How Do We Make Art of Databases?

The themes and projects presented in the master classes and the essays in this publication, Making Art of Databases, represent different methods for database and content-management use and design. This text aims to provide a short overview on teamwork in the development process, and outlines some related approaches for connecting databases and archives, inspired by research in the field of artificial intelligence and artificial life.

In database and archive-based art projects one can distinguish roughly two poles. On the one hand we find the tradition of pre-defined and domain-specific applications. The content of the database here is familiar to the maker, and the outcome often represents different ways to select the objects from the database and reorganization mechanisms. On the other hand one sees artists who are looking for ways to create dynamic environments and applications that come into existence after participant interaction. This rule or learning-based decision process can connect a wide range of (non- linguistic) parameters in a fluid and dynamic way. Once applied to the enormous information space of the internet, the latter approach easily gets shattered. A combined approach that connects the two poles, might be interesting to consider for projects where the interpretation of a wide range of interaction parameters and non-linguistic input needs to lead to some sort of coherent environment.

The design of an archive is determined by the possibilities of its underlying technology, and the system design is in turn (or should be) based on the user requirements. User requirements and system design ought to go hand-in-hand, so the functionalities of a system are rooted in the daily practice and the work methods of its user/ participant. Another aspect to be taken in consideration is the users' context: designing a system includes an understanding of the application domain. When we take the application domain into account we face a rather complicated situation in interdisciplinary practice, based on connected or intertwined disciplines and domains. Before getting into the technology of archive-system design, I will outline some of the interdisciplinary development aspects.

When we analyze the way software-based art projects are being developed and how the constellation of a team relates to the aims and outcome of a project, we come across several domain-explicit or intertwined domain issues. Examining the work process in artistic research and development (aRt&D) shows a reoccurring confusion about the terms multidisciplinary and interdisciplinary collaboration [1]. Although multidisciplinary and interdisciplinary are often used as synonyms, I prefer to retain the distinction: multi = many, and inter = between, among, mutually.

In aRt&D collaborations where various disciplines are united, we observe, on the one hand, a multidisciplinary model, where the artist directs the process and team members bring in their expertise from different backgrounds, like design, engineering etc. In this directive model the team members' functions are comparable to those of a crew in the film industry. The technology is often used here as a tool or a facility to build or produce the artistic concept. Although there can be small innovative aspects included in the production, this is not the main purpose of the collaboration. In this model the art-cultural domain is facilitated and informed by the scientific and technical disciplines.

The interdisciplinary collaboration model is based on a collective team effort, where different disciplines (from the arts, design, engineering, science etc) bring in crucial conceptual aspects and research objectives from their own field of expertise. The organization chart of interdisciplinary collaborations is often more fluid and non-hierarchical, because the collaboration process is crucial. Usually the artistic concept and the code are intertwined: the technology becomes an important element of the concept, while the concept is the drive to develop or adjust the technology. Interdisciplinary collaboration focuses on experiment and innovation rather than on a pre- defined final product. The experimentation and innovation in interdisciplinary collaboration often spring from a range of research agenda brought in by team members. The research objectives are merged into the process and the innovative outcome is therefore not domain specific. Unlike most research and development, this interdisciplinary aRt&D does not aim at problem solving or task- specific applications, but focuses on audience participation instead, through which it generates valuable new insights in mediated social interaction and cultural user experiences. From a bird's-eye view of the art and science landscape, some interesting similarities in collaboration models can be observed. In addition to the traditionally separated disciplines in the academic field, information and communication technology brought about a need for crossing traditional boundaries. Informatics (or computer science) for example can be looked upon as an umbrella for a range of scientific disciplines, and is comparable with the art and technology field, where a range of (media) art disciplines are represented. Whereas in technology-driven research and development the different disciplines represented in informatics 'inform' each other in a way comparable to the multi-disciplinary aRt&D collaboration, the recent shift towards human interaction research in technology has brought about a trend towards more interdisciplinary collaborations among a range of divergent scientific disciplines.

It could be highly beneficial to establish collaborations between the interdisciplinary aRt&D and the interdisciplinary scientific field, as cultural experiences and practices from the aRt&D field are hardly represented in the latter. However, the proposed cross-domain variant of interdisciplinary (collaboration between the interdisciplinary aRt&D and the interdisciplinary science field) turns out to be harder and less obvious than one might expect. The interdisciplinary aRt&D projects often include engineers and scientists, but scientific interdisciplinary research and development rarely includes artistic innovation. Although there is a growing number of people operating in various domains, their effort often remains invisible because it's hardly acknowledged as a valuable asset to the traditional disciplines, and nor do the domain knowledge systems or archives support this cross- domain innovation. This causes a loss of knowledge and potential cross-fertilization among the domains and disciplines. Which brings us to some fundamental issues of technology design, which are reflected in the ontology of a database and content- management system.

In the field of knowledge representation and reasoning, one can distinguish roughly two models [2] which are applied in database, metadata and content-management system design: the symbolic model and the connectionist model. These two approaches, including all their variations and branches, affect the way information is organized, annotated, stored and retrieved.

The symbolic model [3] represents concepts and content by means of a code (numbers, mathematical formulae, words, musical notation etc). The explicitness of a symbolic model has the advantage of predictability or reliability, based on pre- determined descriptions. These descriptions (eg terms and classifications) usually work with conventional vocabularies, such as the terms from domain-specific thesauri. In the symbolic approach, the content selection process is pre-defined by the maker (editor).

This symbolic approach is used for non-linear movies, where the stored media objects can be reorganized according to certain topics, themes or preferences, as we can see in the work of Lev Manovich. Manovich also investigates the limitations and borderlines of this approach - see the essay he wrote for his master class 'Metadating' the Image. The work of Rafael Lozano-Hemmer is another example of a symbolic model: for the Huge and Mobile (HUMO) master class he used this approach when images and selected sites for projections made the decorum for temporary changes in the city landscape. The connectionist model [4] is a network of processing elements (nodes) that can receive and/or send information. The connectionist model allows for the construction of temporary attractors that correspond to relations and mappings. The openness of the connectionist model makes it less discipline specific, although it needs to be kept in mind that the connectionist model needs to be 'fed' in order to learn, and during this machine-learning process unpredictable effects or connections can come about. Music Visualization, the master class lead by Joel Ryan, researches the application of new tools for scientific visualization of music according to this approach.

The main difference between the two techniques is that symbolic techniques deal with representation languages, whereas the connectionist technique 'learns' best from patterns and repetitions. When comparing the scope of the two models, it seems clear both are useful for different purposes.

Symbolist models seem most appropriate to preserve the integrity of a specific domain. The symbolic aspects of reliability and domain-specific annotations provide the tools for a useful expert archive. The database ontology most used for this type of application is built as a fixed classification system. To ensure reliability, symbolic models demand authorized maintenance by a specialist, and allow little or no user participation, let alone (co- )creation of content by unknown users, for this would disturb the reliability and intervene with the expert protocol. For expert archives, domain-specific purposes, libraries and documentation, the symbolic model provides an appropriate rigid approach. The symbolist approach is most suitable for single or multidisciplinary purposes.

Connectionist models behave probabilistic, which is especially interesting for dynamic and quickly changing environments. The connectionist model is relevant for process-based participatory and collaborative environments. The required flexibility for these applications fit an object-oriented database ontology. In artistic research and development areas there is a growing interest in interaction, based upon natural languages and non-linguistics; in this respect the physical aspect of artistic practice should be underlined, especially in these expressions which do not have a linguistic equivalent, like dance, performance, music, improvisation at large etc. The experimental aspects and the process-oriented approach, as characteristic features of an interdisciplinary collaboration, seem best aligned with the connectionist model. Non-verbal kinesthetic experiences such as Brian Massumi describes in his essay The Archive of Experience [5] are expression forms one could imagine suitable for this approach.

Yet, for many purposes, it could be most interesting to work with a combined or integrated model. In artworks one can often find a formal side, which is suitable for symbolic techniques, whereas the less tangible aspects and intuitive, spiritual or improvised parts could benefit from connectionist techniques. One can imagine cross connections between (parts of) domains which are based on linguistics, and those areas where natural or physical vocabularies are dominant. This offers opportunities for both experiencing non-verbal kinesthetic expressions and sharing these by means of language. Sher Doruff researches this mixed approach, with special emphasis on the distributed co-creation process and its embedded social and human machine-control issues, in the master class Collaborative Culture.

The combined or integrated model could be highly useful for people working in cross-domain practice. However, this requires a certain amount of openness from the systems used in the symbolic camp. A combined or integrated model could also bridge the semantic gap between the different domains, and provide a useful tool for cross-domain information and knowledge exchange. Not only could this knowledge and achievements be preserved in different domains, but these models could also function as an interface between two different realities: the artistic and (computer) scientific. As aRt&D is one of the few areas where the different technological and artistic R&D and (cultural) audience experience come together, this might be worthwhile researching in detail for future applications.

Notes

[1] More ideas on different collaboration models can be read in Human Factors in Multi and Interdisciplinary Collaborations: http://lab.v2.nl/%20home/_docs/nigten_2002_humanfactors.pdf [link dead] [back]

[2] A model in this context is usually an abstraction or simplification of (a) reality or concept constructed to explore particular aspects or properties [back]

[3] See also the paper Making a Mind vs Modeling the Brain: AI Back at a Branchpoint by Stuart Dreyfus and Hubert Dreyfus, at: http://nibbler.tk.informatik.tudarmstadt.de/public_www/arun/ Mind_Modelling_Brain.pdf [back]

[4] A technical introduction on connectionist models can be found at: http://www.indiana.edu/~gasser/Q351/connectionism1.html [back]

[5] In Joke Brouwer, Arjen Mulder (ed), Information is Alive: Art and Theory on Archiving and Retrieving Data, V2_/NAi publishers, Rotterdam 2003 [back]

© 2003 Anne Nigten / V2_