Introduction

The Reciprocal Research Network (RRN) is being co-developed by the Musqueam Indian Band, the Stó:lō Nation/Tribal Council, the U’mista Cultural Society, and the Museum of Anthropology at the University of British Columbia.  A dozen cultural institutions, national and international, are also participating in the development process. They include the Royal British Columbia Museum, Burke Museum, Glenbow Museum, Royal Ontario Museum, Canadian Museum of Civilization, McCord Museum, American Museum of Natural History, National Museum of the American Indian, National Museum of Natural History, Cambridge University Museum of Archaeology and Ethnology, and the Pitt Rivers Museum.

The four co-developers have demonstrated leadership in diverse political, scientific and historical areas. Each of the three First Nations’ institutions already has long-standing relationships with the Museum of Anthropology (MOA) or the Department of Anthropology at UBC.

The Musqueam Indian Band (Musqueam) is a leader in bringing issues of First Nations rights to national and international attention through legal processes, including groundbreaking Supreme Court cases (Guerin (1984) and Sparrow (1990)). The Museum of Anthropology is located on Musqueam traditional territory, and the community has co-curated several exhibits with MOA. Through the hard work of many band members, Musqueam has been instrumental in the fundamental shifts at MOA towards decolonizing its museum practice and engaging with originating communities (Ames, 1999; Holm and Pokotylo, 1997; Phillips, 2000).

Situated up the Fraser River from the Musqueam Indian Band, the Stó:lō Nation and Stó:lō Tribal Council (Stó:lō) are pioneers in developing research protocols and internal licensing systems for outside researchers. They have consistently welcomed researchers, including undergraduate and graduate field schools in archaeology and anthropology, to their territory, and have used these opportunities to educate a new generation of scholars in collaborative research and applied anthropology.

The U'mista Cultural Society (U’mista) was founded in 1974 to ensure the survival of all aspects of the cultural heritage of the Kwakwaka'wakw First Nations, of northern Vancouver Island. U’mista has been instrumental in bringing questions of museum ethics and policies to the forefront.  In 1921, with Canada’s infamous anti-potlatch law firmly in place, the Indian agent seized the regalia of those attending Chief Dan Cranmer’s potlatch. Coppers, masks, boxes and other treasures found their way into museums and to a few private collectors. U’mista conducted successful negotiations for the repatriation of the items illegally confiscated at Chief Cranmer’s potlatch. Along with running the U’mista Cultural Centre, the Society supports artists, researches museum collections, and works on language programs.

The Museum of Anthropology at UBC is renowned for the innovation of visible storage and its work on collaborative museology. ‘A Partnership of Peoples’, the largest grant given by the Canada Foundation for Innovation (CFI) to a social science project, was awarded to MOA and the UBC Laboratory of Archaeology (LOA) for the creation of physical and virtual research infrastructures to foster collaborative museology and the ‘… belief that originating communities should have a major voice in shaping research questions and should benefit from the new knowledge produced’ (MOA, 2000:3a). Work on the physical infrastructure has commenced, and these facilities will include a First Nations Community Suite and reconceptualized visible storage. The RRN is the virtual research infrastructure of this project. It seeks to ‘support interactive research partnerships among geographically dispersed researchers and across culturally distinct knowledge systems …’ (MOA, 2000:3a), thereby deconstructing silos of knowledge.

Through the RRN, MOA is working to move beyond current notions of collaboration. Thus, in the original CFI grant application Musqueam, Stó:lō and U’mista were named as researchers. The long-term relationships linking MOA to these communities made co-development the next logical step in redefining relationships between museums and First Nations and re-dressing issues of past power imbalances. Therefore, the RRN, while physically based at MOA, moves forward through the work of its Steering Group, composed of a member from each of the co-developers.

All work on the RRN is based on the principle of respect for the cultural values, laws and cultural knowledge of originating communities. Through this network, communities and institutions will be able to build relationships and share knowledge. The RRN development team’s intentions are to structure the RRN so that the information it contains and the way it is used respect originating communities’ cultural values, support the integration of clearer understandings of First Nations’ cultural knowledge into standard museum practices and philosophies, and further global discussions on cultural knowledge and intellectual property. In this way, the Reciprocal Research Network is being designed to facilitate a paradigm shift in the way research into cultural heritage is conducted, whereby community members will be active participants and one of the drivers of research.

The first task of the RRN Steering Group was to create, and then have their respective institutions approve, a Project Charter. Outlined in the charter are the RRN development principles. These are to:

• Respect the intellectual property framework as laid out in the Research Guidelines document.
• Build on previous or existing research projects and technical infrastructure wherever possible.
• Interface with existing digital collections and/or digital asset management systems. This will require extensive research and work on bridging data.
• Deploy commonly available standards and tools where possible and appropriate.
• Adopt, wherever possible, established best practices and internationally accepted standards. 
• Recognize digital data will be owned and managed by original repositories; the RRN will reference data and make it accessible under the terms agreed upon in the Framework Document.
• Develop the pilot projects necessary for a number of components such as research and development for collaborative tools and interfaces.
• Be flexible enough to allow for anyone with an Internet connection, with the minimal requirements as specified by the RRN, to use it.
• Acknowledge that development will be ongoing – there will not be a “finished” network since sustainability will require on-going research and development. (RRN Steering Group, 2006:15)

The scope of the RRN is detailed within the Project Charter to ensure that all requirements, as agreed upon by the co-developers, are met. This section is used to measure progress and to respond to the requests of other interest groups with a unified voice.

The scope is divided into three sections:

1. items that are in-scope and funded,
2. those that are in-scope but unfunded and
3. those that are out of scope and unfunded. For example:

In-scope and funded: Support for multi-media products such as audio, video, and maps.

In-scope but unfunded: Technology to enable viewing of 3-D images.

The ability to view 3D images was not a mandatory requirement at the time the RRN was initially proposed. However, subsequent advances in technology should be evaluated to determine how this item can be accommodated within the existing funding envelope. Many 3D maps are available through LOA, and some 3-D images are already available at MOA. It is expected more will be available from the partner institutions by the time the RRN is implemented.

Out of scope and unfunded: Providing 3-D content for the RRN (RRN Steering Group, 2006:9).

For the RRN to succeed, a software design approach that meshed with the philosophy of the project was required. The Steering Group was concerned that projects taken out of house for production sometimes fail because of their inability to work within the operational mode of the home institution. They were also aware of the rapid pace of change in software and hardware. The Steering Group saw that some of the concepts  originally seen as difficult, or too expensive in the ‘A Partnership of Peoples’ grant application, were suddenly doable, while others were becoming dated. Therefore, a decision was made to actively pursue resources available at the University of British Columbia. This included the use of the RRN as a project in an undergraduate Computer Engineering course on requirements engineering for software-intensive systems.

Additionally, the RRN software development process needed to be flexible and adaptable. Achieving the target of the RRN’s adoption as a primary research tool by users with differing backgrounds and varying degrees of computer knowledge would require on-going input from potential users during the entire development cycle. It was simply not feasible to gather all data, hand the project over to software designers and engineers, and have them create a product. Many diverse players needed to be heard, and where possible, accommodated.  Agile software development philosophy has proven to be the best ‘fit’ because of its focus on collaboration, interaction, and quick response to change (www.agilemanifesto.org).

Scott W. Ambler has recently outlined 16 components for agile design philosophies; those most applying to the RRN software development philosophy are:

1. Use emergent technologies,
2. Use test-driven development,
3. Designers should also code,
4. Feedback is your friend and
5. Iterate, iterate, iterate (Ambler, 2007).

The purpose of the Reciprocal Research Network is to enable geographically dispersed users to access information from multiple institutions through a single interface, and to facilitate relationship building across user groups through the provision of a collaborative research environment. Ensuring uptake is possible only through a design process driven by feedback from potential users. Individual interviews, focus groups, testing, and an on-line form provide on-going feedback to the team. These are dealt with at daily development team meetings and bi-monthly Steering Group meetings.

The RRN faces several challenges. One lies in developing a network capable of not only accessing but also combining the diverse collection management systems of all participating communities and institutions.  Another is the requirement that the user interface support deep, comprehensive search explorations and yet be intuitive for users with little experience with computers, databases, and repository systems. 

The System Architecture

A difficult problem in supporting research with museum-held data is to create and maintain a comprehensive view of objects and archives spanning a variety of different museums, access models, collection management systems (CMS), standards for terminology and data management. A plan has been developed to meet some, if not all, of these challenges.

To understand the challenge simply of gathering the data for the RRN collection, consider the sources. Within the RRN, there are four co-developers, each with a collection and database, and twelve partner institutions, each with a CMS, only part of which is relevant to the RRN. Within that set, there are a variety of different SQL and non-SQL database systems to integrate. Additionally, a key constraint in developing the system was the realization that museums providing data must be able to do so at virtually no cost. Clearly the model for continuing integration is critical.

To achieve this, a federated model in which a central RRN server is provided with data by mediators at each participating museum is proposed. Each mediator is responsible for monitoring the local CMS, mapping changes into the RRN’s data model, and providing a local view of how each partner museum’s collection appears and is being used. A diagram of this model is shown in Fig 1, with each of three partner institution’s CMS’s maintaining their connection to the RRN via a mediator. The mediators are each responsible for:

• gathering and updating that portion of the host CMS that is relevant and to be shared with the RRN;
• translating the data from its local form into the RRN data structure; and
• gathering usage statistics and a feed of relationships established with the CMS’s data back to interested parties in the partner institution.

In this way, institutions can join at very low cost, and be provided with a valuable tool for relating their collection to other similar collections. Moreover, due to the reciprocal nature of the network, they will be provided with feedback about their collection data.

Some institutions are XML ready, in the sense that they are able to dump their collection data to a transportable XML format. For these institutions, their XML standards can be mapped to the RRN data standards. This can be done either by the museum’s regularly writing an update file, or by an update request sent by one of the RRN servers.

Fig 1: RRN System Architecture and Federation Model

The means by which each mediator gathers and updates local data to the RRN collection must be matched to that collection’s CMS and data standards. Initially, we are manually developing customized data gatherers and creating the semantic and structural data mapping (crosswalk) for each collection. Eventually, with a library of existing mediators, one of the existing mediator implementations will be selected and minor adaptations made for each new partner.

These crosswalks depend on the target data: the data structure and semantics relied on by the RRN itself. To that end, the data structure has been designed hopefully to provide a neutral, efficient “common ground” data structure for these systems. After considering the possibility of building on an RDF (Resource Description Framework) database and relying on the CIDOC CRM (http://cidoc.ics.forth.gr), this possibility was rejected because of the difficulty of building a fast, interactive  application on top of an RDF database, and the additional difficulties of maintaining an accurate semantic mapping from legacy data to such a complex target with well-defined semantics. Instead, we decided to design a simple, adaptable data structure that would allow for efficient implementation and adaptable queries on a traditional RDBMS (Relational Database Management System). The ability to participate in CRM-mediated data systems is potentially important, so care was taken to avoid any obvious incompatibilities in the design; we intend to develop CRM-based RDF feeds of RRN collections in the future.

The data design is based on the Entity Relationship diagram shown in Fig 2: Basic RRN Data Structure. At the center of the of the diagram and the data structure is the Item, a “collection item” defined as “an object (either physical or digital) subject to cataloguing as a separate artifact in a collection. It is the primary locus of information about specific items in a collection.” Each Item has a name, catalogue number and Site (a record identifying its source), and then a set of associations with generic holders of certain kinds of related information (Notes, Categories, DigitalObjects and Events).

Fig 2: Basic RRN Data Structure

Notes: A set of textual annotations on this Item. Each Note has a type that identifies its relation with the Item (e.g. “description”), a block of text, and an associated Event that records an attribution of the Note.

Categories: A set of Category elements that represent all of the various categorizations of this Item. Each Category is related to an Item by an ItemCategory that describes the relation (e.g. “made-of”, “kind-of”) and links to an Event that records an attribution of this categorization (i.e. who made the categorization and when). This intervening association also allows the Item-Category relation to be many-to-many.

DigitalObjects: A set of digital objects related to this item. Each DigitalObject describes itself in terms of a title, description, MIME type and URL where the object may be accessed. Items and DigitalObjects are related by an ItemDigitalObject that describes this relation (e.g. “self” for Items that are digital objects, or “depiction” for images of the Item). Again, many DigitalObjects can be related to any Item and vice-versa (e.g.,. a photograph that shows a number of collection items on site).

Events: An event associated with the Item. Each Event has a type (e.g. “creation” or “collection”), start and end date, and optional associated People and Locations. These are used for all attributions, including provenance, and for recording the dates of any actions taken with respect to the Item (e.g. when it was entered into the CMS and by whom;  when it was restored or repaired).

It should now be clear that this structure is compatible with CIDOC-CRM. Each of the related attributes of a collection Item is defined as a generic holder of a type of information (e.g. Notes hold text, DigitalObjects hold media, Events hold time and place) and describes its relation to the Item by a textual type. This is essentially how RDF (on which CIDOC-CRM is based) works, and similar structures to this are used when storing RDF data in relational database systems. Therefore, one could build a fully functional CMS on top of this model. In essence, this was one of the fundamental constraints in the development of the model - that it could naturally represent any of the CMS data being imported.

This model easily accommodates multiple categorizations of the collection items. Knowing that a variety of categorization standards exist for simply describing First Nations communities (e.g. by language or nation) and that the partner institutions subscribe to different standards, it becomes critical that these categories be represented not only as they are within the partner museums’ CMSes, but also that translations are provided to other standards. This will allow convenient access to users familiar with either standard. Therefore, Categories are organized into Vocabularies, each of which contains a hierarchy of Category elements and some restrictions as to what ItemCategory relations they may be used for. Multiple Vocabularies may represent the same attributes of an item, and we may be able to derive mappings that automatically relate categorizations in one Vocabulary to those in another (e.g. two different material classification models). The RRN data model can support multiple Category associations for the same attribute (e.g. an Item may be composed of multiple materials). Putting these two capabilities together, once there is a categorization of an attribute with a Category that can be mapped to a different vocabulary, we can simply associate the Item with the equivalent Category in the other Vocabulary and support users familiar with both Vocabularies simultaneously.

User Interface

In designing the user interface to allow researchers to explore and manage their own use of the aggregated collection data, there were a number of significant and sometimes conflicting imperatives:

1. Clearly, a traditional keyword-based search model is necessary in today’s world, if only because it is the most familiar means of accessing any body of information in the post-Google world.
2. Since the goal of the entire project was to consolidate a variety of research collections and make them available to researchers, it was important that complex, research-specific queries be supported, and that the results allow researchers to compare, analyse and find gaps and inconsistencies, not only in individual items, but also in the entire set of items returned.
3. Given that the targeted research community includes a wide variety of formal education levels (from university faculty to youth) and of familiarity with computers, it is necessary to support these complex research-specific queries, especially for people with little or no experience with either computers or formal vocabularies.

To accomplish this, the interface was approached from the point-of-view advocated by Marchionini (2006): he defines the principle of “exploratory search”. In this model, we are searching not merely for “a few most-representative examples” (the criteria that lead  to the Google PageRank algorithm [Brin & Page, 1998]), but for an entire, complete collection of related information that is further contextualized by user interpretation and various relations to other items and other collections. A successful method for achieving this goal has been the hierarchical, faceted data browsing approach with embedded pivot points pioneered in Flamenco (English et. al. 2002).

In this model, a user starts with an interface to the entire collection and a high-level description of the various “facets” available for filtering the current collection. For example, when shopping for electronics, a “product type” facet might hierarchically  organize all products sold by their department, category, and finally by name. A parallel, “manufacturer” facet might simply have a non-hierarchical list of manufacturers in alphabetical order. Finally, a “price” facet would typically be divided into price categories such as “< $10” or “between $10 and $20”. The interface allows the user to select one of these facets and choose a restriction on that facet. The current list of products shown is immediately updated to include only those products conforming to the selected criteria. Furthermore, the range of values available in the parallel facets is restricted to only those values available within the newly restricted set. The simultaneous selection of restrictions in parallel facets has the desired effect of simultaneously restricting by all of the associated criteria. The result is a much more exploratory interface that allows even inexperienced users to build complex queries interactively.

In implementing this basic principle in the RRN, the data facets have provisionally been organized into four groups: “Who”, “What”, “When” and “Where”. Each of these interrogatives groups together those attributes of an item (including associations) that have a particular data type and provides a set of questions that may be asked of the current set of items.

• Who implies an interest in the people associated with the item and presents a set of facets derived from Events (e.g. “Who created the item?” or “Who collected the item?”)
• What implies an interest in inherent properties of the item, how it is described or, in our data terms, its various associated Categories (e.g. “What kind of item is it?” or “What is the item made of?”)
• When implies an interest in the timeline of the item and again queries the Events associated with the item (e.g. “When was the item created?” and “When was the item collected?”)
• Where implies an interest in geographical data and once again queries the Events associated with the item (e.g. “Where was the item collected?”)

Opening the “Who” tab in the interface presents the user with a list of questions that may be asked of the current set. Selecting one of these questions sends an Ajax call to the server to gather the range of values that answer the question within the current set and presents these values to the user in a fashion that 1) shows the range, and 2) allows the user to select a specific answer or range of answers representing  restricted interest. In each case, an effort is made to display the range of answers in a natural fashion, clearly exposing the number of items available should the user select that restriction.

An example is shown in Figure 3: RRN Query Formation: here the collection has already been geographically restricted through the selection of  “Where” and “Creation Location” and then “Kingcome Inlet.” Opening the “What” tab and choosing “Materials” creates a list of all the materials used to make the Kingcome Inlet items. These are shown as single-word “top-level” categories, alphabetically ordered and presented using a “tagcloud”-style (words are larger and bolder when they represent more items). In the example, the mouse hovers over the word “fibre.” A tool tip informs the user that there are “103 items” from Kingcome Inlet in this category. If the user clicks on “fibre”, the screen immediately updates to Figure 4: RRN Query Specialization. Note that the sentence describing the current restriction has been updated to include “made of fibre”; the header of the Materials section now shows “Materials " fibre”; and alternative subcategories of fibre, namely “cotton fibre” and “wool fibre,” are listed. Unseen in the figure, the list of items below this query block has also been updated to display only those 103 items.

Fig 3: RRN Query Formation

Fig 4: RRN Query Specialization

This textual presentation of alternatives works well for many queries, but falls short when displaying or selecting in the “Where” and “When” tabs. When displaying a timeline for a collection, we display a linear, logarithmically-scaled histogram of item counts within a range of dates. When selecting “Where” an event occurred, we can display either a place name hierarchy (from standardized gazetteers) or an experimental Google maps interface that can select regions from geographical coordinates.

We have demonstrated this interface to a wide variety of users and have been gratified by the results. For example, some First Nation elders with little computer background were easily able to find sets of items of interest to them. They began immediately to provide feedback as to the accuracy of the collection data. A formal usability study investigating presentation models, query labels (e.g. should we use the “question style” discussed above, or the simple headers shown in the figures) and other features will be conducted in conjunction with less formal methods for gathering feedback.

One on the most important steps in engaging with potential RRN users will come with the testing of the collaborative tools. These include the ability to create collections of interest, annotate them, organize them, and engage in a dialogue related to the items. In order to accomplish this, we plan to integrate a separate, open-source social note-taking and annotation system called Opntag (Iverson et. al. 2008, http://opntag.net). This system provides social bookmarking, note-taking, tag-based organization, and awareness functionality with a deeply integrated privacy-control mechanism that supports an application class we are calling Social/Personal Information Management (SPIM). It has been deployed and used for a number of years by a group of users, and though it is still evolving and maturing, people are integrating this new tool into their day-to-day information processing activities. Once integration is complete, users will be able to annotate items or item sets derived from faceted queries, engage in conversations about those items, and see interest markers and tags applied to items by other users. They will also be able to set privacy levels and invite others to participate with them in research projects. Furthermore, the partner institutions will be able to gather records on this use and gauge interest in the items from their collections. We plan to complete a preliminary integration of these two major components in the summer of 2008.

Conclusions

The RRN pilot is available for testing at www.rrndevelopment.org. It exists as a Web-based sandbox and currently has access to three very different databases:

1. a MimsyXG database of anthropological collections at the Museum of Anthropology,
2. a FilemakerPro database of archaeological collections at the UBC Laboratory of Archaeology and
3. an Access database of collections at the U’mista Cultural Centre. 

It features an easy-to-use interface with faceted search capabilities and pivot points. Users have their own profiles and can create/share collections. Simple collaborative tools are being tested at this time. Community members, elders, experts in fields related to the RRN, and volunteers are regularly testing the RRN pilot.

The co-development process so far has effectively modeled the vision for the RRN.  When the Steering Group started, separate relationships with MOA were the main links between the four co-developers. Through this project, the ties between the co-developers have been strengthened, and the working relationships mirror the proposed functioning of the RRN.

Feedback from multiple user groups, with very different objectives and perceived ways of using the system, is critical to this project. In many cases, it has forced a re-evaluation of priorities. When this project started, there were no equivalents for people to explore. The sandbox approach has been transformative in assisting potential users to see the system and imagine how they might use it. Workshops and focus groups inevitably lead to discussions of the RRN’s research potential and the questions that can be posed and investigated. These range from explorations of stylistic change of an artist over her career, to discussions of the meaning of culturally sensitive, to opportunities to change the way exhibits are created.

The provisional rollout date for the Reciprocal Research Network is October 2009.

Acknowledgements

Funding for this project was awarded to the Museum of Anthropology in 2002 with a 17.2 million dollar grant from the Canada Foundation for Innovation followed by a matching grant from the British Columbia Knowledge Development Fund (2005). Additional support has been received from Canadian Heritage, the Social Sciences and Humanities Research Council of Canada, the Vancouver Foundation, the University of British Columbia, the UBC Faculty of Arts, and MOA.

References

Ambler, S.W. (2007). Agile Design Philosophies. March 03, 2007 www.agilemodeling.com/essays/agileDesign.htm Consulted January 31, 2008.

Ames, M.M. (1999). “How to Decorate a House: The Re-negotiation of Cultural Representations at the University of British Columbia’s Museum of Anthropology”. Museum Anthropology 22(3):41-51.

Brin, S. and L. Page (1998). “The anatomy of a large-scale hypertextual Web search engine”. Computer Networks and ISDN Systems, 30(1-7): 107-117.

English, J., M. Hearst, R. Sinha, K. Swearinghan and K-P. Yee, (2002). “Hierarchical faceted data in site search interfaces”. In Proceedings of CHI, 2002.

Holm, M. and D. Pokotylo, (1997). “From Policy to Practice: A Case Study in Collaborative Exhibits with First Nations”. Canadian Journal of Archaeology 21(1)1-11.

Iverson, L., M. Razavi, and V. Mirzaee (2008). “Personal and Social Information Management with Opntag”. Submitted to International Conference on Enterprise Information Systems, June 2008 (in review).

Marchionini, G. (2006). “Exploratory search: from finding to understanding”. Communications of the ACM 49(4): April 2006.

Museum of Anthropology (2001). A Partnership of Peoples: A New Infrastructure for Collaborative Research at the University of British Columbia’s Museum of Anthropology. Grant application submitted to the Canada Foundation for Innovation in 2001.

Phillips, R. B. (2000). “APEC at the Museum of Anthropology: The Politics of Site and the Poetics of Sight Bite”. Ethnos 65(2):172-194.

RRN Steering Group. (2006). Reciprocal Research Network: Project Charter. Approved by the co-developers in October 2006.