Collecting bits and pieces: What happens when a crowd engages with a digital museum collection?

(1)

Collecting Bits and Pieces

What happens when a crowd engages

with a digital museum collection?

Master Thesis Heritage Studies - Cultural Information Science M.D.P. (Marco) Roling – University of Amsterdam - 11149736 Dr. ir. J. Kamps, Dr. K. Beelen (supervisors) 30-04-2019 (finish date)

(2)

2

List of Figures

Figure 1 Example webpages of Rijksstudio collection exhibitions and individual artworks ... 10

Figure 2 Inside the Rijksmuseum depot (photo: Rijksmuseum) ... 13

Figure 3 Digital collection composition showing visibility and use in Rijksstudio sets ... 15

Figure 4 Increase in number of artworks discovered in depot versus museum collection .... 16

Figure 5 Accumulated discovery percentage of depot versus museum collection ... 17

Figure 6 Size of the generated user sets as resulting from mastermatcher ... 21

Figure 7 Popular choices made by visitors using the mastermatcher ... 21

Figure 8 Popularity ranking for the top-20 individual artworks ... 22

Figure 9 List of top-20 popular artworks in Rijksstudio excluding the mastermatcher effect 23 Figure 10 Popularity ranking versus occurrences in Rijksmuseum sets ... 24

Figure 11 Popularity ranking versus occurrences in app tours ... 24

Figure 12 Number of items in personal sets grouped by object type (logarithmic scale) ... 25

Figure 13 Number of ‘likes’ in personal exhibitions in Rijksstudio ... 26

Figure 14 Common words in titles of works in non-curated Rijksstudio sets ... 27

Figure 15 Common words in titles of single inclusion artworks in user sets ... 27

Figure 16 Inclusion of artworks grouped by the century of creation ... 28

Figure 17 Number of sets per number of users (logarithmic scale) ... 34

Figure 18 Number of included collection items (artworks) per number of users ... 34

Figure 19 Number of users per activity duration in months ... 35

Figure 20 Number of users per number of cut-outs ... 35

Figure 21 Example cut-out of ‘De Nachtwacht’ (image by Rijksstudio) ... 36

Figure 22 Number of cut-outs per cropping percentage ... 37

Figure 23 Number of cut-outs per displacement value versus cropping ... 37

Figure 24 Heat map of ‘De Nachtwacht’ by Rembrandt van Rijn ... 38

Figure 25 Number of downloads per number of users ... 39

Figure 26 Number of downloads in time ... 39

Figure 27 User groups with sets and/or downloads ... 39

Figure 28 Activity based subdivision of user groups ... 41

Figure 29 User groups comparison of user behaviour characteristics ... 41

Figure 30 A- and H-group comparison in relation to popularity ranking ... 43

Figure 31 A- and H-group comparison in size and contribution to collection visibility ... 44

All figures and images are created by the author unless stated otherwise.

Images by the Rijksmuseum are freely downloadable from their website, as well as open data.

With special thanks to: Jaap Kamps, Maarten Heerlien, Monika Cunnington, my dear parents.

Dutch:

Stukjes bij beetjes verzamelen

Wat gebeurt er wanneer een menigte zich bezighoudt met een digitale museumcollectie? The following creative common licence applies to this document.

(4)

4

1. Introduction

Museums are in some ways like our homes. They are tastefully decorated buildings with different rooms with different purposes, nicely designed and filled with carefully selected objects for use or just for display. But they also have overstuffed closets, limited space and outdated possessions. Some have an attic and a basement where rarely anybody enters or looks around. What could happen if you would invite a museum curator to your home to have a look at your belongings and ask to assess what you have? Maybe you instantly debate on the meaning of your personal items, what to show to others or better hide away, where and why. The other way round, what would to happen when a crowd is introduced to the museum collection and invited to interact with it? This is what engagement is all about, the desire to interact with tangible objects carefully collected around us.

1.1. Background

Over the past centuries, Western-European countries developed large, public institutions designed to house many objects of great cultural and historical importance and we started to call them museums. Later on we even defined a museum as ‘an organized and permanent non-profit institution, essentially educational or aesthetic in purpose, with professional staff which owns and utilizes tangible objects, cares for them and exhibits them to the public on some regular schedule.’1_{But for a long time, the leaders of major museums thought the size} of their collections mattered more than the ability of visitors to actually see all the objects and learn from them. Collections grew exponentially. It appears that the large national museums through time have been hoarding on an epic scale, constructing common history and

heritage for their audiences. On average, less than 8% of museum collections in the Netherlands is currently on display, and only about 20% was shown in public exhibitions in the last ten years. So most of the artworks are actually hidden away in depots and are never seen nor studied.2_{Radical thoughts pop up saying that museums should go back to be} modest, small, free, personal and more individualistic as this would be the only way to ever tell stories on a human scale.3_{It shows both ends of the spectrum and immediately urges to} somehow merge them. The cost of maintaining a large collection is increasing and the act of deaccessioning and disposal is occasionally opted as one of the solutions.4_{Maybe some} artworks could be sold in order to finance the buying of other works that actually contribute to the collection. But these suggestions are causing a lot of debate on ethics, art historic value and museum management.5_{A first step could be for a museum to start grading their} collection, to assess what part is without doubt for the keeping and what part may be not.6 Detailed regulations and guidelines in the Netherlands for deaccessioning and disposal intend to support transparency as much as possible for all stakeholders but also turn it into a very slow process.7_{Most of the museums avoid the discussion and will keep a large}

collection and managing directors will defend this with quotes like: ‘Different generations value different things, and so we should be the collective memory and not be arrogant in thinking that we can dispose of part of our collection, because we are the memory of this

1_{Falk Dierking 2011, p69}

2_{Volkskrant February 19, 2016} 3_{Masumiyet Muzesi 2014} 4_{the Federalist August 1, 2016} 5_{NPR August 11, 2014}

6_{NYT March 11, 2019}

(5)

5

country.’, or ‘Our collection was formed by donations under the condition to keep it at the museum.’ 8

It is easy to understand that a large collection leads to high costs of housing and conservation. On the other hand, the revenues and sponsorships remain insufficient and some museums will struggle for their existence, although museums are still very popular and in general visitor numbers are growing despite economic crises. The urge for an increase in the number of visitors is high and competition with other museums almost tangible. Not only small regional museums are aware of their costs and benefits, but also some of the largest national ones. For example, the Metropolitan Museum of Art has a deficit of millions of dollars, despite six million visitors a year. The result were staff cuts and a reduction of special exhibitions in 2016. On top of that, even the CEO had to resign.9_{The Amsterdam} Scheepvaart Museum was closed for some years during a large and costly renovation, and at the time also government support for museums was cut back only to be provided again later on under new conditions.10_{In the last few years, museums in the Netherlands have} been forced to find new and creative ways to generate income in order to be less reliable on government funding.11

A museum has to connect to people, educate, inspire and entertain them, and to supply content for reuse in commerce and innovation. Today more and more people are able to go to a museum, more than ever before in history. They have a critical view and question the museum as an authority. The way for the museum to remain connected to the public is to be open and authentic and to keep encouraging interaction and engagement using its unique collection.12_{Maintaining a sizeable museum collection and managing it, feeling the need for} more visitors and generating new income, defining the meaning of the museum for the present and future – facing all these challenges museums are encouraged to rethink their objectives and discuss what the museum of tomorrow should be like. A clear focus and a defined branding are essential to be able to develop a vision for the museum and a strategy to operationalize it.13_{Museums appear to be in a constant transition process nowadays,} where the largest museums seem to be the early adaptors because they are more publicly exposed, having many stakeholders and a large and loyal base of supporters and

sponsors.14_{For organisations opening up their metadata, advantages to the museum} mission have been identified. These include an increase in their relevance to digital society, the fulfilment of their public mission to open up access to the collective heritage, and the value of opening up access to new users, who are prompted to engage with the object in its digital form and subsequently with its real-world source.15_{Major potential benefits of open} data (both images and metadata) are being identified. A few are mentioned here, such as the increase of relevance in today’s digital society, the increase of opportunities that users have to see the museum collection, discoverability of the collection through open data, linked data and semantic data initiatives driving traffic to the museum website, and innovative ways to interact with and relate to customers.16

8_{Volkskrant February 19, 2016} 9_{New York Times February 28}, ₂₀₁₇

10_{Dutch Raad van Cultuur 2016, BIS 2017-2020}

11_{Museum Vereniging, Musea voor Mensen 2013}_{p12. In 2013 about sixty percent of museums in the}

Netherlands still partially relied on public government sponsorship.

12_{Museum Vereniging, Musea voor Mensen 2013}_p17 13_{Museum Vereniging, Musea voor Mensen 2013}_p22 14_{Museum Vereniging, Musea voor Mensen 2013}_p14

15_{Verwayen et.al 2011, p4}

(6)

6

The large museums have already put much effort into the development of new websites, online platforms and the use of social media networking. With increased visitor numbers on these channels, around half of museums already encounter an increase in the number of people visiting the institution itself.17_{When visitors have a satisfying experience, they will} recommend the museum to others and return in future. The museum experience is crucial to visitors for acceptance and appreciation, online or on-site or even simultaneously. But each experience is different and difficult to grasp. In the ‘interactive experience model’ Falk and Dierking (2011) show that each museum experience is filtered by personal context, mediated by social context and embedded in physical context. Museum curators should be aware of these contexts when creating displays, exhibitions and events in order to be effective. The mutual influence of the contexts makes each experience unique as every person is unique. It results also in a difference between a potential museum visit and the actual one.18_The challenge could be to minimize this difference for each visit, but it is impossible to know each visitor personally and to assess how this can be achieved. The dilemma can be avoided altogether by just giving visitors full access to a museum collection and see how they engage with it.

Providing full access is not feasible for the physical museum on-site because of practical reasons like the size of the full collection, spacial limitations and the condition of artworks and artefacts. But when high quality digital images and up-to-date metadata of the majority of collection items have been created, the option to provide free open access to the full museum collection can be discussed. An online digital museum can be built for a global community that has the liberty to look around and interact with all the available collection items, not only the popular ones already on display in the physical museum. But inviting visitors to a digital museum collection and the need to provide them with a satisfying experience immediately raises the question of how to find out in the first place what visitors actually want to look at in the collection, and how they express their appreciation for what they discover.

Not realizing enough that visitors actually create their own experiences, traditional museum exhibitions often seem to be designed by professional curators under the assumption that visitors will stop, look, and absorb all the information as presented in a sequential way. Research done on museum visitors illuminates the flaws in this assumption. Only when a teacher or staff member is there to guide them, do all visitors follow the

designed sequence. Left on their own, some will not visit exhibits as intended, viewing artworks in a different order or even skipping parts of the exhibition.19_{Without contextual} information and background stories, visitors will look at artworks and wonder: ‘What is it?’ and ‘How old is it?’ Rarely do they wonder: ‘Why is this painting a landmark in the history of art?’ It is argued that most visitors, whether adults or children, deal with exhibits on a concrete level, rather than on an abstract level.20_{This leads to questioning the role of a} curator. Some say that the term curator is increasingly misused and it is abhorrent to claim that sitting behind a computer and passively clicking on appealing images is a form of curation.21_{According to the American Alliance of Museums (AAM), a curator traditionally} would be highly knowledgeable, experienced, and educated in a discipline relevant to the museum’s purpose or mission and serving the public good. Curatorial work is multifaceted

17_{Axiell 2016, p4}

18_{Falk Dierking 2011, p14-15}

(7)

7

and a curator can have different backgrounds.22_{Many people will be thankful for the} curator’s work in managing collections and will appreciate guidance in art historical significance and cultural heritage background. But at the same time, the curator is biased and can be disconnected too much from the public. Another approach would be to see the curator as a collection broker, ready to collaborate. The curator of tomorrow can have an even greater impact by becoming a curator of information in the public domain, and an expert communicator and interpreter, stimulating interest and helping audiences navigate to the information sources that satisfy their curiosity.23_{This opens the doorway for the crowd to} participate in some way. Maybe a diverse crowd can be just as ‘wise’ as the trained experts, or maybe they can complement each other. According to Maris et.al. (2013), designing outreach is no longer a matter of drawing first and then developing exhibitions and event; it needs to be a continuous process, engaging the public and responding sincerely to their expected and unexpected reactions.24

In order for the museum to be responsive, we need to measure and understand the size and nature of the crowd reaction to the online museum collection. Interactions are a crucial component in the architecture of a cultural heritage information system.25_{When providing} visitors with some tools to interact with the collection, potentially we can study online visits and observe what visitors like to see and experience. We may be able to find out if they select and collect artworks and look at them differently from museum curators. Providing tools with an online collection is in line with the idea that the outcome of the interactions between the museum and the public can no longer be programmed. Online spaces are ‘living’ and the museum should enable participation instead of scripting it. Designing a museum where close interaction and collaboration with the public can take place is like ‘designing without a product’.26_{It is in line with the development of the web with social media} and crowdsourcing platforms where people cannot be seen as only consumers anymore, but also as producers, thus naming them so called ‘prosumers’.

According to Fois (2015), heritage institutions should invest time and capital in appointing in-house digital curators who can create experiences with the user in mind, designing online platforms with the help of the audience to ensure that these are effective and successful.27 Also Maris et. al. (2013) argue that the museum need to transform from an educational institution to one in which joint learning with the public is promoted, from collector to co-producer of life stories, from authority to host. The boundary between online and on-site slowly disappears as the majority nowadays bring a mobile device to the museum, fusing the on-site experience with online data and tools. Most visitors use their devices for taking photos and sharing their experience on social media, followed by people searching for information about an artist or artwork. Only a small proportion of the visitors use their mobile devices with museum apps to seek guidance, to learn and to obtain context.28_{Opportunities} lie ahead to turn them into devices that really support visits and enable creativity, prolonging the stay at the museum and increasing fulfilment. Transforming the museum into a

communal area can be done but only through constant and intense collaboration between 22_{AAM 2009, p3} 23_{Proctor 2010, p38} 24_{Maris et.al. (2013), p251} 25_{Stiller Petras 2015, p155} 26_{Maris et.al. 2013, p252, p257} 27_{Fois 2015, p292} 28_{Axiell 2016, p7}

(8)

8

museum managers, employees, curators, volunteers, visitors and other stakeholders together.29

In an attempt to improve and increase connection with museum visitors, understanding their motivation to visit is not easy as they cannot be interviewed constantly about their objectives and experiences. According to Clari (2012), we can at least assume that there are no ‘general visitors’, people who are just browsing through the collection, and there is no ‘audience’, but simply people who use the museum (or the web for that matter) for their own purposes. Hence, focusing on finding out who these people are is of interest but targeting them based on demographic data could be pointless. Instead, it might make more sense to pay attention to what visitors actually do and the context in which they do it.30_{The behaviour} and interaction result in a new type of artefact and experience. This new type is a

combination of different elements; it includes the original object itself with its institutionally validated meaning, but transformed from museum object to ‘social object’, and it also becomes enriched with the traces of the interaction which has taken place around it.31

1.2. Research questions

In the previous paragraph of this chapter we have discussed that many museums have acquired large collections of artworks and artefacts over time, and that they can display only a small portion on-site. It is not easy to assess what part of the collection to promote and museum curators try their best to assist in presenting a relevant part of the collection, educating the public and encouraging engagement. Museums are challenged in a changing cultural and economic landscape, facing an increasingly critical audience, working to keep generating income and to continuously attract visitors. Reconsidering the museum meaning and rethinking what the museum of tomorrow should be like has become a vitally important part of the museum vision and strategy. Building an online presence will lead to more

visitors, but as it is impossible to know each visitor personally, the dilemma of how to ensure a fulfilling museum experience for each can be avoided by just giving them full access to a museum collection and see how they engage with it. Responding sincerely to their expected and unexpected reactions requires to measure the behaviour of the crowd and analyse its dynamics.

To support the museum in its existential mission, the main research question of this exploratory study is therefore: ‘What happens when a crowd engages with a digital museum collection?’ This is a very general question, and we have to try and specify examples of interaction with a digital museum collection in particular that can represent engagement. When we study crowd dynamics, we have to try and understand the composition of the crowd that we are studying and that is expected to represent the museum visitors on-site and online. In order to support the main question, three sub questions are formulated here that can be researched with concrete data.

We have discussed in the previous section that opening up a museum collection by publishing open data (images and metadata) and providing open access has potential benefits. But these benefits can only become a reality when we understand the part of the collection that is actually being discovered by the crowd and the rate at which it is being discovered. Is the large and formerly ‘hidden’ depot collection seen at all by visitors? This

29_{Maris et.al. 2013, p254}

30_{Clari 2012, p37-38}

(9)

9

aspect of crowd behaviour is the topic of: ‘How is the crowd discovering the digital collection?’

One step further into the crowd dynamics is to try and find out what artworks the crowd actually appreciates when using the collection. A fulfilling museum experience is important for the museum to stay connected with visitors and measuring the appreciation of the collection can be regarded as an indication of this experience. We ask ourselves: ‘How popular is the digital collection?’

Finally, speaking of a crowd engaged with a digital collection, we need to discuss the crowd itself in more detail. A large group of visitors cannot be regarded as homogeneous. We can look at the results of their engagement and interaction with a digital collection to assess the composition of this group. Looking for similarities between crowd members based on their behaviour, we can ask ourselves: ‘How do we get to know the crowd?’

1.3. Approach

There are many museums, many digital collections and many implementations of crowd engagement. It is not the objective of this research to compare them all in detail and to conclude on the best way how to understand the museum crowd. This would require an enormous effort and given the time and resource limitations is not feasible for the research as presented here. We can, however, study one of the leading examples of a published digital museum collection and its tools for user interaction. For this research the Dutch Rijksmuseum with its Rijksstudio website has been chosen because it has a long history of publishing a very large digital museum collection online, with open access and free use of the digital images and metadata, and moreover with recorded data on user interaction.32

The decision by the Rijksmuseum to start showing the collection in full to the general public, not only the objects on display in the museum but also the majority of the ones that are stored in depots, was made in 2011. It followed the large scale and high resolution photography of the collection that had taken place for more than a decade. Development of a new website was done during the years of ongoing restoration of the Rijksmuseum

building. The launch of the new website in October 2012 was followed by the grand opening of the building in April 2013. Besides a brand new design, dominated by high resolution images of the digitized museum collection, the website started a section called Rijksstudio, presenting the collection and allowing online visitors to register and create their own personal exhibitions along the curated ones of the museum. The Rijksmuseum has implemented an open data interface on their website and a described API (Application Programming Interface) that enables anyone with programming skills to freely access and download data of the digital collection and user data of the personal exhibitions. This data is constantly changing and therefore a snapshot was taken in March 2018 for this research. The downloaded data covers five years in which the digital collection had grown immensely plus the personal exhibitions.

The downloaded data contained the metadata of 626,441 artworks, identified 143,427 registered users within the data of 249,275 different personal exhibition collections. Collections items (also referred to in this research as ‘artworks’) were included 3,543,413 times in these personal exhibitions. Although the number of registered users is substantial, compared to the millions of on-site visitors to the physical museum and the online visitors to the website and its extended online outreach (platforms like Wikipedia, Europeana, etc.), the group of registered users is relatively small. On the other hand, the recorded user data can

(10)

10

be regarded as high quality, well-structured and clearly defined. The data set provides excellent opportunities to operationalize the research questions as stated before.

When an artwork is shown online on the website there are several options for a user to respond and interact. Most prominent is the option to save the work (or part of the work) in ‘your Rijksstudio’, a personal exhibition collection (see figure 1). By doing so, this artwork becomes part of the collection of the registered user and given an implicit ‘like’. It also becomes immediately searchable and visible as a personal exhibition collection next to the curated museum collections. Collections can be given a ‘like’ as well. The image of the artwork can be downloaded and used freely. These user interactions are recorded and especially the interactions leading to the creation and changing of personal exhibition collections (also named hereafter: ‘user sets’) are available as open data.

(11)

11

When looking at the research questions, we expect to be able to use the open data of the personal exhibition collections to analyse the online visibility of the full digital collection and its popularity. As soon as an artwork is included in at least one personal exhibition collection, it is regarded as being discovered and the online visibility of the artwork is increased. The implicit ‘liking’ of the artwork when including it in a personal collection leads to a measurable influence of its popularity as we will discuss later on. Finally, the act of downloading an image of an artwork and ‘liking’ someone else’s personal exhibition set can also be regarded as user interaction, indicators of crowd dynamics and suitable for analysing the crowd

composition. The publicly available open data of the Rijksmuseum digital collection and the user data from Rijksstudio are used as the primary data sources in this comprehensive case study research.

1.4. Outline

In the previous paragraphs, we have discussed the museum institution in general in the context of society and economics, an ever changing environment in which the museum’s objectives are reconsidered and the need to provide visitors with a fulfilling experience has become crucial. The ability to connect and interact with visitors is challenging, and opening up the full digital collection online supports the attempt to understand crowd interests and dynamics. The Rijksmuseum website with its open data of the digital collection and Rijksstudio user data have been chosen for this research as a case study, and in the next chapters we will focus on the different aspects of the main research question to analyse what happens when a crowd engages with a digital museum collection.

Chapter 2 focuses on the visibility of the digital collection as a result of providing access and interaction tools to the crowd. How is the crowd discovering the digital collection? The formerly hidden depot collection has the potential to be seen and appreciated, but so far we do not know if this is actually the case. The discovery of museum and depot collections is the focus of this chapter.

Chapter 3 focuses on the popularity of artworks as an indicator of crowd dynamics. How popular is the digital collection in the eyes of the crowd? What mechanisms are influencing the popularity building and what part of the collection becomes popular?

Chapter 4 focuses on the group of visitors who created personal exhibitions and together form the crowd that can be studied using user data. The question on how to get to know the crowd is discussed in this chapter, where we are combining user data on the individual level to infer on what similarities in common behaviour characterizes parts of the crowd.

Chapter 5 will summarize chapters 2, 3 and 4 and finalize the conclusions of this research coming back to the central question of what happens when a crowd engages with a digital museum collection.

Chapter 6 contains referenced literature, mainly on the topics of crowd engagement with digital collections, museum experience, and curation.

Chapter 7 describes technical details about the open data interface, used tools and best practices that were used to analyse the large amount of data files available for this research.

(12)

12

2. How is the crowd discovering the digital collection?

In this chapter we discuss how an online crowd is discovering a large digital museum collection, given full access to this collection and tools to interact with it. The dawn of online digital media that allowed many museums to share their collection with the public more than ever before shed light on just how enormous some of these collections are. Following the Rijksmuseum in Amsterdam and the National Gallery of Art in Washington, the Metropolitan Museum of Art (Met) in New York announced their change in open access policy only a few years later in 2017, and opened up the vast majority of their collection of digitized works to the public domain (more than 375,000 artworks). But this is still only a relatively small portion of their complete collection of incomprehensible size. It is interesting to know why museums share their images with the general public. It seems that the Met already gives part of the answer in their press release by using phrases like ‘proven itself a leader among the world’s great cultural institutions’, ‘other institutions to follow’, working closely together with

‘Wikipedia that has hundreds of millions of users’ with ‘an enormous gift to the world’.33_{It is} tempting to say that the Met is trying a new marketing strategy, reaching out to a greater number of potential visitors, increasing exposure and impact. It could be a way to show sponsors that the museum is serious about its social responsibilities and contribution to global cultural heritage, thus being a good cause for donations. In the Netherlands, there is a growing tendency to become more externally oriented and look for public support and impact as a result of the changes in funding and the need to account for results.34_{It is important to} note that digitised collections act as a showcase for museums and generate interest that will eventually lead to physical visits, thus generating income.35

A clear vision and strategy for opening up the collection sometimes follows the

unpredicted effects of developments in other areas of cultural heritage digitization and open access initiatives. According to Pekel (2014), the Rijksmuseum agreed, after internal

discussions, to make high resolution images available for the first time in a Dutch Open Data competition.36_{Only after seeing the creative reuse of its images and realizing that pursuing} copyright violations would become too laborious (for example over 10,000 poor quality image copies of the Vermeer painting ‘Het Melkmeisje’ were already circulating online without permission37_{), the Rijksmuseum decided to publish all available high resolution} images of artworks under the public domain status38_{on the new website that was launched} in 2012.

Although the largest part of the collection, physically hidden in the depot, has been photographed and published online in order to generate interest by the public, it can be questioned if this formerly unseen part of the collection is actually becoming more visible. The artworks have to be searched for, discovered through browsing and recommended by others to look at them. According to Anderson (2009): ‘The things we’re most passionate about tend to be niche as opposed to mass market.’39_{This means that there is high potential}

33_{MET February 7, 2017}

34_{Museum Vereniging, Musea voor Mensen 2013}_p13

35_{Axiell 2016, p10}

36_{Pekel 2014, p7}

37_{Verwayen et.al 2011, p2}

38_See_{https://creativecommons.org/publicdomain/}

39_{Quote by Anderson in 2009 (see Proctor 2010 p40), author of ‘The Long Tail: Why the Future of}

(13)

13

of this formerly unseen and very diverse part of the collection being of great satisfaction to visitors, generating positive experiences and more interest.

Figure 2 Inside the Rijksmuseum depot (photo: Rijksmuseum)

As the crowd comes and goes online, visitors leave traces in search log files and web statistics, but by itself this does not lead to a measurable increase in the collection’s visibility, let alone a more permanent impact. According to Tanner (2012), impact is defined as: ‘The measurable outcomes arising from the existence of a digital resource that demonstrate a change in the life or life opportunities of the community for which the resource is intended.’40 His ‘balanced value impact model’ is a framework to support the process of decision making of those wishing to engage in impact assessment related to digital resources. Measuring impact is extremely difficult. Terras (2015) argues that the measuring and analysis of impact is a new discipline and only beginning to emerge and most museums do not have significant data about the use of their images because they do not have a long enough policy of open data. Besides anecdotal evidence on the re-use of images and website traffic numbers suggesting an increase of image use, there seems to be little other impact data as research tools to track and report the use of open access images still remain limited.41

The Rijksmuseum has implemented personal exhibitions in their Rijksstudio website and this provides an excellent opportunity to measure impact and to study real online user behaviour. Visitors can include images of artworks in their personal exhibition sets and because these sets remain visible, accessible and searchable within the website this provides a way to permanently increase collection visibility and encourage findability of artworks. Real data on the personal exhibitions and the full digital collection can be

combined to research the impact of the digital collection in terms of online visibility, a benefit also described as ‘bringing collections out of the dark’.42_{This very pragmatic approach to} exploratory research impact data is common ground in the GLAM43_{space, focusing upon} what can be measured rather than to worry about overarching definitions and connectivity back to core impact assessment concepts.44

40_{Tanner 2012, p4}

41_{Terras 2015, p744}

42_{Tanner 2012, p 21}

43_{GLAM (Galleries, Libraries, Archives, Museums)}

(14)

14

2.1. The online visibility of the digital collection

Before presenting results, we first look briefly at the available two data sets used for the analysis of collection visibility. First, the Rijksmuseum digital collection data contains metadata for each collection item separately. Some metadata can be used in determining the potential visibility of the online collection, grouping items for example according to location and image presence. The relevant data elements are:

- unique identification (xml: <artObject><objectNumber>) - object type category (xml: <artObject><objectTypes[1]>)

- on display physical location (xml: <artObject><location>), in case this element is absent for an artwork, it is assumed that the item is located in the depot

- an indicator for the presence of an open access digital image (xml:

<artObject><hasImage>; xml: <artObject><showImage>); both should be ‘true’ to indicate open access of the image

Second, Rijksstudio user data on personal exhibition sets contains data about the set itself and the artworks included, real proof of user behaviour. The relevant data elements are:

- user identification (xml: <userSet><user><id>) - set identification (xml: <userSet><id>)

- set date of creation (xml: <userSet><createdOn>)

- references of artworks included in the set; the artworks point to the digital collection data with their unique identification (xml: <userSet><setItems><objectNumber>) In March 2018, according to the downloaded open data at the time, the Rijksmuseum collection consisted of 626,441 artworks in 2,001 different categories (like paintings, prints, sculptures, plates and many more).45_{Only a mere 1.4% (8,605 items) was indicated as} shown on-site in the museum and the remaining 98.6% was kept in the depot. Museums in the Netherlands in general are able to display about 8% of their collection.46_{The difference} to the Rijksmuseum collection can be explained by considering the large number of prints, drawings and photographs (together forming a staggering 84% of the total collection) that cannot be displayed on-site mainly because of conservation reasons. If these types are ignored in the counting then the museum displays actually 14.2% of its collection, an above average percentage.

The full digital collection can be subdivided into a museum collection and a depot

collection, based on where the artworks are kept. This subdivision is made here because of the inherent difference in visibility between artworks on public display in the museum and artworks in the depot that are not physically displayed but only shown to online visitors. When studying crowd engagement with a digital museum collection, this difference is relevant because we suspect it influences the size and nature of the engagement. A second subdivision can be made between the collection part that has a digital image online and the part that does not have an image or may not be shown for reasons of authorship or portrait rights. This subdivision is chosen here because for this research the crowd engagement with a digital collection is studied by looking only at the use of images in personal exhibition sets. It relates to the imagery nature of the Rijksstudio website. The ‘collection without image’ consists of 245,124 items in the depot and 1,938 items in the museum (247,062 artworks, 39.5% of the total collection) and it is unlikely to be chosen by online visitors because of the

45_{Open data changes constantly and therefore the captured data is stabilized on the reference date.}

(15)

15

prominent focus of the website on artwork images and the sets that include these images. Excluding the ‘collection without image’, what remains are the collection items that do have a digital image (242,512 + 136,867 = 379,379 artworks, 60.5% of the total collection). This collection can be subdivided again into two parts, where one part counts for the collection items included in one or more Rijksstudio personal exhibition sets, and the other part for the ones not included in any set. Inclusion of an item in a set is regarded as a measurable increase of the collection visibility for that item, because it proves that it has been seen by at least one online visitor and inclusion of the item in a set subsequently helps in the discovery of the item by other online visitors. Subdividing the collection this way results in six groups that can be visualized in a tree map to show their relative sizes (see figure 3).

The visibility of the ‘collection with image’ can now be defined as the percentage of artworks of this collection actually included at least once in a user set, and this can be calculated as 36.1% (= 136,867 / 379,379 items). When we subdivide the ‘collection with image’ into a museum and a depot collection, we see that for the museum part the visibility is 80.3% (= 5,352 / [1,315 + 5,352] items) and for the depot part it is 35.3% (= 131,515 / [241,197 + 131,515] items).

Figure 3 Digital collection composition showing visibility and use in Rijksstudio sets

Considering these visibility percentages, the museum collection is far more visible at the moment than the depot collection. It is not clear what causes this difference. The collections are very different in size and composition when looking at the types of artworks. The

digitization and online publication of the collection could have been done in a particular order no image image, not included image, and included total museum 1,938 1,315 5,352 8,605 depot 245,124 241,197 131,515 617,836 total 247,062 242,512 136,867 626,441

(16)

16

and search and retrieval (and subsequent inclusion in personal sets leading to more

exposure) of collection items could have been in favour of museum items at the beginning of the online publication. The promotion and recommendation mechanisms (discussed in more detail in the next chapter) might also be a factor.

2.2. The discovery rate

Publishing artworks online with their metadata and a high resolution image does not immediately imply that all these artworks are actually seen by visitors. Inclusion of an artwork in a personal exhibition set in Rijksstudio is proof of the fact that at least one visitor has ‘discovered’ it and included it in a set, thus initiating a more permanent online visibility because user sets are searchable and findable on the website. Downloading an image of an artwork could also be regarded as a discovery moment, but for two reasons download data is not used here. First, although the artwork is seen and in that sense discovered by an individual online visitor before downloading its image, the downloading itself does not add a visible instance of the image to be seen by others within the context of the website, as opposed to inclusion of the artwork in a user set. It does not really contribute to the collection visibility. Second, from the download data it appears that at some points in time large parts of the collection were downloaded massively by only a few ‘users’ (see chapter 4.1). These users are not regarded as part of the crowd that we intend to study in this research and so we chose to exclude download data altogether from the discovery analysis.

For each artwork in the collection we can derive its ‘earliest discovery moment’ in Rijksstudio by taking the oldest creation date of the user sets containing the particular artwork. Because we lack transaction data for each individual inclusion of an artwork in a user set we can only use the creation date of the set itself. This will obviously lead to discovery dates that will sometimes precede the actual date of the inclusion, but working with a large amount of collection items and user sets this will temper the effect. Overall, the goal will be to see if the time lines for the discovery of museum and depot items show any similarity or not. Plotting out the accumulated number of discovered artworks in time (see figure 4) shows particularly clearly the large depot collection that is increasingly being seen and used by online visitors and this strong positive trend seems to continue at a rather steady pace.

(17)

17

As calculated earlier, almost two thirds of the depot collection still awaits discovery, whereas the overall discovery rate of the museum collection is already quite high. We can compare the discovery rates of the depot and the museum collection to see if there is any similarity. First of all, we can plot the monthly increase percentage (= number of discovered artworks for that month / total number of artworks to be discovered) for both the depot and the museum collection separately (see figure 5, blue and yellow line). The difference between these monthly percentage increases (named hereafter: delta increase) is plotted as well (the red dotted line). Judging by the strong upward trend of the delta increase, the museum collection seems to have been discovered much faster between the early start in October 2012 and about February 2014. This kick-start of the museum collection discovery cannot be explained from the available open data alone, and further analysis might be possible by linking additional data on the digital publication process and promotion.

The delta increase seems to stabilize in 2014 as the depot collection is discovered at more or less the same discovery rate as the museum collection during the next two years. In 2016, the delta increase turns to a downward trend because the museum discovery appears to slow down whereas the depot discovery remains rather constant. As discovery of both the museum and the depot collection continues over time, it can be expected that the ‘collection with image’ will be fully discovered in the near future. Based on the calculated average discovery rate of about 0.5% monthly of the depot collection, and assuming that there will be no significant changes in publication and promotion activities, the collection is expected to be fully seen and used within the next sixteen years, pointing somewhere towards the end of 2028.

Figure 5 Accumulated discovery percentage of depot versus museum collection

2.3. Summary

In this chapter, we have discussed how an online crowd is discovering a large digital museum collection, given full access to this collection and tools to interact with it. Especially the depot collection part, formerly hidden because it is not displayed in the physical museum, waited to be found and appreciated by online visitors. The high potential of this formerly unseen and very diverse part of the collection lies in it being of great

satisfaction to visitors, generating positive experiences and more interest. We have argued that assessing and measuring impact is still difficult and that a pragmatic approach is

common ground where actual user data is collected by facilitating access to and use of a full digital collection online and record user activities.

(18)

18

The Rijksmuseum published its collection online and provided features for visitors to include artworks in personal exhibition sets in Rijksstudio, the collection discovery

environment of the Rijksmuseum website. Inclusion of an artwork in a personal exhibition set is a concrete individual user action that proves online discovery of the artwork. We have analysed the open data of the collection together with the open data of the user sets to find out that the collection is indeed being discovered by the online crowd and that this is occurring at a steady rate.

The Rijksmuseum open data was downloaded for this research in March 2018. The division between the museum and the depot collection turned out to be 1.4% to 98.6%, but the collection has a large number of prints, drawings and photographs in the depot that is not representative of the sheer diversity of the whole collection that covers 2,001 different

categories. When filtering out the mentioned three categories, we conclude that 14.2% of the collection is displayed in the museum on-site, an above average compared to museums in general.

Not all artworks have a digital image, and because the nature of the Rijksstudio personal exhibition sets is image-based, we regard this part of the collection not relevant for the analysis of the online discovery of the collection. Looking at the difference between the museum and the depot collection for the remaining part that actually has a digital image, we have observed that there is a clear difference in discovery between the museum and the depot collection. The museum collection discovered so far is 80.3% and the depot collection 35.3%. This profound difference cannot be explained by the difference in collection size and composition alone and the used data is insufficient at the moment to provide a clear

understanding. A lot of different factors are expected to play a role in discovery, such as the order in which digitization and publication might have taken place and the deliberate

promotion and recommendation of artworks.

The rate of discovery has been calculated as a monthly percentage of the collection that is included for the first time in that month in a personal exhibition set. Although we have seen that a limitation in the available user data requires some assumption as to the ‘earliest

discovery date’ of an artwork, we can however compare the museum and the depot

collection as we are using the same calculation for both. There is a clear distinction between the museum and the depot collection in size and composition, and the museum collection seems to have had a kick-start in discovery. But the crowd indeed discovers the formerly hidden depot collection rapidly and a strong positive trend is shown. Because the depot collection is much larger in size compared to the museum collection it takes more time for it to be discovered. The discovery rate looks to be rather similar for both the museum and the depot collection with about 0.5% monthly, but the rate starts to slow down once the

discovery percentage approaches its maximum. This is already happening for the museum collection. For the depot collection, this maximum is expected to be reached in 2028.

To finalize this chapter, the Rijksmuseum case study shows that when a crowd is

engaged with a digital museum collection, this collection is being discovered by that crowd at a more or less steady rate and visibility is maximized over time as a result of the inclusion of artworks in personal exhibition collections. Especially the formerly hidden depot collection is brought out of the dark and into the appreciation of the crowd.

(19)

19

3. How popular is the digital collection?

In the previous chapter, we have seen that the digital collection is increasingly being discovered by an online crowd once it becomes accessible, findable and usable. But we have not looked into any of the mechanisms for the online discovery of artworks, and for the subsequent influencing of the same crowd through promotion or recommendation. How do visitors discover artworks and tell others about it? Do they start searching for artworks from scratch, do they stumble upon an artwork somewhere, or do they get recommended by others? In this chapter, we attempt to study crowd dynamics and we start by using the popularity of artworks as an indicator for this.

Serendipity is the art of chancing upon things, of making unexpected discoveries. Environments can be designed so as to encourage or discourage serendipity. According to Chan (2007), when a museum visitor browses along showcases, serendipity plays a considerable role in the ‘user experience’. Even in highly specialised exhibitions, visitors inevitably come across objects that are of interest to them but that they were unaware of prior to visiting.47_{And when they subsequently recommend their discoveries to others, they} influence the popularity of artworks. Popularity (a word that has become popular by itself after its introduction in the 17th century) can be defined as ‘the fact that something or someone is liked, enjoyed, or supported by many people’.48_{It implies that the combined} choices of individuals in a crowd together lead to consensus on who or what becomes popular. As individuals are socially influenced, the crowd is influenced as a result.

Museum visitors have heard that they must see a certain painting, or they must visit a particular exhibition. They feel that if they don’t, no matter how wonderful the rest of the visit, it will somehow be incomplete.49_{These hopes and expectations vary between visitors; they} derive from a variety of sources, but they are always there. The structure of the visitor’s agenda determines, to a large extent, the museum experience eventually recalled. Indeed, understanding and influencing the visitor agenda is fundamental to the museum’s ability to attract visitors, meet their expectations and create a successful museum experience.50 Bhowmick and Mitra (2019) have demonstrated that popular businesses mostly attract customers through social influence, while unpopular businesses, on the other hand, mostly rely on customers who have an inherent preference towards those category of businesses.51 If we project this onto artworks and visitors, it becomes interesting to analyse the crowd as it probably doesn’t behave in a homogeneous way. In chapter 4, we will look more closely at the crowd to see if the difference between social influence or inherent preference can also be traced back to personas, but for now it suffices to argue that measuring popularity is a step in understanding crowd dynamics.

Finding out what visitors choose to look at, despite museum ‘advertising’, is not very easy. At least two mechanisms need to be taken into account. First, deliberate promotion of selected artworks by the museum is a common way to influence visitors. In the first place, the visitor’s perspective of the museum is that of a consumer. Besides the collection, this perspective includes impressions of the building, the friendliness of staff, the gift shop and restaurant. It is a holistic view of the museum.52_{Even outside the museum (or website) there}

47_{Chan 2007}

48_{Popularity as defined by the}_{Cambridge Dictionary}

51_{Bhowmick and Mitra 2019, p66}

(20)

20

is influence when an artwork is seen in many different places and the more it will be seen, the more it will be remembered and shared and the more popular it will become. The same goes for the creators of artworks, like Rembrandt being promoted by several museums in the Netherlands throughout 2019, 350 years after his death on October 4th_{, 1669. The museum} curator working on the collection is the one who professionally decides what artworks are most significant and most representative of the story that a museum wants to tell. Besides predesigned tours (guided by a brochure, a mobile app or a real person) that take visitors to the curated artworks, the museum shop sells products bearing the images of artworks and banners, while flyers and billboards attract attention to the top artworks that are tightly linked to the museum. On the Rijksmuseum website, used here as a case study, a concrete

example of online promotion is the so-called ‘mastermatcher’, a feature that helps visitors to create their own personal exhibition, but also promotes a selection of artworks. In the next paragraph we will analyse further what influence this tool has on the recommendation of artworks and subsequent popularity.

Second, word-of-mouth is an important recommendation mechanism which helps individuals form opinions about their world. In general, word-of-mouth is perceived as a highly credible source of information because it is free from the bias of the people who make, sell, or deliver a product or service. Visitors likely to hear about an artwork through word-of-mouth are also most likely to tell others about it as well after their visit.53_Presenting personal exhibitions on the Rijksstudio website is a strong implementation of this

recommendation mechanism, as online visitors are triggered to look at what others find pretty and valuable and decide to actively ‘like’ it or include the work in their own collection. From the museum’s point of view, Rijksstudio with the personal exhibitions was proclaimed as an innovative collection discovery interface.54_{Indeed, it supports the visitor’s feeling of} serendipity when browsing along the exhibition sets. Visitors can react and recommend by starring other visitors and their personal exhibitions, or include an artwork found in someone else’s exhibition set into their own. The Rijksmuseum itself has curated and published more than 280 exhibition sets on certain themes, styles and artists to provide guidance and to influence online visitors when browsing the collection. These sets can be regarded as promotion alongside recommendation, as these collections are shown side-by-side with personal exhibition sets in the browsing section of the website.

3.1. The mastermatcher effect

One way to promote artworks is the use of the wizard-like tool on the Rijksstudio website called ‘mastermatcher’ that helps visitors create an online personal exhibition. It has been implemented from the beginning in 2012 to encourage visitors to start collecting themselves. The mastermatcher is a five-step wizard in which the visitor is asked to choose a word for each step and in the end, the wizard generates a small set of artworks based on the choices made. Because 4,080 pre-selected artworks are configured in the content management system of the website for this wizard, it can be regarded as a promotion tool from the museum’s point of view. To the visitor, the resulting set creates the feeling of serendipity as unexpected artworks are discovered. The mastermatcher sets can be traced from the open data because of their formatted set name (xml: <userSet> <id>), and the word choices made in each step of the wizard can be traced from the recorded keyword with each artwork included in the set (xml: <userSet> <setItems> <relationDescription>). This deduction is not

(21)

21

fool proof because visitors can change set names afterwards, but it is assumed here that most of the times they will not. Looking at the number of items for mastermatcher sets, we see that most of them have 25 items, followed by 20 items and a few in between (see figure 6). Looking at the large variation in the number of items per set (up to 266 items when looking at the horizontal axis), it does show that some sets have been altered extensively after the initial generation, adding artworks or even deleting some that were generated by the wizard.

From the open data we can conclude that 64,847 users (45.2% of the total number of 143,427 unique visitors) have performed the wizard one or more times, generating 76,538 user sets (30.7% of the total number of 249,275 sets). A staggering 1,740,566 times has a mastermatcher artwork been included in these sets (58.9% of the total of 2,950,764 inclusions). Given these numbers, the impact of the mastermatcher on the creation of personal exhibitions can be regarded as high, influencing the popularity of the 4,080 collection objects included in the wizard. Popular choices made within the wizard are for example the colour blue in step 3 and the city of Amsterdam in step 4 (see figure 7).

Figure 6 Size of the generated user sets as resulting from mastermatcher

Figure 7 Popular choices made by visitors using the mastermatcher

We can now divide the personal exhibition sets into two types: the mastermatcher sets curated and promoted by the Rijksmuseum, and the non-mastermatcher sets, created by the registered user. Popularity has been made measurable in this research by defining it as the

(22)

22

number of times the artwork is included in a personal exhibition set. Popularity listings created for individual artworks were based on the total number of their inclusions in these sets. This was done for both types of sets separately and for the combination of all sets together (see figure 8, the three right columns). Notice the differences in the ranking of artworks in the overall top-20 when comparing the mastermatcher and non-mastermatcher sets. For example, ‘Het Melkmeisje’ by Johannes Vermeer is overall second, but first in the non-mastermatcher sets. This difference is the result of the painting not being part of the mastermatcher and thus not contributing to the overall ranking. ‘De Nachtwacht’ by

Rembrandt van Rijn, however, is included in the mastermatcher and also popular in the non-mastermatcher sets, and the painting becomes the overall number one. These examples show clearly that popularity is influenced by the promotional influence of the wizard. The right two columns in figure 8 show the individual ranking in the mastermatcher and non-mastermatcher sets and the highlighting indicates which ones have contributed most to the overall ranking.

Figure 8 Popularity ranking for the top-20 individual artworks

3.2. Other promotion and recommendation mechanisms

We can exclude the mastermatcher effect from the available open data by ignoring the mastermatcher sets and focus only on the remaining sets that are somewhat less influenced by promotion. These concern 78,580 users (54.8% of the total number of 143,427 registered users) with 172,737 user sets (69.3% of the total number of 249,275 sets). Mastermatcher is only one example of a promotional effect and of course there are many other mechanisms that play a role in popularity building but not all of them can be linked directly to the open data of the personal exhibition sets. Despite these limitations, we can at least add some data to the popularity listing for non-mastermatcher sets for further analysis (see figure 9).

It makes sense to argue that seeing an artwork in the museum, as opposed to it being hidden in a depot, makes it more popular in personal exhibitions. In this context, the 8,605 artworks displayed in the museum are promoted by the museum curators. For this reason, the location of the artwork (depot or museum) has been added to the listing. Of the top-20

Object number Title Principal maker

Overall ranking Master matcher Non master matcher

SK-C-5 De Nachtwacht Rembrandt van Rijn 1 186 2

SK-A-2344 Het melkmeisje Johannes Vermeer 2 missing 1

SK-A-2382 Kinderen der zee Jozef Israëls 3 7 4

SK-C-214 Stilleven met bloemen in een glazen vaas Jan Davidsz. de Heem 4 missing 3

SK-A-3584 Meisje in witte kimono George Hendrik Breitner 5 83 7

RP-P-OB-23.417 Twee engelen Pieter Willem van Megen 6 2 462

SK-A-1577 Abraham ontvangt de drie engelen Ferdinand Bol 7 1 517

SK-A-1718 Winterlandschap met schaatsers Hendrick Avercamp 8 290 5

SK-A-3602 Morgenrit langs het strand Anton Mauve 9 6 24

SK-A-4688 Italiaans landschap met parasoldennen Hendrik Voogd 10 48 15

SK-A-2860 Het straatje Johannes Vermeer 11 536 6

RP-T-1993-174 Hovelingen op bezoek bij een kluizenaar anoniem 12 3 1,401

SK-A-370 Een meisje plaatst een kaars … Godfried Schalcken 13 4 833

SK-A-3597 Ezeltje rijden langs het strand Isaac Israels 14 13 56

SK-A-4226 De zee bij Katwijk Jan Toorop 15 11 79

SK-A-4 De bedreigde zwaan Jan Asselijn 16 797 9

BK-1963-101 Zittende Amor Étienne-Maurice Falconet 17 37 45

SK-A-3580 De Singelbrug bij de Paleisstraat … George Hendrik Breitner 18 100 34

RP-T-1933-44 Het slagnet Adriaen Matham 19 5 2,105

(23)

23

most popular artworks, fourteen were located in the museum in 2018, as opposed to six in the depot. This suggests that museum artworks are more popular than the ones in the depot, but it is profoundly reassuring to see that depot artworks can also become popular with the crowd. It appears that the crowd (or part of it at least) is indeed seriously interested in the depot collection. The third most popular painting by De Heem showing flowers in a vase is the top artwork stored in the depot. Possibly also the most visible image in the physical Rijksmuseum shop, it is displayed larger than real and with backlight illumination behind the counter where customers can wait for their gift to be wrapped by one of the staff members. It seems that visitors occasionally point to the image of the flower painting behind the counter and ask where to find the original in the museum because they appreciate it so much, leaving the staff member hesitant and a little embarrassed to respond that the original unfortunately cannot be admired in reality.55_{The appearance of images of the top-20} artworks on products in the Rijksmuseum shop has therefore been added to the popularity listing. It turns out that nineteen out of twenty are actually used on one or more products. This suggests a correlation, but it would require additional data on the full product catalogue and sales numbers to actually analyse if artworks depicted in the shop become more popular as a result of the shop turnover.

Figure 9 List of top-20 popular artworks in Rijksstudio excluding the mastermatcher effect

55_{Personal note as observed by the author}

ID Title Principal maker Type Year Location

Count Non-MMsets Count MM-sets Count total Number of downloads Number of apps Number of RM-sets Shop

SK-A-2344 Het melkmeisje Johannes Vermeer schilderij 1660 HG-2.30.3 16313 16313 30110 0 2 yes SK-C-5 'De Nachtwacht’ Rembrandt van Rijn schilderij 1642 HG-2.31 15064 2748 17812 32969 3 3 yes SK-C-214 Stilleven met bloemen in een glazen vaas Jan Davidsz. de Heem schilderij 1650 depot 13073 13073 15892 0 1 yes SK-A-2382 Kinderen der zee Jozef Israëls schilderij 1872 HG-1.18 8484 6380 14864 9113 0 2 yes SK-A-1718 Winterlandschap met schaatsers Hendrick Avercamp schilderij 1608 HG-2.6 8371 1534 9905 14513 1 4 yes SK-A-2860 'Het straatje’ Johannes Vermeer schilderij 1658 HG-2.30.3 8111 926 9037 14331 0 3 yes SK-A-3584 Meisje in witte kimono George Hendrik Breitner schilderij 1894 depot 6570 4270 10840 8456 0 2 yes SK-A-3262 Zelfportret Vincent van Gogh schilderij 1887 HG-1.18 6482 6482 12742 2 4 yes SK-A-4 De bedreigde zwaan Jan Asselijn schilderij 1650 HG-2.30.5 6475 418 6893 7716 2 3 yes SK-C-251 Brieflezende vrouw Johannes Vermeer schilderij 1663 depot 6234 6234 9929 2 2 yes SK-A-799 Stilleven met bloemen Hans Bollongier schilderij 1639 depot 5907 491 6398 9070 0 2 yes SK-C-216 'Het Joodse bruidje’ Rembrandt van Rijn schilderij 1665 HG-2.30.7 5080 5080 8865 2 2 yes SK-C-229 Het vrolijke huisgezin Jan Havicksz. Steen schilderij 1668 HG-2.30.4 4885 4885 8207 1 5 yes SK-A-1505 'In de maand juli’ Paul Joseph Constantin Gabriëlschilderij 1889 HG-1.18 4871 1 4872 7641 2 4 yes SK-A-4688 Italiaans landschap met parasoldennen Hendrik Voogd schilderij 1807 HG-1.12 4702 4895 9597 7170 0 2 yes SK-A-268 Stilleven met bloemen en een horloge Abraham Mignon schilderij 1660 depot 4642 4642 6425 0 2 no SK-A-4050 Zelfportret als de apostel Paulus Rembrandt van Rijn schilderij 1661 HG-2.30.8 4343 4343 8067 1 5 yes SK-A-3064 Portret van een meisje in het blauw Johannes Cornelisz. Verspronckschilderij 1641 depot 3865 3865 5165 0 2 yes SK-A-4821 Stilleven met kazen Floris Claesz. van Dijck schilderij 1615 HG-2.30.2 3776 3776 6563 1 1 yes SK-C-211 De molen bij Wijk bij Duurstede Jacob Isaacksz. van Ruisdael schilderij 1668 HG-2.30.6 3721 3721 6520 2 3 yes

(24)

24

There are other promotion and recommendation mechanisms that might influence the popularity of individual artworks. For example, the Rijksmuseum curators themselves have acted as online visitors by creating sets dedicated to a particular artist, style or historical period. In this way, they have promoted 1,639 individual artworks curated in 287 different sets.56_{All of the top-20 artworks are present in one or more of these Rijksmuseum sets (see} figure 9, column ‘Number of RM-sets’), suggesting a correlation. Visualizing the popularity ranking versus the number of sets that contain the artwork appears to support this, as the high ranking artworks also have more occurrences in these sets (see figure 10). The mobile app that visitors can use as a museum tour guide also contains 136 individual works of art in fourteen tours. Eleven of the top-20 artworks are present in one or more of these app tours (see figure 9, column ‘Number of apps’). The numbers here are much smaller, but when we visualize the ranking and occurrences in the same way as before, we again see a similar pattern where the higher ranking artworks are more likely to also be in the mobile app (see figure 11).

It is obvious to argue that when a work of art is visible in different places (museum, website, Rijksstudio, mobile app, shop), it is also more likely to become popular. All the mentioned recommendation mechanisms add up to about 14,000 artworks (still only 3.8% of the 379,379 artworks with a digital image online) being promoted by the museum curators to increase the popularity of these artworks. The painting by Mignon (number sixteen in the popularity listing) might be an example of an artwork that has become popular despite the fact that it is not promoted heavily. The painting is not visible in the museum, not used in the shop nor in the app, but only in two Rijksmuseum sets online. But maybe because one of these sets is dedicated to the painter himself and the work of art is a still life with flowers (see next paragraph on search keywords), it had a strong upwards popularity push.

Figure 10 Popularity ranking versus occurrences in Rijksmuseum sets

Figure 11 Popularity ranking versus occurrences in app tours

So far, we have only looked at individual artworks in trying to analyse what kinds of recommendation mechanisms might influence popularity. We can also look at the type of artwork and deducting from the top-500 listing we can conclude that an overwhelming 66% are paintings and almost two-third of these are actually displayed in the museum. Next are drawings (12%) and prints (9%) of which none in the top-500 are displayed in the museum.

(25)

25

Sculptures follow with a mere 2% and the remaining 38 types together make up the final 10% of the top-500. It is fair to say that paintings are a really popular type of artwork and this hardly comes as a surprise considering this type is also the most visible in the museum itself. But comparing paintings to prints or any other type in terms of popularity seems pointless as complicating factors get involved in the analysis, such as the number of

artworks of a particular type as well as their difference in the composition between the depot and the museum. Is it fair to compare one hammock in the collection (actually popular because it is included in the mastermatcher) to thousands of paintings (some popular and others not at all)?

Popularity changes over time, which at any moment provides the opportunity for the niche types that are not so visible in the collection to become more appreciated. A particular category can become more popular once it has been the topic of an exhibition as is

illustrated when plotting the inclusion of artworks of a particular type in personal exhibitions on a timeline. Looking for trends, it seems that a lot of the types follow the same pattern in time, but for example the type ‘fashion print’ (see figure 12, indicated by the red line) seems to behave differently, picking up an upward popularity push after June 2015. This coincides nicely with the Rijksmuseum exhibition at the time where fashion and flowers were

prominently displayed together.57_{However, the most popular flower painting (green line,} identified by SK-C-214) does not seem to have benefitted from the exhibition. Maybe a future exhibition particularly focusing on flowers would show similar effects in Rijksstudio personal collections, pushing flower paintings and tulip vases upwards.

Figure 12 Number of items in personal sets grouped by object type (logarithmic scale)

In Rijksstudio, anyone can vote for personal exhibition sets created by other visitors. By starring a set, thus giving it a ‘like’, a visitor shows engagement with the collection and appreciation for that particular set. The set can be regarded as becoming more popular as a result of this ‘like’ action. We can assume that this will increase the popularity of the items in that personal exhibition set, is likely to make them more visible online, and can act as additional recommendation. Although we cannot measure the absolute impact of this

Collecting bits and pieces: What happens when a crowd engages with a digital museum collection?