqueness of any event occurring in an unrepeatable context that gives it meaning. People are always in a context and they can be understood only by taking into account their involvement in the world. That implies a methodology of analysis carried out in the workplace that takes care of routine procedures applied by users and of strategies aimed at solving breakdowns. The researcher becomes part of the scenario she is studying, and the system cannot be considered independently of the observer. The analysis phase must always be followed by a first implementation of a system which is an "experimental platform" or prototype for eliciting the actual needs of the users. This concept is well explained in the following sentences (Olson and Olson, 1991):
33
"...In all of these cases, the features of these systems are based on the designer's intuitions rather than on a user-centered observational base. This has helped to surface many issues about what functions might be useful and how to present them to users, but iterative design with users will be needed to sort out what is useful..." And therefore: "...Another aspect of our research is to build an experimental platform on which we can begin the process of iteratively designing a shared workspace that might help our collaborators..." In fact, by using the system described in this paper other projects have been developed with all the functionalities that users appreciated in this first implementation and some further issues resulting from user requests (De Michelis and Grasso, 1993; Agostini et al., 1993).
In line with Marca's (1989) considerations, we think it is necessary to have a 'facilitator' of the meeting. The facilitator was made popular by the Japanese quality circles methodology. This role is usually held by a person in charge of assisting the group in the fulfilment of its duties, and she must run the meeting to make it more productive. The weight of the facilitator's role within the groups (Viller, 1991) has at this point been ascertained. Providing an electronic meeting support system without specifying tools to help her in her duties, will limit the group's efficiency; on the other hand, if her activities are wholly automatized by the machine, this system would lose flexibility. It is our opinion therefore that the tool must not interfere with the natural evolution of the meeting; rather, it has to help the facilitator
34
without forcing her in any way.
Before the meeting, UTUCS helps the facilitator to prepare the environment, making it easier to provide the necessary information that will be used in the discussion: for example, retrieving the history of previous meetings. During the meeting she takes notes of the on-going conversations by filling in a form displayed on her workstation. She can visualize it on a projection device which is useful in guiding and focusing the meeting discussion. As information does not exist outside interpretative structures, what is said in a meeting is not autonomous with respect to the utterer. Besides, there is an intrinsic duality between what is said and what is listened. Because listening means to interpret (in the hermeneutic sense) from a particular point of view, there is no 'objective' report. The minutes generated on line, avoid a number of the problems that arise when either everyone takes her own notes, or a single person takes minutes without much interaction and control on the part of the other participants. It is of great importance that users be aware of what is going on during the meeting. This awareness is reached thanks to the sharing of the events context, accurately explained by the formula WYSIWIS (What You See Is What I See) (Stefik, 1988). During the meeting the facilitator can shift from one conversation about a commitment to another in a very easy way, so she can reproduce the flow of the discussion without difficulties. Like in the face to face couple colloquies, it is not necessary that the commitment runs through all the states and transitions of the conversation; but, as an example, it
35
can be created at once in the "Taken" state. At the closure of the debate all the commitments will be automatically recorded in the CRRs of the participants and, if that is the case, of all people indicated in the carbon copy. The final version of the minutes is stored in a way that are available at the next meeting in order to recreate the context of the previous ones and have available the history of the meetings about a project. Thus far, many Electronic Meeting Systems have supported meetings as independent, autonomous events. While improving meeting outcomes is important, it is also important to capture the additions to organizational memory and to provide access to them in subsequent meeting(s) (Nunamaker et al., 1991) or in any other communicative event.
36
4. Example of a working session
Taking inspiration from the internal use of UTUCS within the Cooperation Technologies Laboratory, let us sketch a plausible scenario, in which UTUCS supports, in a very simple and natural way, the negotiation of commitments through different media of communication. In our Department there is a Committee for Seminars in charge of planning and organizing seminars about specific topics, settled each time by the committee members. The coordinator of this committee, in this case Giorgio De Michelis, meets another member, Renata Tinini, in a face to face colloquy and from the discussion the idea for a spring cycle of lectures about new suggestions in the CSCW field comes up. They decide to call a meeting in which the whole committee will be involved in this project. During their conversation the two interlocutors select the face to face couple colloquies module (see Fig. 6) to record what has been discussed and decided and to bring up to date the respective CRRs with the new commitments.
Figure 6 In particular, Renata undertakes calling the meeting (for December 12th) and organizing it, managing the speakers' accommodations and looking for rooms available in the next spring; meantime, Giorgio devotes himself to finding
37
sponsorships. A form for each commitment is filled in; in figures 7 and 8 two forms are shown, the first one involving Renata as active and Giorgio as passive, and the latter vice versa. To get the other committee members acquainted with this decision they are inserted in the carbon copy, so that they have visibility over the future history of the conversation.
Figure 7
Figure 8
Besides being the organizer of the meeting, Renata also engages herself as the facilitator. Afterwards, she sends an e-mail message to all the committee members, asking for comments and suggestions to be delivered before the meeting about topics and possible invited speakers. In the preparation phase she has available a number of functionalities that make arranging the environment easier: for instance, she is able to retrieve information from previous meetings by looking at the minutes in the agenda. In this case the
38
meeting does not become part of an already begun history, since it is the first one starting a new series; so the facilitator cannot take advantage of pre-existent meetings and she chooses the 'New' option in the 'Prepare a meeting' menu (see Fig. 9). Otherwise, if the current meeting represents the last link of a chain, the 'Continue' option would be chosen.
Figure 9
She draws up the lists of program items and of participants, and inserts in the carbon copy field the names of committee members unable to attend the meeting. In our case Elsa Bignoli, who is at present at Urbana University is entitled to participate, will receive the minutes all the same. From this template the facilitator may also send invitations to the attendees. When the meeting is convened, the facilitator -the only one who has a workstation-
39
selects the right meeting among those previously prepared. If some of the invited members are missing, their names are shifted to the carbon copy. The form to fill in, displayed also in the on line projection device, consists of two parts. The upper part contains general information about the meeting and is a landmark all along the reunion; in the lower one commitments mutually relating couples of participants are recorded. Figure 10 depicts the fact that Alessandra, as far as time scheduling is concerned, submits a draft of the lectures calendar, depending on speakers and rooms availability. Various optional fields allow us to characterize both the meeting and the conversations ongoing within it, depending upon user needs. This is the result of a negotiation between Alessandra and Renata: it started with an e-mail in which Renata requested that Alessandra prepare a draft of the calendar on the ground of both speakers and logistic requirements. Alessandra accepted and promised to bring the draft to the meeting; at this point the commitment reaches the state "Concluded", as figure 10 shows.
40
Figure 10
When the facilitator goes through all suggestions she received in reply to her e-mail, she notices that Elsa has been able to secure the presence of an important unexpected speaker. So, since there are some changes to be made in the draft calendar, the information sent by Renata to Alessandra is checked. In fact, from each report template the retrieval of the CRR of the participants related to the topic under discussion is allowed. In this case the previous step in the history, that is, Renata's mail, is made known to people present at the meeting (see Figure 11). This permits the context not to be missed, and creates links among the various
41
building bricks of the conversation. As the calendar must be rearranged, the commitment Alessandra took comes back to the "Taken" state.
Figure 11
At the closure of the debate all the minutes will be automatically sent to the attendees and to whom it may concern.
42
5. Conclusions In developing UTUCS we had three different objectives: 1) to propose a radical shift from systems supporting users on a single communication medium to systems supporting them on any communication media; 2) to design communication support systems going beyond message rooting, so as to make the context of the communication available to its actors; 3) to build a conversation handler within the Language/Action perspective, that on the one hand originally exploits a pragmatical interpretation of human communication and on the other enriches the communication possibilities of its users.
After the development of the UTUCS prototype as described above, we carried on the integration of a new release of it, supporting only electronic mail conversations, within the office support system WooRKS, and developed an experimental application of it within a Bank, see (De Michelis and Grasso, 1993; Agostini et al., 1993). The experience we had developing UTUCS has strongly encouraged us to feel that we are moving in the right direction in attaining the three objectives above. It has also shown us new steps, insipiring the themes of our current research, which we briefly describe in the following: 1) Our point of view with respect to human communication, to its pragmatic dimension, and to the whole debate on the Language/Action perspective, has been deeply influenced by the design of UTUCS and by the discussions we have had about it with numerous colleagues (e.g. with some anonymous referees of this
43
paper), inspiring us to develop a new conversation model that tries to overcome some of the limitations of the classic Winograd-Flores model without exiting from the Language/Action perspective. The development of a conversation handler based on it is ongoing and will be the most relevant follow up of UTUCS. 2) We are also studying the complete variety of communication channels a person can use within her (computer supported) working environment. The aim is to extend the ideas on which UTUCS is grounded in order to specify a fully general communication support system able to bear conversations carried on through any communication channel and to allow switching between two of them with the minimal structure necessary in order to integrate the knowledge generated through a conversation. We therefore need to review all the existing proposals for supporting communication through any type of channel, and to conceive a semi-structure (Malone, 1987) for the support of a conversation through different channels in such a way that it preserves integration and parallels at utmost non electronic communicative behaviour (Bullen and Bennett, 1990).
Appendix. Implementation issues
The UTUCS prototype was developed on workstations Sun 3/60 (UNIX Operating System) using a functional and object-oriented language denominated LeLisp (Chailloux, 1984). The 'User Interface' is implemented in FORMS (Texier, 1985), a graphic package that allows us to build and manage windows, forms and
44
objects like buttons, menus, input fields and so on. The 'Information Base' of Current Relation Records, Talks and Minutes is implemented as lists of objects, taking advantage of the object-oriented features provided by Le-Lisp. Each conversation, no matter how it is supported, updates the CRR, which is characterized by the following structure: Active: the partner that performs the action involved by the commitment; Passive: the other interlocutor; State: the current state of the commitment; Subject: a description of the commitment given by the actor at the opening of the conversation; Category Set: a list of meaningful classes, defined by the user to group her conversations; Deadline: the time in which the commitment must be accomplished; Carbon Copy: the users who have visibility on the conversation, that is people who can read and comment but not affect the conversation state; Conversation: all the conversation steps, 'Message', 'Talk' or 'Minutes', making up the historical context of the discussion; Support: the module of UTUCS through which the last conversation move has occurred; System State: the state of mailboxes of the user; useful in accomplishing, both the acknowledge and the consistency mechanisms.
The minutes of meetings which either the user attended or is involved in bring up to
45
date the 'Minutes' file. It has the following definition: Number: the progressive number characterizing a meeting within a sequence of related meetings, automatically updated by the system; Name: the meeting identifier; Date: when the meeting happens; Place: where the meeting is held; Program: the complete list of items on the agenda; Topic: a synthetic description of subjects under discussion; Key Words: the relevant words; Participants: the complete list of people attending the meeting; Distribution List: the list of users who must be informed; Facilitator: the person who runs the meeting; Minutes: the "minutes" of the meeting or, better still, the list of commitments taken by couples of participants; State: the hidden field holding the current state of the meeting, that may be "new", "open" and "closed". Let us talk a bit about this last field. It allows us to mark the different phases of the meeting's life. During the preparation phase the state is "new", and all the values can been changed. After the beginning of the meeting the state becomes "open" and the facilitator is able to add or modify the minutes. In the end, after the closure of the meeting and the sending of the minutes, the state is "closed" and no changes can be made.
Lastly, the Talk structure, by means of which the private colloquies are managed,
46
includes the following slots: Me: the owner of the workstation at which the talk is recorded; You: the other one; Date: when the colloquy took place; Speeches: the report of the undertaken commitments.
General purpose methods have been defined to handle the objects within each of the main modules. Specific methods have also been defined for particular tasks, such as sending and analyzing messages, meetings and talks. The user interface is based on windows, icons, buttons and menus. In our windows normal and scrollable text are available both in input and in output. It has been designed to be as user-friendly as possible, avoiding useless complications in using procedures. UTUCS windows are interchangeably operated and displayed, and so allow the user not only to compile and send e-mail messages, to prepare face to face group meetings and to fill in meeting minutes and face to face talks, but to visualize the contents of CRRs, Minutes and Talks Information Base as well.
Acknowledgements
The UTUCS prototype was developed within the EEC (Esprit-2 Research Project, nr.2121 ITHACA). The authors wish to thank all members of the ITHACA project, in particular Martin Ader and Gang Lu from Bull-Massy, Josep Monguio and Josep
47
Marti from TAO-Barcelona. We also express our thanks to Elsa Bignoli, Fiorella De Cindio, M. Antonietta Grasso, Giuseppe Omodei Salé, Alberto Pozzoli, Carla Simone and the entire staff of the Cooperation Technologies Laboratory for their valuable help. Moreover we thank the anonimous referees of our paper for their constructive comments which allowed us to improve its readability as well as to deepen the explanation of our ideas. Last but not least we gratefully acknowledge Betty Hewitt for her suggestions to improve the quality of its English.
References
Action Technologies Inc. (1988): The Coordinator Version II. Emeryville, California. Ader, Martin, Gang Lu, Sudarshan Murthy, Patrick Pons and Kumar Venkataraman (1990): WooRKS-1, The First Prototype of an Object Oriented Workflow System for Offices. Technical Report ITHACA.Bull.90.D.1.#1, Bull S.A., Paris, France. Agostini, Alessandra, Giorgio De Michelis, Maria Antonietta Grasso and Stefano Patriarca (1993): Reengineering a Business Process with an Innovative Workflow Management System: a Case Study. In COOCS '93. Proceedings of the Conference on Organizational Computing Systems, Milpitas, California, November 1-4, 1993, ed. Simon Kaplan. New York: ACM Press, pp. 154-165. Austin, John L. (1962): How To Do Things With Words. Oxford: Clarendon Press.
48
Bullen, Christine V. and John L. Bennett (1990): Learning from User Experience with Groupware. In CSCW '90. Proceedings of the Third Conference on Computer-Supported Cooperative Work, Los Angeles, California, October 7-10, 1990, New York: The Association for Computing Machinery, pp. 291-302. Chailloux, Jérôme, Matthieu Devin and Jean-Marie Hullot (1984): Le-Lisp: a Portable and Efficient Lisp System. In ACM Symposium on Lisp and Functional Programming. Austin, Texas: Association for Computing Machinery. De Cindio, Fiorella, Giorgio De Michelis, Carla Simone, Raffaela Vassallo and Annamaria Zanaboni (1986): CHAOS as a Coordination Technology. In CSCW '86. Proceedings of the First Conference on Computer-Supported Cooperative Work, Austin, Texas, December 3-5, 1986. Microelectronics Computer Corporation, pp. 325-342. De Michelis, Giorgio, Fiorella De Cindio and Carla Simone (1989): Groups in a Language/Action Perspective. In Proceedings of Second European Meeting on Cognitive Science Approaches to Process Control, Siena, Italy: CEC - JRC Ispra, pp. 205-221. De Michelis, Giorgio and Maria Antonietta Grasso (1993): Routines and Conversations. Structured Programming, Springer-Verlag, vol. 14, pp. 110118. Grice, Paul H. (1975): Logic and Conversation. In Syntax and Semantics - Speech Acts, ed. P. Cole J.L. Morgan. New York, New York: Academic Press, pp. 4158. Heath, Christian and Paul Luff (1991): Disembodied Conduct: Communication through Video in a Multi-Media Office Environment.In Proceedings of CHI'91,
49
Association for Computing Machinery, pp. 99-103. Malone, Thomas W., Kenneth R. Grant and F.A. Turbak (1986): The Information Lens: An Intelligent System for Information Sharing in Organizations. In Proceedings of the CHI'86 Conference on Human Factors in Computing Systems, Boston, Massachusseth, April, 1986. New York: Association for Computing Machinery. Malone, Thomas W., Kenneth R. Grant, Kum-Yew Lai, Ramana Rao and David Rosenblitt (1987): Semistructured Messages Are Surprisingly Useful for Computer-Supported Coordination. ACM Transactions on Office Information Systems , vol. 5, no. 2, pp. 115-131. Malone, Thomas W. and Kevin Crowston (1991): Toward an Interdisciplinary Theory of Coordination, CCS TR 120, MIT. Marca, David (1989): Experiences In Building Usable Meeting Support System. In Proceeding of IFIP The Groupware Technology Workshop. Palo Alto, California. Marti, Josep and Josep Monguio (1990): COP Kernel Design. Technical Report ITHACA.TAO.90.U.2.#17, TAO S.A., Barcelona, Spain. Nunamaker, J.F., Alan R. Dennis, Joseph S. Valacich, Douglas R. Vogel and Joey F. George (1991): Electronic Meeting Systems to Support Group Work. Communication of the ACM, vol. 34, no. 7, pp. 40-61. Olson, Gary M. and Judith S. Olson (1991): User-Centered Design of Collaboration Technology. Journal of Organizational Computing, vol. 1, no. 1, pp. 61-83. Reder, Stephen and Robert G. Schwab (1990): The Temporal Structure of
50
Cooperative Activity. In CSCW '90. Proceedings of the Third Conference on Computer-Supported Cooperative Work, Los Angeles, California, October 7-10, 1990, New York: The Association for Computing Machinery, pp. 303-316. Searle, John (1969): Speech Acts: an Essay in the Philosophy of Language. Cambridge, United Kingdom: Cambridge University Press. Searle, John (1975): A Taxonomy of Illocutionary Acts. In Language, Mind and Knowledge, ed. K. Gunderson. Minneapolis, Minnesota: University of Minnesota Press, pp. 344-369. Suchman, Lucy (1993): Do Categories Have Politics? The Language/Action Perspective Reconsidered. In ECSCW '93. Proceedings of the Third European Conference on Computer-Supported Cooperative Work, Milan, Italy, September 13-17, 1993, eds. Giorgio De Michelis, Carla Simone, and Kjeld Schmidt. Dordrecht, The Netherlands: Kluwer Academic Publishers, pp. 1-14. Stefik, Mark, Gregg Foster, Daniel G. Bobrow, Kenneth Kahn, Stan Lanning and Lucy Suchman (1988): Beyond the Chalkboard: Computer Support for Collaboration and Problem Solving in Meetings. In Computer-Supported Cooperative Work: A Book of Reading, ed. Irene Greif. San Mateo, California: Morgan Kaufmann Publishers, Inc., pp. 335-366. Texier, M. (1985): Sistème de Gestion de Formulaire pour des Applications Bureautiques. Bull MTS, DEA, Massy, France. Viller, Stephen (1991): The Group Facilitator: a CSCW Perspective. In ECSCW '91. Proceedings of the Second European Conference on Computer-Supported Cooperative Work, Amsterdam, The Netherlands, September 24-27, 1991, eds. Liam Bannon, Mike Robinson, and Kjeld Schmidt. Dordrecht, The Netherlands:
51
Kluwer Academic Publishers, pp. 81-95. Whiteside, John and Dennis Wixon (1988): Contextualism as a World View for the Reformation of the Meetings. In CSCW '88. Proceedings of the Second Conference on Computer-Supported Cooperative Work, Portland, Oregon, September 26-28, 1988: New York, New York: The Association for Computing Machinery, pp. 369-376. Winograd, Terry and Fernando Flores (1986): Understanding Computer and Cognition: a New Foundation For Design. Norwood, New Jersey: Ablex Publishing Corporation. Winograd, Terry (1988): A Language/Action Perspective on the Design of Cooperative Work. Human Computer Interaction, vol. 3, no. 1, pp. 3-30.
52