This article was downloaded by: [Northeastern University]On: 11 October 2014, At: 17:45Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

Journal of OrganizationalComputing and ElectronicCommercePublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/hoce20

Providing for OrganizationalMemory in Computer-Supported MeetingsGerhard SchwabePublished online: 22 Jun 2011.

To cite this article: Gerhard Schwabe (1999) Providing for Organizational Memory inComputer-Supported Meetings, Journal of Organizational Computing and ElectronicCommerce, 9:2-3, 151-169, DOI: 10.1080/10919392.1999.9681092

To link to this article: http://dx.doi.org/10.1080/10919392.1999.9681092

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinions and viewsexpressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of theContent should not be relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not be liable for anylosses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly orindirectly in connection with, in relation to or arising out of the use of theContent.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan,

sub-licensing, systematic supply, or distribution in any form to anyone isexpressly forbidden. Terms & Conditions of access and use can be found athttp://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

Providing for Organizational Memoryin Computer-Supported Meetings

Gerhard SchwabeDepartment of Information Systems

Hohenheim University

Meeting memory features are poorly integrated into current group support systems(GSS). In this article, I discuss how to introduce meeting memory functionality into aGSS. The article first introduces the benefits of effective meetings and organizationalmemory to an organization. Then, the following challenges to design are discussed:How to store semantically rich output, how to build up the meeting memory with aminimum of additional effort, how to integrate meeting memory into organizationalmemory, and how to protect the privacy of the meeting participants. Finally, using thegroup-object object-oriented model of a GSS, the article shows how meeting memoryfunctionality can be implemented in a GSS.

organizational memory systems, meeting memory, group support systems,computer-supported cooperative work, object-oriented models

1. INTRODUCTION

Organizational memory has been defined as the “stored information from an orga-nization’s history that can be brought to bear on present decisions” ([1], p. 61) or“the means by which knowledge from the past is brought to bear on present activi-ties, thus resulting in higher or lower levels of organizational effectiveness” ([2], p.89). An organization needs information from its immediate past to support ongoingprojects and coordinate its actions. It needs information from its history to providemodels for future actions, to explain past decisions, to allow forecasts, to counterdamaging myths, to function as a basis of change, and to avoid repeating mistakes[3].

A significant amount of information relevant for future decisions or actions isgenerated in meetings. However, in many conventional meetings, the minutes andthe individual memories are the only records an organization keeps. As humansare poor at remembering the past [4] and particularly poor at remembering the ra-tionale of their decisions [1], they are unreliable containers for meeting memory. Itmay not be feasible to persuade an organization to transfer its memory from indi-

JOURNAL OF ORGANIZATIONAL COMPUTINGAND ELECTRONIC COMMERCE 9(2&3), 151–169 (1999)

Correspondence and requests for reprints should be sent to Gerhard Schwabe, Department of Infor-mation Systems, University of Koblenz–Landau, Rheinau 1, D–56075 Koblenz, Germany. E-mail:schwabe@uni-koblenz.de

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

viduals to a computer, but the organization should make the best use of the infor-mation captured by the computer to enhance organizational memory.Computer-supported meetings offer an opportunity to store organizational mem-ory in the computer because in such meetings much information is entered in thecomputer.

Group support systems (GSS) introduce unique opportunities and challengesfor keeping more extensive and flexible meeting records. Yet, current GSSs such asGroupSystems [5–7] or Vision-Quest [8] are ill prepared to store meeting informa-tion for a longer time than the meeting itself. Morrison and colleagues [9, 10] tack-led this problem by providing a database that can import and store meeting output.In this article, I go one step further and show how one can enhance the GSS itselfwith meeting memory features. To this end, I have developed an object-orientedmodel of a GSS that enables data collection during a meeting in a way that allowseasy access to all organizational context and content information.

The following sections discuss the unique opportunities GSSs provide for keep-ing better meeting records. Challenges to record keeping that are not completelymet by current systems are also addressed. Finally an object-oriented model of aGSS shows how these challenges can be met in system design. A section on imple-mentation, limitations, future research, and a summary close the article.

2. OPPORTUNITIES FOR IMPROVING ORGANIZATIONAL MEMORYTHROUGH GSS

Traditional meeting minutes are based on a meeting style that depends heavily onoral discussions. In some instances participants hand out documents beforehand orgive a prepared presentation, but the heart of face-to-face meeting activities is oral.To collect a more elaborate meeting memory, one first has to change the meetingstyle to one that relies on joint work on common material [11], also called common ar-tifacts [12]. Professional facilitators therefore introduce flip charts, white boards,cards, and other material [13], making meetings less talk and more work. These ma-terials are distributed all over the room (in the hands of participants, on the walls,etc.) and participants work on them, sometimes alone (e.g., writing a card), some-times in subgroups (on separate flip charts), and sometimes as a complete group. Inthe course of the meeting, the entire room resembles the group memory. Anyonecan refer to anything that he/she or someone in the room has put on paper. Unfortu-nately, this group memory is destroyed once the group has finished its work and theroom is prepared for the next meeting. Flip charts with cards or drawings on themare difficult to archive, difficult to reuse, and impossible to share among the groupmembers outside meetings. Photographs of them are a poor substitute with regardto their expressive power. Furthermore, the whole work has lost its context once it isremoved from the room.

GSSs expand the idea of joint work on common material. The material can beshared in a more flexible way, participants can contribute in a true parallel fashionand in new forms of work (e.g., anonymously or asynchronously) [11, 14]. In thisarticle, I focus on the memory abilities of a GSS: It automatically stores all informa-tion that has been entered in the course of the meeting and makes it accessible to

152 SCHWABE

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

anybody interested during and after the meeting. Once the information is both cre-ated and stored on a computer, it is even possible to maintain much of its originalcontext. Thus, the group memory is not automatically destroyed after the meetingas in the flip chart meeting. It can be integrated into a project memory or an organi-zational memory, retrieved, reused, and built on by individuals, by the samegroup, or by other groups of the organization. As the stored information has a con-text, anybody can reconstruct the history of the meeting, including product cre-ation and the rationale of decisions made in the course of a meeting. Thus, ameeting memory does not only archive what was produced, but also how it wasproduced, when a contribution was made, who contributed, and why decisionswere made.1

Specifically a meeting memory can contain:

• What has been produced: The products of a meeting are archived in the form oftexts, collections of comments, outlines, tables, drawings, votes, and diagrams. Theproducts of the meeting contain what has been worked on, what has been agreedon, or what has been decided.

• How each product has evolved: The meeting memory collects information on theproblem-solving procedure used (e.g., brainstorming) for each agenda item, the ac-tivities (e.g., presentation, group writing, voting, outlining), and the content con-texts (e.g., as an answer to a question, as a rebuttal, as a free-standing idea) ofcontributions. A meeting memory can even contain a history of all changes thathave been applied to materials if it stores multiple versions of products.

• When any product has been created: Both the absolute time (e.g., June 1, 1993 at1:25 p.m.) of any product’s or contribution’s creation, and the creation time in rela-tion to any other contribution’s creation time (i.e., has it been written before, dur-ing, or after another contribution) are stored. Based on extensive time-stampinginformation, a whole meeting can be replayed on the computer, making it mucheasier for new members to enter an ongoing activity. The duration of any activityindicates how much effort has been put into the activity. For example, a decisionmade after a long discussion is probably much better grounded than a decision thatwas not discussed at all.

• Where a meeting has taken place: Times and places trigger memories of the meet-ing content [15] and can serve as filters when searching for information.

• Who has contributed: Archiving information about who has made a contribu-tion gives the participants a sense of ownership and responsibility for meeting out-put. Contributions can be made by a single identified participant, an anonymousparticipant, the facilitator (on behalf of the group), several participants, a sub-group, or the whole group. Authorship information is also important to judge thevalue of a contribution (e.g., from an expert). Finally, ownership allows each partic-ipant to take his/her personal cut at one or several meetings, filtering what peoplethat interest him/her have contributed.

• Why a decision has been made: The more time that has passed since a decisionhas been made, the harder it is for an organization to find out why a decision wasmade the way it was, especially if the persons who originally made the decision

ORGANIZATIONAL MEMORY 153

1 Walsh and Ungson [1] also used the journalistic questions of how, why, when, and where.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

have long since left the company. Without knowing the rationale behind a deci-sion, nobody can determine if a decision was a particularly clever one or if it wasbased on assumptions that are no longer valid. Even worse, without knowing therationale behind a decision, often one is not able to follow the intentions of deci-sions even if one wants to. It is exactly the qualitative information needed to under-stand the rationale behind decisions that is being exchanged in meetings. Ameeting memory can make decision rationale explicit if participants have usedspecial tools such as issue based information systems (IBIS) [16, 17]. It can also al-low persons to reconstruct the rationale from the meeting output and the meetingcontext information [18]. The richer the meeting documentation is, the easier it is toreconstruct the decision rationale.

3. CHALLENGES TO GSS

3.1 Storing Rich Output

Why not just store the GSS output as documents and forget about the rest? One basicargument has been elaborated on already: Reducing GSS output to conventionaldocuments takes the contributions out of their context. Looking at typical meetingminutes, one senses how much effort the writer goes through to reconstruct themeeting context (describing who said what, when, and how).

The other argument against storing GSS output as conventional documents isthat typical meeting output does not resemble a conventional document, but looksmore like a hyperdocument structure [19]. Take, for example, a meeting in which agroup first brainstorms and discusses ideas with a brainstorming tool, then orga-nizes the ideas with an outliner, and finally votes on selected issues. The gatheredinformation resembles a network in which several structures overlay one another:Brainstorming typically organizes information into several sheets of comments.Comments on these sheets can refer to one another (i.e., have content links). Someof these comments are linked to an author and others are anonymous. Afterward,the comments are organized into the treelike structure of an outline. It would beuseful not to be forced to destroy the structure of the brainstorming output by thisreorganization and at the same time avoid having to duplicate the information. Du-plication of information (e.g., copying all brainstorming remarks before reorganiz-ing them into an outline) tends to make the information inconsistent and makes asearch for a particular comment much more burdensome (i.e., which copy do youwant?). If an issue has been voted on and this issue was an element of an outline, auser should be able to access the voting information from the outline.

Thus, the meeting memory has to be stored in finer grains than conventionaldocuments. According to my experiences, a comment or paragraph as an atomicunit for meeting output that is mainly textual is a fair compromise between theneeded flexibility and the danger of getting lost in hyperspace. A paragraph canbe seen as the smallest unit of text that makes sense in itself. Using words or sen-tences as atomic units leads to a large overhead for storing information andthreatens the performance of a system. Using speech acts [18] places an excessiveadditional burden on the participants because it requires participants to make

154 SCHWABE

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

their intentions explicit all the time. These basic units need to be linked to othercontent information and to context information such as authors, dates, andagenda items.

3.2 Building Up Memory With a Minimum of Additional Effort

Participants want to concentrate on the meeting itself. They have little understand-ing for additional work that is not directly related to the meeting purpose. Askingparticipants to organize information in a particular way required by an organiza-tional memory system (e.g., into flat tables of a relational database) is not accept-able. The challenge therefore is to make the best use of the data collected in thecourse of the meeting. For example, time-stamping information can be collected au-tomatically in the background and authoring information requires one login of eachparticipant at the beginning of the meeting. A detailed discussion of informationthat can be automatically collected in a GSS is given in the discussion of the ob-ject-oriented model of GSS.

3.3 Making Information Easily Accessible and Maintainable

Organizations that memorize their history must also be able to organize, access, andforget this memory in large chunks (organizations may be forced to forget informa-tion, e.g., by laws protecting privacy). Although it is necessary to store veryfine-grained information, it must be possible to find and recover large chunks of in-formation (e.g., all meeting output of a project) for further inspection or for deletion.As meeting work and meeting decisions may directly influence day-to-day work,information should be accessible from anywhere.

Generic techniques for finding information in unstructured or semistructureddocuments are discussed in information retrieval. Keyword indexes or search treesimpose an artificial order within and across documents; query languages allow so-phisticated questions. These techniques can be applied to meeting memories andare useful for retrieving meeting information from an organizational memory [20].Including meeting information in a database that allows flexible search mecha-nisms becomes paramount once a larger institution archives meeting informationfor any length of time. This requirement might look obvious, but typical GSSs (e.g.,GroupSystems, VisionQuest, SAMM) are neither built on top of nor provide an in-terface2 to database management systems (DBMS). The main difficulty lies in thestructure of traditional relational DBMS that do not support storing data of arbi-trary length and type as is typical for meeting output. Extended relational DBMS[21], object-oriented DBMS [22], and office form DBMS (e.g., Lotus Notes® [23])should improve possibilities for archiving meeting data in DBMS and integratingmeeting information with other office information.

ORGANIZATIONAL MEMORY 155

2Except for ASCII export.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

3.4 Integration With Other Organizational Memory Data

Meeting output needs to be integrated with other office information. A prerequisitefor integration is a network connection between the meeting environment and theoffice environment. The easiest way to integrate meeting output into the personaloffice environment is to convert the output to documents and allow access to thedocuments via a central file server. Yet, this approach has several disadvantages.Converting meeting output to documents takes the information out of its context.Participants might lose valuable clues for understanding the meaning of a contribu-tion, making it much harder to synchronously or asynchronously continue a meet-ing. If one duplicates the output, keeping a snapshot of a meeting for possible con-tinuation, in the GSS and a conventional document in the office environment asorganizational memory, both copies quickly get unsynchronized. Furthermore, filesystems have only rudimentary access and search mechanisms. Another approachis to keep only the GSS meeting snapshot and access the GSS directly from an indi-vidual office. This allows “anytime, anyplace” access to the meeting memories, butit does not integrate the meeting memory with the rest of the individual data and theorganizational memory. A lack of integration leads to duplication of data, inconsis-tency of data, longer search times, increased teaching efforts, and loss of links be-tween meeting data and other data.

There are three mechanisms that allow for a closer coupling of a given meetingmemory and other information. External links connect meeting contributions withoffice information via a mechanism of the office information system. For exampleMicrosoft Windows supports information linkages with object linking and embed-ding. In theory, if a piece of information (e.g., a comment) embedded in another ob-ject (e.g., a document) is updated, the embedding object reflects these changesimmediately. Current external linking mechanisms have only rudimentary syn-chronization facilities and basically target single-user support. As the navigationalfeatures of external links (how one finds information) and their security features(how I prevent somebody else from unintentionally removing objects that I need)are poor, these links remain problematic as a basis for an organizational memory.

Internal links connect meeting outputs with office information via a mechanismof the GSS. Any information that is related to a meeting is imported into the GSSand becomes an integral part of the meeting memory.3 Although internal linkswork well during the meeting itself, they are very cumbersome for an organiza-tional memory. Very soon, a lot of information that is produced in an organizationis being used both in and outside of meetings. During a meeting one can only usethe tools of a GSS; outside the meeting one wants to use personal tools (such as aword processor). Typically this conflict results in a vast duplication of informationthat again almost certainly will be inconsistent. If one keeps only a reference to re-lated office information in the meeting memory (e.g., storing the file name), thesame problems as with external links apply.

Integrated organizational memory seamlessly embeds the meeting environmentinto the office environment and the meeting memory into the organizationalmemory. The basis for an integration is a DBMS that contains and organizes both

156 SCHWABE

3This is the approach of current group support systems (e.g., GroupSystems).

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

meeting data and other office information. This DBMS imposes a common orga-nizational structure on both meeting and individual data. Typically an organiza-tion collects data under projects or departments. Each member can then accessand work on all of his/her information in meetings and in his/her normal officeenvironment. If a person wants to seamlessly move from individual work togroup work and vice versa, the DBMS must store a common representation foroutput produced by personal tools and group tools. Optimally the same set oftools works for both individual and group work. GO (described later) follows theintegrated organizational memory model, although it is probably unrealistic toexpect an organization to switch to the new DBMS required. However, I believeit is better to start with a clean architecture and later make compromises in theactual organizational implementation than to design for compromise from thevery beginning.

3.5 Protecting Privacy

Archiving who said what, when, why, and to whom offers good opportunities for abetter organizational memory. In many instances the stored information is in-tended to be freely shared by any interested persons within the organization. How-ever, in other instances meeting information should remain confidential to thegroup, shared with only part of the group or intentionally only mentioned orallyand not put on paper. Conventional meeting minutes are carefully crafted docu-ments containing only what they have been agreed to contain (or what the minutetaker thinks the group has agreed on). A GSS that interferes with these habits andacts like a “Big Brother” will not be accepted. However, allowing for anonymouscontributions turned out one of the most important features of GSS in the eyes of theparticipants [24, 25]. Its additional privacy protection can be one of the major facili-tators for introducing GSS to business groups. Thus, a meeting memory should notonly protect sensitive information and the privacy of the meeting participants (as isalready being done in a conventional meeting); it should also make the best use ofinformation technology to enhance confidentiality when necessary to allow for afree discussion or to protect the interests of the individual. Measures to achieve thisend can be taken before, during, and after a meeting.

Preplanning for privacy before the meeting: Privacy issues should be addressed inthe preplanning of a meeting. A facilitator plans for each agenda item whether it isdiscussed anonymously, semianonymously,4 or openly. At the beginning of themeeting the facilitator briefly introduces new participants to the meeting memoryand privacy protection capabilities of the system. Privacy measures during themeeting include the following:

• Allowing pseudonyms: Any person who wants to keep his/her participationconfidential logs in under a pseudonym. Contributions by a person using a pseud-onym are owned by the group.

ORGANIZATIONAL MEMORY 157

4See later discussion of leveled precision of authorship information.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

• Duplicate control of anonymity: During the meeting anonymity, awareness(who is doing what now) [26], and authorship (who has contributed what in thepast) remain under the control of both the facilitator and each individual member.Leaving anonymity under facilitator control is cumbersome because participantsmight want to quickly switch between different anonymity modes. Leaving ano-nymity under participant control only can endanger anonymity because a singleindividual cannot remain anonymous on his/her own if all other participants con-tribute openly. There are also instances in which a facilitator wants to enforce ano-nymity to avoid group pressure for openness (as it happened in the Eastern bloc,where all citizens “voluntarily” voted for the Communist party). Leaving anonym-ity under duplicate control means that a contribution remains anonymous if one(i.e., the facilitator or the participant) wants it this way.

• Leveled precision of authorship information: For protecting privacy in the meetingmemory the control of authorship is important. To reconcile the need to under-stand information out of context and the need to protect privacy, a leveled ap-proach to authorship is useful: The (real) author can decide if he/she wants thecontribution to be stored under his/her individual authorship (e.g., John Miller),his/her subgroup authorship (e.g., the soap marketing task force), his/her group’sauthorship (e.g., the marketing department that has gathered for a meeting at thismoment), his/her organization’s authorship (e.g., the perfume company), or abso-lute anonymity (anonymous). Usually each participant does not decide each con-tribution’s authorship on his/her own, but he/she uses default values that he/sheonly changes in a few instances.

• Strong and weak memory: In a typical meeting there are contributions that areexplicitly made for the record and other contributions that are not intended toleave the room. There should not be any records of such information because theauthor does not want to be committed to it. A meeting memory can support contri-butions without trace by differentiating between weak memory and strong mem-ory. Strong memory contains all information that is to be archived; information inweak memory is automatically deleted once the meeting is over. The decisionwhether a contribution belongs to weak or strong memory can be made at any timein the meeting: when the contribution is made, after the related agenda item hasbeen discussed, after it has been voted on, or at the end of the meeting. Strong andweak memory may not protect participants within a meeting from one another, asany participant may very easily make a copy of the weak memory, but they protectthe group members from others outside the meeting. Thus a spirit of confidential-ity is supported.

Privacy measures after the meeting include the following:

• Access rights: Just as any other office information, the meeting memories in theorganizational memory must be protected against misuse and sabotage. Tradi-tionally DBMS and operating systems take care of data protection by giving accessprivileges only to selected persons. This requirement appears obvious for DBMS.However, currently GSS are typically not connected to DBMS (discussed earlier).

• Formal meeting minutes: In conventional meetings often one person has the roleof a minute taker. During the meeting he/she takes notes of the participants’ con-

158 SCHWABE

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

tributions. After the meeting he/she goes through his/her notes and tries to findout what the group has done and agreed on, in the process filtering all commentsthat he/she thinks participants want to be “off the record.” Afterward, the meetingminutes are distributed to the participants and the minutes are formally agreed on.Only then do they become part of the formal organizational memory. A similarpostmeeting review of the meeting output can be done in computer-supportedmeetings: A minute taker is chosen and this individual goes through all the contri-butions after the meeting, correcting or deleting problematic or irrelevant com-ments. This condensing process can even result in the usual brief meetingminutes.5 The minute taker then distributes the meeting output to the participants,allowing each to voice objections. Only then does the meeting memory becomepart of the organizational memory.

• Fuzzy time stamps: Time-stamping information can corrupt other privacy mea-sures, as other information can be reconstructed from timing information. Storingthe exact time of the contribution makes it possible to pin down the meeting andthe participants. Detailed time-stamping information can be important during themeeting for synchronization of work, but the storing of detailed time-stamping in-formation can be relaxed after the meeting. One can store fuzzy time stamps.Storing only the hours or date of a contribution protects the individual participantwithin the meeting from being identified with a comment. Storing only relativetime stamps (starting a clock at 0:00:0) helps to protect a contribution from beingidentified with a certain meeting.6 Relative time stamps keep enough informationto replay a meeting.

• Control of authorship: Authorship of contributions does not necessarily need tostay the same after a meeting as it has been during a meeting. A GSS can allow par-ticipants to generalize authorship after the meeting if they do not want to be identi-fied with a contribution outside the group. A minute taker or any other participantcan generalize authorship by making individual contributions into subgroup con-tributions, subgroup contributions into group contributions, and group contribu-tions into organizational contributions. In addition, a contribution may be madeanonymous.

How can privacy protection be reconciled with building up organizationalmemory using a minimum of additional effort? Do the strategies of pseudonyms,duplicate control of anonymity, controlled authorship, strong and weak memory,and fuzzy time stamps put too much of an additional cognitive burden on theparticipants? One can assume that an additional cognitive burden is acceptableto the participants in two cases: (a) if they must perform the actions only once ina meeting, or (b) if the actions are of their own interest and do not have to be per-formed continuously. Participants have to log in with a pseudonym only once ina meeting, fuzzy time stamps are applied to all meeting output only once afterthe meeting, and strong and weak memory have to be cleaned up only once afterthe meeting. Individual control of anonymity, control of authorship, and mark-

ORGANIZATIONAL MEMORY 159

5Even if this would destroy many of the enhanced organizational memory features.6This can of course only be achieved if other context information is also removed.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

ing weak and strong memory are in the interests of each participant. Workingwith default values that are preset by the facilitator for a meeting or session andcan be easily overridden by each participant should be a feasible approach.

4. AN OBJECT-ORIENTED MODEL OF A GSS

I have developed the Group Objects (GO) GSS architecture and a GO prototype thattake organizational memory requirements for GSS into account [3] and show how aGSS designer can integrate meeting memory features into a GSS. The GO architec-ture is an object-oriented model of a GSS. GO is based on numerous observations ofactual and experimental computer-supported sessions7 and an analysis of currentGSS. The GO prototype is a Smalltalk–80 program intended to be used in conjunc-tion with an object-oriented DBMS. The following paragraphs discuss those parts ofthe GO architecture that are relevant for meeting memory and organizational mem-ory. The object-oriented description language is described in the Appendix.

The model consists of two parts: The left part (Subject 1) is called the com-puter-supported cooperative work (CSCW) part, as it describes sharing of informa-tion among people. The right organizational part (Subject 2) consists of objects thatare typically needed to run a meeting in an organizational setting (see Figure 1).The attributes and services of class and objects8 just list a few features and activitiesnecessary for understanding the model, and are by no means complete.

4.1 CSCW Part

The CSCW part centers around the class and objects tool9 and material (Figure 2) [27].Participants use tools to manipulate their material. Tools can be work tools such asoutliners, voting tools, or simple text editors; or memory tools such as privacy tools orquery tools. Material can be a simple text, an outline, an outline node, a paragraph, agraph, a drawing, a matrix, a matrix element, a vote, and so on. The subclasses of tooland material serve as examples for a much larger potential range of text- andgraphic-based tools. Note that neither tools are the same as (PASCAL) procedures,nor are materials the same as data: Tools have their own state (e.g., a text editorknows where to insert the next character) and their own attributes (e.g., the pre-ferred font). Material is not plain data, as it has the sole knowledge of how it shouldbe manipulated (e.g., a tree knows how to insert a node or any material knows howto broadcast the fact that it is locked).

The dependency between tools and material is one way: A tool knows the in-terface procedures of a material, but the material does not have any informationabout the tools that work on them [28]. To make this idea plausible, I offer an ex-

160 SCHWABE

7This was conducted at the University of Arizona, University of Michigan, and Hohenheim Univer-sity.

8The term class and objects is roughly equivalent to the term class, for futher explanation read the de-scription of the modeling language in the Appendix.

9All class and objects names in the following paragraphs are italicized when they are introduced forthe first time.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

ample. The meat does not need to know anything about the knife to be cut, butthe knife needs to take the features of the meat into consideration to be able tocut it. Group work is then seen as common work with individual tools on ashared material. This way one is able to build all kinds of different tools withoutever having to change the material.

A major benefit of a clean separation between tools and material lies in the factthat the meeting memory only needs to store the material and forget the tool infor-mation as soon as the tool is not needed any more. Material contains the informa-tion to be stored and administrated in an object-oriented DBMS.10 Tools containthe information that should be stored locally, if at all. Storing material in a DBMSenables DBMS support in locking, synchronization, distribution, security, andquery functionality, fulfilling many of the requirements mentioned earlier.

Creating one abstract class of material with concrete subclasses (outline, vote,etc.) concentrates all the authoring, locking, time stamping, and weak and strongmemory administration in one place (see Figure 3). Explicitly storing authoring in-formation of a node protects privacy with access control and allows users to searchfor what a given person has contributed (for a deeper discussion, see [29]). Timestamps allow for searching on time and order in which contributions were made;

ORGANIZATIONAL MEMORY 161

Figure 1. Overview of basic structural components of the Group Objects architecture.

10Object-oriented database management systems store class and objects with all their attributes andservices; that is, they are able to store the complete material, not just the data part.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

fuzzy time stamp mechanisms can be implemented at one place. The privacy pro-tection scheme of strong and weak memory can be reduced to one simple attribute(weak–strong) and a garbage collector that is automatically started at the end of themeeting. The privacy tool allows participants to control privacy during and afterthe meeting. Because material can be as large as a document with thousands ofpages and as fine-grained as one paragraph11 or one drawing element, it is rich andeasily accessible and maintainable.

4.2 Organizational Part

The organizational part (right side) supports administration of meetings and putsgroup work into an organizational context. As argued earlier, the meeting context isan important part of the organizational memory. Specifically one wants to store the

162 SCHWABE

11Or even more fine-grained; the administration overhead is huge, however, if one, for instance, useswords as the finest grain.

Figure 3. People and material.

Figure 2. Tools and material.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

agenda (i.e., the sessions a meeting consists of, see Figure 4). Each agenda item islinked with the material that has been produced during this session. If the sessionhas been part of a problem-solving procedure, this information is also stored. Thelink between a session and a tool type records how the material was created.

A hierarchical structure of projects12 collects meetings into ever larger organiza-tional units (see Figure 5); projects also integrate material that has been producedoutside a meeting. The meeting context gives clues about how the product hasevolved. Storing the agenda and linking the sessions with their material and proce-dures helps to indicate later why a decision was made.

GO strictly separates standing groups (called teams) and acting groups [30](called groups; see Figure 6). Teams and their members (called persons) representthe long-term organizational structure; groups and their members (called partici-pants) represent the actors present in the meeting room. Teams and groups fre-quently overlap but are not necessarily identical. In a meeting, a member of theteam can be missing or an outsider can be invited to join the group. Information ongroups is needed to run the actual meeting (e.g., for session management or forawareness of what others are doing). Information on teams is interesting for orga-nizational memory (e.g., who has contributed, what project does the work belongto). Persons are identified by their name, but participants need only be internallyidentified by their workstation number.

During the meeting, participant information (e.g., Participant1) and person in-formation (e.g., Tom) are linked. The link allows participants to be aware wherepeople work. For example, Tom can see in his/her outliner that Kathy is workingon the last paragraph of the shared outline: The system follows the link from partic-ipant data to the person data to provide the information. The person information is

ORGANIZATIONAL MEMORY 163

Figure 4. Organizational part.

Figure 5. Projects.

12The hierarchical structures of project, group, and team are left out in all other drawings to increasereadability.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

logged by the GSS; the participant information leaves no trace. While Kathy isworking on a paragraph of a shared outline, both her participant object and herperson object are connected to the paragraph material. After she has finished, thelink to her participant object is cut, but the link to her person object (called author-ship) is kept for organizational memory. After the meeting the link between partici-pant and person objects is cut and there is no way to trace a contribution to itsauthor other than by the stored person information. The separation between per-son and participant information allows for a clean separation between sessionmanagement and organizational memory management. The stored person object(e.g., Tom) can freely be manipulated during and after the meeting without affect-ing the system performance in the meeting: Whenever a group member wants togeneralize authorship (i.e., make a contribution group owned or totally anony-mous), he/she just relinks his/her participant to another pseudoperson. Thispseudoperson can have the name of a subgroup, group, organization, or no name.Because the linking information is not stored, there is no way to trace the actualidentity of an author of a comment after he/she has relinked to his/her originalidentity. Hiding participant information also protects the privacy of all groupmembers who have logged in under a pseudonym.

There remains one problem to be addressed: An organizational memory thathas been growing over the years should serve as an encyclopedia. Encyclopediasorganize information by topics that cut across time, space, and persons. Thus, thereneeds to be a topic class and object that serves as a container for all meetings, pro-jects, and materials regarding a particular topic (see Figure 7) [9, 10]. In GO, topicsalso contain a reference to persons who have identified themselves as experts on aparticular topic. One can use this person information to scan through meetings inwhich the expert has participated.

164 SCHWABE

Figure 6. Groups and teams.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

5. IMPLEMENTATION, LIMITATIONS, AND FUTURE RESEARCH

The GO architecture has been successfully implemented as an ObjectworksSmalltalk prototype (described in detail in [3]). This prototype is running on a singlecomputer simulating several participants. However, the prototype has not beentested in a real meeting. Currently, we are leading the Cuparla research project sup-porting the collaboration of the Stuttgart City Council [31].13 In this project, meet-ings are supported with GroupSystems and asynchronous collaboration is sup-ported with Notes. GroupSystems has an interface that stores meeting output in aNotes database. The participants can continue to work on the meeting documents inNotes and reload the documents into GroupSystems. Collaboration in a city councilraises the organizational memory and privacy issues discussed so far. Therefore,we currently work on including organizational memory features into Notes data-bases and applications. Using Notes comes close to the described ideal of an inte-grated organizational memory. However, some new issues have appeared duringthe implementation efforts.

1. Collaboration support has to be very easy to use and make the best use of thelearned working behavior. This behavior is very context dependent. For example,there is much more openness in meetings within the same party than in meetingswith members from different parties. We therefore provide software “rooms” [32]as a working context (e.g., a party room). Rooms give access to information forsome specific group and protect its privacy from other groups. The first usage ex-periences are encouraging.

2. Once the organizational memory grows, finding information becomes an is-sue. Particularly, awareness of what has changed is important. We are experiment-ing with intelligence agents to identify changes to information that are relevant toparticular users. An event mechanism would then notify the users of the changes.

ORGANIZATIONAL MEMORY 165

Figure 7. Topics.

13Cuparla is part of the research and development program of the DeTeBerkom GmbH, a whollyowned subsidiary of German Telekom.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

6. CONCLUSIONS

In this article, I describe how collecting organizational memory information duringmeetings benefits an organization: It allows participants to access what has beendone and agreed on, when, how, and by whom. With this information, a participantcan reconstruct much more of the meeting context than conventional meeting min-utes or the document output typical of current GSS allow. The meeting context willhelp determine why decisions have been made. Challenges that organizationalmemory pose to GSS include storing rich output, allowing participants to build upthe meeting memory with a minimum of additional effort, making information eas-ily accessible and maintainable, integrating current outputs with other organiza-tional memory, and protecting privacy. An extract of the GO object-oriented modelwas presented to show how a GSS designer can cope with challenges posed by orga-nizational memory in a simple way. Organizational memory requires databasefunctionality of GSS. The longer an organization uses a GSS, the more importantthis database functionality becomes. In this article, I propose enhancing GSS to storemeeting information in a DBMS, integrate meeting memory into organizationalmemory, make meeting information available to people outside meetings, and al-low them to coordinate their work via sharing material and context. First steps to-ward the realization of this aim are currently being taken in the Cuparla project.

REFERENCES

[1] J. P. Walsh and G. R. Ungson, “Organizational memory,” Academy of Management Rev., vol. 16, pp.57–91, Jan. 1991.

[2] E. Stein and V. Zwass, “Actualizing organizational memory with information systems,” Informa-tion Systems Research, vol. 6, no. 2, 1995, pp. 85–117.

[3] G. Schwabe, Objekte der Gruppenarbeit (Objects of Group Works). Wiesbaden, Germany: Gabler,1995.

[4] G. Huber, “A theory of the effects of advanced information technologies on organizational design,intelligence, and decision making,” Academy of Management Rev., vol. 15, no. 1, pp. 47–71, 1990.

[5] A. Dennis, J. George, L. Jessup, J. Nunamaker, and D. Vogel, “Information technology to supportelectronic meetings,” MIS Quarterly, vol. 12, no. 4, pp. 591–624, 1988.

[6] GroupSystems Manual, Basic Tools Version 4.0. Tucson, AZ: Ventana Corp., 1990.[7] H. Lewe and H. Krcmar, “GroupSystems. Aufbau und Auswirkungen (GroupSystems: Tools and

effects),” Information Management, vol. 7, pp. 32–41, 1992.[8] Visionquest Users Guide. Austin, TX: Collaborative Technologies, 1991.[9] J. Morrison, M. Morrsion, and D. Vogel, “Software to support business teams,” Group Decision and

Negotiation, vol. 1, no. 2, pp. 91–115, 1992.[10] J. Morrison, “Team memory: Information management for business teams,” in Proc. 26th Hawaii

Int. Conf. on Systems Sciences, 1993, pp. 122–131.[11] G. Schwabe, “Computerunterstützte Sitzungen (Computer supported meetings),” IM-Information

Management, vol. 9, pp. 34–43, July 1994.[12] M. Robinson, “Keyracks and computers: An introduction to ‘common artefacts’ in computer sup-

ported cooperative work,” Wirtschaftsinformatik, vol. 35, pp. 157–166, Oct. 1992.[13] R. Bostrom, R. Anson, and V. Clawson, “Group facilitation and group support systems,” Dept. of

Management, University of Georgia, Athens, Working Paper, July 1991.[14] H. Krcmar, H. Lewe, and G. Schwabe, “Empirical CATeam research in meetings,” in Proc. 27th Ha-

waii Int. Conf. on System Sciences (HICSS94) Vol. IV, 1994, pp. 31–40.

166 SCHWABE

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

[15] W. Newman, M. Eldridge, and M. Lamming, “PEPYS: Generating autobiographies by automatictracking,” in Proc. ECSCW91, 1991, pp. 175–188.

[16] W. Kunz and H. Rittel, “Issues as elements of information systems,” des Institut für Grundlagender Planung igp, Stuttgart, Germany, Working Paper No: S–78–2, 1970.

[17] J. Conklin and M. Begeman, “gIBIS: A hypertext tool for exploratory policy,” ACM Trans. on OfficeInformation Systems, vol. 6, pp. 303–331, Oct. 1988.

[18] K. Sandoe, L. Olfman, and M. Mandviwalla, “Meeting in time: Recording the workgroup conver-sations,” Proc. 12th ICIS, 1991, pp. 261–272.

[19] U. Hahn, M. Jarke, S. Eherer, and K. Krplin, “CoAUTHOR—A hypermedia group authoring envi-ronment,” in Studies in Computer Supported Cooperative Work, J. Bowers and S. Benford, Eds. Am-sterdam: Elsevier, 1991, pp. 79–100.

[20] H. Chen, W. McHenry, K. Lynch, and S. Goodman, “Computer-based support for organizationalmemory: Frame-work and design,” MIS Dept., University of Arizona, Tucson, Working Paper91–37, 1991.

[21] M. Stonebraker and G. Kemnitz, “The POSTGRES next-generation database management sys-tem,” CACM, vol. 34, pp. 78–92, Aug. 1991.

[22] M. Atkinson, F. Bancilhon, D. DeWitt, K. Dittrich, D. Maler, and S. Zdonik, “The object-orienteddatabase system manifesto,” in Proc. Conf. Deductive and Object-Oriented Databases, 1989, pp. 40–57.

[23] S. De Jean, “Lotus Notes, Version 2.0: Connectivity and new pricing broaden its appeal,” PC Maga-zine, vol. 10, pp. 33–35, May 1991.

[24] A. Dennis, Parallelism, Anonymity, Structure and Group Size in Electronic Meetings, University of Ari-zona Graduate College, Tucson, unpublished doctoral dissertation, May 1991.

[25] J. Valacich, L. Jessup, A. Dennis, and J. Nunamaker, “A conceptual framework of anonymity ingroup support systems,” Group Decision and Negotation, vol. 1, no. 4, pp. 219–242, 1992.

[26] P. Dourish and V. Belotti, “Awareness and coordination in shared workspaces,” in Proc. CSCW ’92,1992, pp. 107–114.

[27] R. Budde and H. Züllighoven, Softwarewerkzeuge in einer Programmierwerkstatt. Berichte der GMD(Software Tools in a Programming Workshop). Munich, Germany: Oldenbourg, 1990.

[28] G. Gryczan and H. Züllighoven, “Objektorientierte Systementwicklung—Leitbild undEntwicklungsdokumente (Object oriented software development—Role model and documents),”Informatik Spektrum, vol. 15, pp. 264–272, Oct. 1992.

[29] M. Mandviwalla, K. Sandoe, and S. Clark. “Collaborative writing as a process of formalizing groupmemory,” in Proc. 28th Annual Hawaii Int. Conf. on System Sciences, 1995, pp. 342–351.

[30] J. McGrath, Groups: Interaction and Performance. Englewood Cliffs, NJ: Prentice-Hall, 1984.[31] G. Schwabe, “Understanding and supporting knowledge management and organizational mem-

ory in a city council,” in Hawaii International Conference on System Sciences (HICSS99)[CD–ROM], 1999.

[32] D. Henderson and S. Card, “Rooms—The use of multiple virtual workspaces to reduce space con-tention in a Window-based graphical user interface,” ACM Trans. on Graphics, vol. 5, no.3, 1986, pp.211–243.

[33] P. Coad and E. Yourdan, Object-Oriented Analysis. Englewood Cliffs, NJ: Prentice-Hall, 1991.

APPENDIXThe Object-Oriented Description Language

GO is described using the Coad and Yourdan notation [33]. Class and objects14

stand for existing or conceptual unities (e.g., Tom, Tom’s Volkswagen, UnitedStates, or antiapartheid legislation). To be more precise, class and objects serve as

ORGANIZATIONAL MEMORY 167

14There is no difference with the term class, normally used in object orientation. However, in this nota-tion the term class has a different meaning (discussed later).

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

“cookie cutters” for such unities, also called instances: A person class and objectserves as a cookie cutter for the instance Tom; a car class and object serves as a cookiecutter for the instance Tom’s VW; and so on. Class and objects are described by theirattributes, which describe their features and their state, and by their services, whichmake known what the object can do for you. For example a car has the attributecolor and a car offers the service of moving persons (see Figure A1).

Classes model conceptual unities like class and objects. The only difference isthat they are purely abstract. They are cookie cutters, but there are not supposed tobe any cookies produced with them. They only make sense in conjunction withgen–spec structures. A vehicle is a superclass of a car, but an instance of a vehiclewill not exist (see Figure A2).

Gen–spec structures describe an abstraction hierarchy between classes. Thehigher class is a generalization of the lower class, or, in other words, the lower classis a specialization of the higher class. A vehicle is a generalization of a car (becausethere are other vehicles such as trucks or carriages) and a car is a specialization of avehicle (see Figure A3).

Whole–part structures describe the relation between a whole instance and itsparts. For example, the parts of a car are its engine, the doors, the seat belts, and soon. It is normally not so easy to differentiate between whole–part structures and in-stance connections. A rule of thumb says that a whole–part relation exists if theparts cease to exist with the whole (e.g., if your car burns, the engine is also de-stroyed; see Figure A4).

168 SCHWABE

Figure A1. Class and objects.

Figure A2. Classes.

Figure A3. Gen–spec structures.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014

Instances connections model all structural relations between class and objects. Justlike in entity relationship models, there exist (1:1, 1:n and m:n) relations. For exam-ple, there is an m:n relation between conferences and authors as authors hand in pa-pers to several (n) conferences and a conference receives several (m) contributionsby authors (see Figure A5).

Message connections stand for service requests by one class and object to another(e.g., a person can request the service “move” from a car; see Figure A6).

ORGANIZATIONAL MEMORY 169

Figure A4. Whole–part structures.

Figure A5. Instances connections.

Figure A6. Message connections.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

17:

45 1

1 O

ctob

er 2

014