To appear in Proceedings of the ACM Conference on Computer-Supported Cooperative Work (CSCW'96)Answer Garden 2: Merging Organizational Memory withCollaborative HelpMark S. AckermanDavid W. McDonaldDepartment of Information and Computer ScienceUniversity of California, IrvineIrvine, CA 92717{ackerman, dmcdonal}@ics.uci.eduhttp://www.ics.uci.edu/CORPS/ackerman.htmlABSTRACTThis research examines a collaborative solution to acommon problem, that of providing help to distributedusers. The Answer Garden 2 system provides a second-generation architecture for organizational and communitymemory applications. After describing the need for AnswerGarden 2’s functionality, we describe the architecture of thesystem and two underlying systems, the CafeConstructionKit and Collaborative Refinery. We alsopresent detailed descriptions of the collaborative help andcollaborative refining facilities in the Answer Garden 2system.KEYWORDS: computer-supported cooperative work,organizational memory, community memory, corporatememory, group memory, information refining, informationretrieval, information access, information systems, CMC,computer-mediated communications, help, collaborativehelp, CSCWINTRODUCTIONMany user communities have a problem with deliveringhelp and general assistance. Unfortunately, the user is oftenleft to sift through reams of documentation, find his waythrough mail archives, or pursue answers through trial anderror. Normally, one attempts to examine thedocumentation or other help sources, and then wanders outinto a hallway in search of friendly colleagues.The problem becomes acute, however, in distributedcommunities. We take for our example the astrophysicscommunity, although this problem exists in most scientificcommunities. In the astrophysics community, the usersmay be spread across the world, they may work inisolation, and they may have need of relatively specializedhelp. What we would like is a surrogate for this hallwaytalk. Such a solution must avoid the broadcast problem offlooding everyone's electronic mail basket with thousandsof questions. Instead, this work reports on a system tonarrow-cast a question to the appropriate others, whetherthose others are experts or colleagues.This research, then, examines a collaborative solution to acommon problem. Earlier work, a system called AnswerGarden, allowed organizations to develop databases ofcommonly asked questions that grow "organically" as newquestions arise and are answered. The subsequent AnswerGarden 2, the focus of this paper, continues this work. It isa second-generation architecture for the same designproblem, investigating some of the issues encountered infield studies of the original system. The new architectureprovides a customizable and adaptable set of softwarecomponents that allow a variety of organizational andinformational configurations. Furthermore, it offers ageneralized solution to the problem of finding help for anyinformation system. We report here on the new architectureand its responses to the context and authoring issues.The paper begins with a brief introduction to the help andmemory problems, as well as a brief overview of theoriginal Answer Garden application and its field studyresults. Answer Garden 2 is then introduced. After anexplanation of its architecture, the paper analyzes twoparticular features of Answer Garden 2. These two features,collaborative help and collaborative refining, are explainedat length. Collaborative help mechanisms provide thenecessary context for information, and collaborative refiningmechanisms provide support for authoring. The paperconcludes with a survey of related CSCW systems andsome conclusions about these design considerations.FRED’S PROBLEMFred (not his real name) is an astrophysicist at the Harvard-Smithsonian Astrophysical Observatory. He, like manyscientists, does not want to know anything about softwaresystems or his hardware. He wants to do his scientificwork, free of the multitude of computer problems that seemto get in the way.In the "old days," everyone sat around in a common room,using their computer consoles with the mini-computer. IfFred had a question, he could ask one of the half-dozen todozen colleagues and programmers sitting in the room.Everyone had to hear the answer, so the community learntfrom the problems of each individual.Now, Fred sits in his office with his workstation near hisdesk. It is quieter, but much more isolated. If he has aproblem or a question, he can look through thedocumentation or send electronic mail for help. If he sendselectronic mail, he may not get an answer from theprogrammers for some unknown period of time, or he maybe given a response that makes him feel stupid for notknowing the answer. Often he resorts to wandering throughthe hallways, looking for people who might know theanswer. He then tries various possibilities until he finds asolution or he gives up.Any community, institution, or organization of any sizeoften has a problem with answering questions in a timelymanner. Yet, solving problems and completing tasks areoften dependent on obtaining timely answers to specificquestions.Fred’s problem is the dual problem of help and of collectivememory. We will use the term collective memory todenote the common attributes of organizational,institutional, and community memory. (The term has arelated, but slightly different meaning in thehistoriographical and critical literatures, but there is nobetter term to denote memory in a range of collectivities.)Within an organization or community, individuals’information seeking requires finding the right part of thecollective memory. Typically, collective memories includeinformation repositories (e.g., information databases, filingcabinets, documents). It can also include people (e.g., otherorganizational personnel) [25]. The collective memory towhich Fred has access includes at least the documentation,the system programmers, and his colleagues. However, hemay have great trouble finding the right piece of thecollective memory that has the answer he needs. In otherwords, his access to the collective memory should beaugmented.Answer Garden and Fred’s problemPrevious work, reported in [4] and [2], considered one wayof doing this augmentation. This work revolved around asystem called Answer Garden. Field studies of its useuncovered a number of important problems in providingcollective memory and help to users such as Fred.