Relational Perspectives as Situated Visualizations of Art Collections

Fig. 1: A relational perspective on the Hausmann collection centered around a selection. Link to the prototype: https://uclab.fh-potsdam.de/hausmann

Introduction
Akin to digitization processes, cultural heritage institutions are digitizing their inventories and are looking for approaches to devise meaningful digital representations of their collections. In contrast to traditional systems that use keyword search as the main mode of accessing collection items, usually requiring familiarity with the collection, a growing body of work demonstrates how visual interfaces can support more exploratory and serendipitous modes of accessing collections (Dörk et al., 2011; Thudt et al., 2012; Walsh and Hall, 2015; Whitelaw, 2015). Interactivity is often used as a strategy to traverse the extent and complexity of information visualizations (Cockburn et al., 2008; Van Ham and Perer, 2009; Meirelles, 2013). Despite the growing variety of visualizations of cultural collections (Windhager et al., 2018), researchers like Drucker argue that many of them are transferred uncritically to the humanities, ignoring fundamental aspects of humanistic research, such as interpretation, ambiguity or uncertainty, and the specificity and situatedness of a given point of view (Drucker, 2011; Drucker, 2016). While visualizations that focus on providing one overview are suitable for the display of patterns across entire datasets (Shneiderman, 1996), they often fall short of visualizing relations and similarities at the level of individual items. Additionally, the limitation of using only one view implies a loss of information and may lead to distorted perceptions.
Referring to the relational importance of an individual’s social context, Latour et al. (2012) conceptualize the individual relations in large social networks as monads. Here, monad refers to a conceptual point of view that defines an entity through its many particular relations to all other entities of a dataset. Monadic exploration is a visualization technique for relational information spaces, in which each node can be selected as a navigation point to trigger changes in flexible layouts (Dörk et al., 2014).
Following the premise that for cultural data the individual relations can be just as important as an overview of the whole, we investigate the potential of relational perspectives in the visualization of cultural collections. We engaged in an iterative prototyping and research process in close collaboration with the
Berlinische Galerie – museum of modern art, using their
Raoul Hausmann collection as a case study (see Fig. 1).

Based on the exchanges with our collaborators, a co-design workshop and prior work on collection visualizations, our research ambitions can be summarized by these design goals:
1. Provide multiple, flexible views on the data.
2. Reveal the individuality and diversity of the artifacts.
3. Promote open exploration and serendipitous discovery.
4. Expose the temporal context of artifacts.
5. Acknowledge uncertainties and gaps in the data.

Taking a relational perspective on a collection
To examine the potential of relational perspectives in comparison to overviews, the design of the visual interface contrasts two approaches: 1) an
overview of the collection, and 2)
multiple perspective views, each centered around one artifact.

The
overview serves as a landing page providing a comprehensive representation of all artifacts arranged by media type/genre and contextualized by visualizations of the temporal and social relations. Although we are particularly interested in relational perspectives, we acknowledge the benefits of an overview as an entrance point to a collection (Shneiderman, 1996; Whitelaw, 2015). The overview consists of three connected visualizations (see Fig. 2), a vertical timeline (left), artifacts displayed by genre (center), and a people/relations diagram (right).

In the center, all artifacts are organized in small thumbnails by genre of the respective items. Hovering over an item displays a bigger thumbnail. Titles of the genres also function as a legend for the colors of the other visualization parts.

Fig. 2: Overview: a) top-left: no selection, absolute timeline format; b) top-right: person selected, absolute timeline format, hovering contextual information; c) bottom-left: year and person selected, relative timeline format, hovering an element; d) bottom-right: keyword selected, relative timeline format
In the timeline, which supports absolute and relative scales, each year inside the time frame of the collection (+undated) is visualized by a horizontally stacked bar consisting of small aligned rectangles, one for each artifact in the color of the corresponding genre, representing the total quantity of each genre per year. Small icons next to a year reveal additional contextual information. The arc diagram displays all persons and relations that are linked to artifacts within a selected time frame. Each person that is connected to one of the items is represented by a circle, whose size represents the total number of artifacts connected to the respective person in the current selection. The ordering, font size, and size of the nodes depend on the number of related artifacts. Selections of a year in the timeline (see Fig. 2b), a person in the relations diagram (see Fig. 2c), or a keyword via the search bar (see Fig. 2d) trigger highlights or activate filters.
A click on an artifact triggers a switch to the
perspective views, starting with the attribute-based perspective (see Fig. 3a). The perspective views are primarily based on individual, yet relational viewpoints anchored by a selected artifact and its unique connections with other items. The prototype offers three views, two faceted perspectives and a temporal perspective, each enabling access to a detail view (see Fig. 3d).

The core functionality of the two faceted perspectives is to display related items in order to stimulate open-ended exploration and increase the likelihood of serendipitous discoveries. While the
attribute-based perspective (see Fig. 3a) displays general metadata of the object, the
content-based perspective (see Fig. 3b) is based on manually authored and automatically extracted keywords.

Fig. 3: Perspective views: a) top-left: attribute-based, b) top-right: content-based; c) bottom-left: temporal; d) bottom-right: detail view
Both views are structured around a selected item, displayed in the center of the page along with an image and basic information. Other data attributes/keywords are displayed once above and once below the selected item; related artifacts sharing the same attribute/keyword are displayed above/below these labels, roughly ordered by their temporal distance to the selected object. All artifacts created in the same year as the selected item or in later years are displayed above the selected element, all artifacts that were created in the past are displayed below, followed by undated elements. In addition to the ordering, temporally close elements are larger than those that lie further in the past or future (see Fig. 4). Additional encodings help to identify genres (colored circles), the number of shared attributes with the selection (number of rings), duplicates in other attribute columns (lines), and uncertainty (striped texture). The selection of another artifact causes a rearrangement of the visualization based on the artifacts and attributes related to the new selection, resulting in many unique visual and semantic fingerprints.

Fig. 4: Close up of a faceted perspective view

Fig. 5: Close up of temporal perspective view
In contrast to faceted perspectives, the aim of the
temporal perspective is to induce iterative navigation from one collection item to the next along temporal vicinity by revealing precise dating of artifacts, including uncertainties (see Fig. 3c & 5). The selected element is positioned on the left side of the interface, marking the moment in time it was created. All remaining elements of the collection are represented in a scrollable vertical timeline around the selection. While a dot indicates a precise point in time and a bar represents an uncertain time interval, dating based on uncertain assumptions is marked by a texture (see Fig. 5).

Conclusion
Based on the observation that many common visualization types cannot represent the complexity and uniqueness of individual artifacts of cultural collections, we developed the idea of multiple visualizations acting as relational perspectives. Findings from a co-creation workshop, frequent exchanges with our collaborators, and a thorough literature review indicated that such individual perspectives on artifacts may offer a richer and more diverse approach to cultural collections.
To examine the potential of such an approach, we carried out a case study using the Hausmann collection with the intention to put focus on the individuality of each collection element by providing a variety of visualizations, in particular an overview of the entire collection and three relational perspective centered around individual artifacts.
Usage logs suggest that, while many visitors actively used the perspective views as exploration tools, the provision of an overview still proved to be useful for others.
The present research suggests that a deliberate consideration of a diversity of situated views holds promise for the digital exploration of cultural collections. By displaying relational perspectives of each individual artifact and by acknowledging uncertainties we propose an approach to promote exploration, user-dependent interpretations, and furthermore we emphasize the consideration of perspective in future projects.
The conceptual ideas are very much open for further development, but we hope that this research contributes to the exciting research carried out by a growing transdisciplinary community at the intersection between design and the humanities, where critical, aesthetic, and functional considerations converge.

Bibliography

Cockburn, A., Karlson, A. K. and Bederson, B. B. (2008). A review of overview+ detail, zooming, and focus+ context interfaces
. ACM Comput. Surv.,
41(1): 2:1–2:31.

Dörk, M., Carpendale, S. and Williamson, C. (2011). The information flaneur: A fresh look at information seeking.
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, pp. 1215–1224.

Dörk, M., Comber, R. and Dade-Robertson, M. (2014). Monadic exploration: seeing the whole through its parts
. Proceedings of the 32nd Annual ACM Conference on Human Factors in Computing Systems. ACM, pp. 1535–1544.

Drucker, J. (2011). Humanities approaches to graphical display.
DHQ: Digital Humanities Quarterly,
5(1): 1–21.

Drucker, J. (2016). Graphical approaches to the digital humanities
. In Schreibman, S., Siemens, R. and Unsworth, J. (eds)
A New Companion to Digital Humanities. Chichester, West Sussex, UK: Wiley-Blackwell, pp. 238–250.

Latour, B., Jensen, P., Venturini, T., Grauwin, S. and Boullier, D. (2012). ‘The whole is always smaller than its parts’–a digital test of Gabriel Tardes’ monads.
The British Journal of Sociology,
63(4): 590–615.

Meirelles, I. (2013).
Design for Information: An Introduction to the Histories, Theories, and Best Practices behind Effective Information Visualizations. Beverly, MA, USA: Rockport publishers.

Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information visualizations.
Proceedings 1996 IEEE Symposium on Visual Languages. IEEE, pp. 336–343.

Thudt, A., Hinrichs, U. and Carpendale, S. (2012). The bohemian bookshelf: supporting serendipitous book discoveries through information visualization.
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, pp. 1461–1470.

Van Ham, F. and Perer, A. (2009). ‘Search, show context, expand on demand’: supporting large graph exploration with degree-of-interest
. IEEE Transactions on Visualization and Computer Graphics,
15(6): 953–960.

Walsh, D. and Hall, M. M. (2015). Just looking around: Supporting casual users initial encounters with Digital Cultural Heritage
. Proceedings of the First International Workshop on Supporting Complex Search Tasks at ECIR 2015.

Whitelaw, M. (2015). Generous Interfaces for Digital Cultural Collections.
DHQ: Digital Humanities Quarterly,
9(1).

Windhager, F., Federico, P., Schreder, G., Glinka, K., Dörk, M., Miksch, S. and Mayr, E. (2018). Visualization of Cultural Heritage Collection Data: State of the Art and Future Challenges
. IEEE Transactions on Visualization and Computer Graphics.