Deep personalization: A case study of systems thinking within an art museum

Paul Fishwick, University of Texas at Dallas, USA


How can museums become more personal—to connect better with each unique visitor? A visitor will have a general profile that is a combination of the visitor’s background, topical subject interests, and learning style. We recommend a deep personalization strategy that begins with learning as much as possible about the visitor, with special attention paid to learning topical interests within STEAM (Science, Technology, Engineering, Art, and Mathematics). Each visitor will enter a museum with a specified combination of these five topics, and so a visitor model can be viewed as a five-dimensional space with each visitor being a point in, or region of, this space. This region is termed a topical visitor profile. Systems thinking is one example of a profile that is pervasive across the four areas of STEM. We chose to discuss systems thinking within an art museum because we hypothesize that a subset of art museum visitors will have this profile, and therefore the museum is tailoring its object interpretations to these visitors by exposing them to systems thinking on any particular object. We describe a hybrid software/hardware architecture based on 1) the Physical Web concept, which represents a natural extension of the Semantic Web; 2) a collaborative wiki-style approach to curating knowledge about each object; and 3) a desire to seamlessly connect the informal museum setting with formal learning classes, subjects, and their associated subject standards. Deep personalization represents the broad goal described in the paper, and systems thinking is an example of personalization for STEM learning.

Keywords: Education, Visitor Studies, Deep Personalization, Concept Map, State Machine, Systems, Modeling

1. Introduction

What attracts visitors to an art museum, and how can we learn more about their interests? Pitman and Hirzy (2010) present a research study resulting in segmenting visitors into four groups: observers, participants, independents, and enthusiasts. The Wallace Foundation published a study on how museums can be more visitor centered (WF, 2001). Over a decade has passed since this publication, and museums are increasingly visitor-centric. The museum’s goal is to reach out to the visitors, thus better engaging them. These studies provide a much-needed evidence-based approach to museum visitation, but more research is needed to allow visitors more degrees of freedom in setting their interests and overall profile.

Consider an example museum engagement that could exist in the near future. Mary goes into an art museum and has selected to receive information on “A”rt and “S”cience (i.e., two letters of STEAM) from her cell phone application. She has chosen art because of a personal interest and science because of her science classes at high school. Mary pauses in front of a painting by Thomas Hart Benton. She presses “like” on her phone. This personal preference results in related artwork being presented to her. Maybe she would like other paintings by Benton or artwork from the same period or genre? Benton’s painting depicts a landscape, which appears similar to the field near her grandmother’s house. It would be nice to take a photo of the painting and engage in social media to elicit reactions from her friends. Could her friends relate to similar landscapes?

Another of Benton’s paintings depicts the prairie (i.e., grassland), which she is learning about in her earth science and biology classes at high school. The earth science class is currently covering prairie soils and geology, and the biology class is covering prairie ecosystems. The connection from informal learning in the museum to formal learning in the high school would have been previously established through a link to state or national standards. Information on prairie ecosystems has been collaboratively edited in the Wikipedia Web page for Benton’s painting. A few university and high school faculty have added their topical material and co-edited the Wikipedia page. The page also includes links to the regional science museum where geological dynamics and ecology are highlighted in special exhibits. The smartphone brings the geology lesson to her attention, and she bookmarks this for later perusal so as not to interrupt her experience of Benton’s work. She is also made aware of a news topic about a Benton exhibit in New York. She reads the headline and bookmarks this news item as well for later reading. The interest in ecosystems generates a mobile-assisted tour throughout the museum for objects that reflect ecology. Mary goes on this tour and then decides to depart the museum.

The following qualities capture Mary’s learning experience and represent deep personalization:

  1. Self-directed visitor learning: Mary is in control of her learning experience.
  2. Topic-based visitor profile: The museum knows what topics Mary likes or may be learning at school (e.g., art, biology, earth science).
  3. Diverse multidisciplinary object interpretation: Knowledge extends beyond a single subject (e.g., art) to include areas that match Mary’s topic-based profile.
  4. Community knowledge: Knowledge is community generated beyond the bounds of the museum (e.g., ecosystem expert from local school or college).
  5. Informal-to-formal knowledge network: There are explicit pathways from discipline-specific interpretation for museum objects to places and curricula associated with formal learning.

These qualities are not collectively present in museums despite their highly innovative projects (e.g., Brooklyn Museum’s “Ask” or Cleveland Museum of Art’s “Gallery One”). Self-directed learning (quality 1) is defined as learning while at an object without necessitating human assistance—although, in practice, such assistance, if available, is desirable before the self-directed stages of learning. A topic-based profile (quality 2) would need to be made by Mary during the visit. The topics are multidisciplinary (quality 3), requiring input from multiple experts in and outside of the museum (quality 4). Experts outside of the museum can include collaborators at schools and universities. The goal of quality 4 is to open up the interpretation to a wider community as a way of strengthening connections to the museum, and fostering richer experiences with objects. Quality 5 is addressed by Falk (1999) as a challenge for museums. Museums are places of learning, but this learning frequently has no formal tracebacks to schools and their curricula.

What makes museums so interesting is that they are home to rare and unique objects, making the visitor-object experience a significant learning opportunity not available in a formal school setting. Each object has connotations afforded by disciplines as well as by relevant concepts such as materiality (Dudley, 2013). Many objects will have representational properties ascribed by an observer or defined by the creator. Placing this object-first perspective into a theory of aesthetics, we advocate a manifold object-discipline experience between the object and the observer. This frees the object from a singular disciplinary scaffolding or context, and is similar to arguments expressed by Dewey (1934) and Duchamp (1957), whose views are described and compared by Leddy (2012). Dewey’s focus was on art as the experience between observer and object, rather than an inherent attribute of an object. Duchamp observed that “the spectator” contributes to the creative act. The object freely represents a wide variety of interpretations, which are independent of physical context, although if a museum has a declared emphasis (e.g., art or science), then this emphasis can be made operational by making it a default topic in the visitor profile when in that physical space. When in an art museum, the visitor would have the “A” topic selected by default. The multidisciplinary interpretive potential for an object may run counter to treating the museum as a ritualistic space (Duncan, 1995).

Deep personalization is defined as spanning the personalization continuum by allowing for learner customization (topic-based selection by Mary) at one end of the continuum, as well as personalization (guidance by the museum based on visitor topical profile) at the other end. Technology would allow Mary to be automatically notified when a painting matches her personal profile, and this would require the profile to be known to the museum. A sensing mechanism would connect the presence of painting to Mary’s physical position. Questions for realizing this hypothetical scenario include 1) who creates the content? 2) how is the visitor position determined? and 3) what is the effect of this personalization on visitor attendance and repeat visits? It is the combination of all five qualities specified above that makes deep personalization a unique line of research.

The article begins with the motivation to form connections. This is followed by a discussion of personalization in the museum, and why deep personalization represents a strategy to be nurtured within museums. We’ll present a case study of learning systems thinking via an Incan cultural object within the Dallas Museum of Art. Consider systems thinking to be a prototypical example of knowledge that may be of personal value to the museum visitor and of relevance to the cultural museum object, but is knowledge not normally conveyed within art museums. Systems thinking cuts across typical disciplinary boundaries in STEM (Science, Technology, Engineering, and Mathematics). STEM and systems thinking represent new “ways of seeing” in “A”rt museums, and so capture a practical approach for STEAM education within these museums.

2. Making connections among multiple disciplines

Manifold experiences are defined by a wide range of interpretations for an object. An object can elicit numerous interpretations, guided from different experiences—for example seeing mathematical perspective in a painting or analyzing the painting for symbology. The “art” is in these object experiences. For STEAM, this means that the “A” becomes more than a body of knowledge within art theory and practice. The “A” represents the very activity of manifesting diverse object-based interpretations.

Museums—especially art museums and their personnel—have long had interests in making connections to and from a cultural object. For example, Pitman, while serving as the Eugene McDermott Director of the Dallas Museum of Art, initiated the Center for Creative Connections. The Brooklyn Museum houses Connecting Cultures: A World in Brooklyn. Fishwick et al. (2016) created Models of X, which was a Web-curated set of different views of Liz Larner’s X sculpture, which was on loan to the Nasher Sculpture Center for its XChange exhibition (XChange, 2013).

Making connections is guided by the notion of discipline. Phrases such as multidisciplinarity, interdisciplinarity, and transdisciplinarity are commonly found in different literature (Stock & Burton, 2011). These are briefly defined as follows: discipline is a well-formulated body of knowledge; multidisciplinarity involves several disciplines combined together; interdisciplinarity involves disciplines that are combined through both knowledge integration and social exchange; and transdisciplinarity represents a strong form of interdisciplinary activity where the emerging project represents an entirely new form. Postdisciplinarity (Muller, 2007) is an attempt to go beyond disciplines, so as not to be confined by any single one. Our project is classified as object-based X through self-directed learning, where X is one of the above phrases. For instance, object-based, multidisciplinary, self-directed learning means that different disciplines are connected to an object by discovery and delivery of multiple-discipline-specific knowledge while someone is attending to the object. Self-directed learning implies knowledge discovery, acquisition, and control by the museum visitor. Through improved personalization, can we meet the “scaling up” challenge by Edson (2014)? If an object in a museum were to offer multiple perspectives for its visitors, then the museum could scale horizontally across multiple disciplines.

3. Deep personalization

Customization refers to the visitor’s ability to select what is desired. Within a business environment, Pine (1992) introduced the idea of mass customization, which charted a new trend by allowing buyers to customize products. For example, a customer can customize the look of an automobile’s interior, as well as some aspects of its exterior. An extreme example from daily life is choosing to make your own pizza by specifying every ingredient via a Web page. One could go even further to suggest that one must first build the pizza oven. When Pine wrote about mass customization, this ability to customize was not as prevalent as in today’s Web, since the Web was in a fledgling state.

The challenge with customization is that it can be arduous for the consumer, or visitor in the case of a museum, to have to manually select everything to be delivered. This is where personalization raises new opportunities. Personalization is defined by a system that considers your personal profile and provides suggestions as to what type of pizza you like based on prior purchases or tendencies already published in a social media site like Facebook. Customization and personalization are part of what we refer to as the personalization continuum. Total personalization requires all “push” on the part of the organization delivering goods, whereas total customization requires total control, or “pull,” from the client. In this article, to simplify wording, we’ll refer to the entire continuum as personalization, since in practice, there is often a combination of both personalization and customization in any Web interaction. Operating at the poles is fraught with issues—nobody wants to have to specify each and every pizza, nor do they want to always be told what they are getting without having some choice.

Horowitz (2013) offers an account of taking the same walk in Manhattan, but through the eyes of eleven experts, and highlights the importance of personal experiences and knowledge based on the same “path traveled.” Each expert views the same walking path differently. For example, an expert in typography will attend to every sign, and a geologist will notice outcrops in Central Park and building materials. The paths walked by Horowitz’s experts are full of object-based experiences. Horowitz’s survey of experts is distinctive because of its diversity of experts. Contrast this with Ways of Seeing by Berger (1972), which contains different critical approaches to “seeing,” and yet the ways fit primarily within discourse involving cultural critique; where are the ways of seeing guided by mathematics, science, and engineering? Berger’s essay is apropos, since there are many ways of seeing within a museum. If we are to see things differently, perhaps guided in part by the STEAM philosophy, we should spread our diversity in how we view the world. This diversity must include an effort to span curricula from substantially different subjects. The issues of diversity and multiplicity stem from an ancient philosophical debate regarding monism versus pluralism. James (1909) provides his view of a “pluralistic universe.” Turkle and Papert (1990) discuss the need for epistemological pluralism with respect to issues arising from considerations of gender and computing.

Within museums, personalization has been described as a desirable goal. Museum objects, regardless of the type of museum in which they are housed, are discipline-neutral; the objects have many stories to tell, depending on what is desired by each visitor. As Falk and Dierking (1992) discuss, the “personal context” is essential if museums are to connect and engage the visitor. Falk (1999) portrays museums as places where personal learning is natural. He observes a key problem with a “museum’s community historic inability to document the educational impact it has on its visitors.” This inability is two fold: the museum does not 1) trace back its informal experiences to formal education, and 2) make the connection explicit. An example of an explicit connection is Mary learning about the geology of the prairie because she bookmarked it while at a painting. The problem Falk introduces will require museums to have personally relevant content for the visitor, and explicit traces back to formal learning. This solution might sound like a burden for the museum, but technologies and museum practices exist now for this personalization to occur.

Bowen and Filippini-Fantoni (2004) extensively document the history of museum-related personalization with regard to the evolution of the Web. They discuss what it means for a museum to engage in customization or personalization (what we are terming deep personalization for convenience). For example, New York’s Metropolitan Museum of Art has a Web calendar that is connected to the visitor through browser cookies.

Other work in personalization within museums includes Sparacino (2002), who creates a personal museum visit using wearable technology. Aroyo and Gorgels (2007) create a profile through a visitor-initiated ranking system (e.g., number of stars). Lord (2007) highlights the inevitability of the personal experience by observing that “even when curators establish a highly structured context for the interpretations of the collections on display, visitors remain stubbornly free to take from the exhibit what they choose.”

Villaespesa and Tasich (2012) describe a spreading “culture of analytics” at the Tate Museum. In DMA Friends (Stein & Wyman, 2014), the visitor to the DMA becomes a “DMA Friend” by registering. This registration creates a visitor profile, which is stored in the DMA’s databases. This profile would not contain information about what Mary is studying in school, but the profile does capture basic information to get a sense of the visitor’s demographics. The visitor provides information and preferences and, while doing so, obtains credits that can be exchanged for tangible benefits like discounts at the gift shop. DMA Friends, which currently has over one hundred thousand registrations (Stein, 2016), is a significant move toward providing creating visitor-centric analytics for the museum. The visitor’s experience is made personal, and the museum learns about its visitors.

These efforts are part of an evolutionary path for museums in personalization, and yet much more work needs to be done. In particular, the sort of personalization available through topic filtering is what we are creating at the University of Texas at Dallas (UTD). Topic filtering is where either the user or machine-learning tool selects what topics are of interest. Online news and magazine media are rife with reader-to-topic matching procedures. A reader may pick topics of interest such as health, science, geology, arts, and astronomy. Content is then filtered through these topics. Commercial phone and tablet apps such as Flipboard and News360 are examples, but many websites and services provide topic filtering. Pariser (2012) expounds upon the limits, and corresponding critiques, of personalization. Pariser carefully documents the current trends in personalization, and how they are pervasive on the Web. He notes that the top Internet businesses (e.g., Facebook, Google, Amazon) are based squarely on personalization as a core business strategy. Personalization, and most Web interchange, is a tradeoff between user benefits versus loss of some aspects of privacy. Total personalization is to be avoided, and there should be transparency in personalization so that users, readers, and potentially museum visitors will know how the filtering works.

Deep personalization for museums represents a goal where the museum knows as much as the visitor is willing to provide. The level of detail for this level of personalization could be enhanced through explicit social media partnerships (e.g., Facebook and Google+). Short of links with popular social media sites, finding out basic visitor demographics and what topics interest the visitor are desirable approaches. Simon (2010) details the importance of personalization in the chapter “Participation Begins with Me.” How can a museum connect on a personal level with the visitor? For example, having a small number of topics for the visitor to choose is a practical way to achieve profiling without getting into advanced algorithms in machine learning. Example topics involve visitor choice of Science/Art or Science/Technology/Engineering/Art/Mathematics (STEAM). Museums need this level of personalization knowledge to 1) create explicit content-based ties with formal education and 2) enhance the visitor’s experience of the cultural object, and not just present a single topical angle for that object.

In education, personalization is frequently positioned by categorizing, clustering, or grouping the learner. For instance, Gardner’s multiple intelligences is frequently cited as catering to different types of learners (Gardner, 2008). This clustering approach is similar to a museum’s categorization approaches (e.g., Pitman & Hirzy, 2010). While the categorical methods involving groups of people are useful, they do not address the individual learner. Bray and McClaskey (2015) observe that personalization is separate from individualized or differentiated instruction, where the learner is not in “the driver’s seat.” Within specialized areas such as Algebra I, personalization has yielded positive learning outcomes (Walkington, 2013) in contrast to a non-personalized approach.

4. Case study: Systems thinking in an art museum

Words such as system and model evoke different reactions and interpretations depending on discipline and context. For our purposes, we define a system as an abstract mathematical structure that contains sub-components, each of which may contain further sub-components. Each component is connected with other components. While this definition may seem broad, it captures two types of systems that we cover: information and dynamic. An information system is a structure about information or knowledge. Concept maps and semantic networks, to be described, are information based. A dynamic system is a structure that defines how something works over time. Systems are defined through the creative activity called modeling. An execution of a dynamic model is called a simulation.

Fishwick provides a comprehensive description of modeling and simulation methods (Fishwick, 1995), with a systems focus, and also discusses the notion of connecting system models to physical objects (Fishwick, 1998; Cubert & Fishwick, 1999). Systems thinking provides a way of thinking about STEM “in the large,” since each of the four disciplines employs models and identifies systems.

The art museum was chosen as the venue for considering systems thinking in a Fall 2015 class in Modeling and Simulation. Students were each given a choice of an object at the DMA. With some guidance, they interpreted these objects through thinking of them from a systems perspective. The guidance consisted of heuristics such as: (1) represent knowledge about the objects and their representations, resulting in a concept map; (2) consider any processes or techniques associated with the object, what is represented in the object, or in the object’s material; and (3) model the object with digital or physical materials. Systems thinking is atypical in an art museum, which is why it was chosen. The goal was to illustrate variety in object interpretation that ventured beyond art history explanations.

Systems thinking represents only one case study. Other case studies might be to analyze objects from other perspectives (e.g., physics, chemical, material, mathematical, engineering). The case study extends beyond the object to the system enveloping the object. The study also includes that which might be represented in an object. Consider the Inca tunic in figure 1, which was highlighted within a recent exhibit (DMAInca, 2016).


Figure 1: tunic with checkerboard pattern and stepped yoke. Image courtesy of DMA.

Information on the DMA website includes the fine weaving (qompi) exclusively for Inca nobles. Numerous questions can be asked about this work, from how it was made to how it was worn and who may have worn it. In art education, students learn to create a “concept map.” Concept maps, as defined within art education, begin with identifying an object and then clustering questions around that object by theme (APAH, 2012, 2015). The art-based concept map is fairly flexible since it can be used by teachers and students. Rao et al. (2015) describe an online system for this type of concept mapping that can be used by Web visitors. The “concept map” here is a special case of the more generalized concept map as defined by Novak and Gowin (1984). Figure 2 illustrates a concept map for the Inca tunic. This map is a directed graph (i.e., a network of nodes with arrows specifying a direction of a relation between two nodes).

Figure 2: Concept map for the tunic organized around three themes: social technique, and study

Figure 2: concept map for the tunic organized around three themes: social, technique, and study.

The Social concept identified in the upper left part of figure 2 involves two questions: 1) who wore this tunic? and 2) How did the Inca interact with each other during its fabrication? The question in the Study concept involves art practice in the form of sketching, but it could as easily have been “Can you code the pattern using Processing?” where Processing is a Java-based computer programming language (Greenberg, 2008). The type of study would be based on the visitor’s interest to use charcoal, to code using a computer program, or both. These are questions based on the visitor’s interpretation of the artifact by way of creative reconstruction.

Thinking in terms of concept maps is a type of systems thinking during a conceptual stage of model development. The word “model” is often used interchangeably with diagrams created for studying systems. The historical approach of diagramming rather than prose is closely aligned with engineering practices (Ferguson, 1992). The practical use of diagrammatic thinking of this type is also present in primary and secondary education using the phrase “graphic organizer,” and in Universal Design for Learning. Systems thinking refers to looking at an object as a collection of interdependent parts. When we think from a systems perspective, the tunic has many interpretations involving different processes: 1) how it was installed in the exhibit, 2) how it was made, 3) how components (e.g., red dye) were made, 4) how people interact with it in a gallery, and so forth. Rather than use natural language to answer “how” questions, the systems modeler frequently employs diagrams. We can think of these diagrams as belonging to the Semantic Web of the Inca tunic. The questioning is still centered on the tunic object, but the object becomes part of a complex web of meaning, process, and object. A next step after raising questions about the tunic is to identify concepts learned while exploring the gallery space (DMAInca, 2016). Sample concepts are shown in a “concept cloud” (figure 3).

Figure 3: A collection of disconnected concepts obtained in the gallery and while exploring the Inca collection

Figure 3: a collection of disconnected concepts obtained in the gallery and while exploring the Inca collection.

Figure 4 takes a subset of the concepts, makes connections among them, and then labels these connections (i.e., relations). This process goes beyond the level of formalization in some concept maps used in art education because each arrow is labeled. Without arrow labels, the relationship between any two concepts is ambiguous. There are two sub-figures in figure 4: on the left of the arrow is represented the introduction of arrows. On the right side of the arrow, these arrows are labeled with their semantics. For example, the relationship between the concepts “Inca” and “culture” is that Inca is a “type of” culture.

Figure 4: Evolving a concept map by starting with Figure 3 and then creating connections, and then labeling each connection

Figure 4: an evolving concept map, starting with a subset of figure 3 concepts (left side) and then creating labeled connections (right side).

In figure 5, we continue this diagram construction process by including some images of select concepts and extending the map.

Figure 5: Extended concept map of knowledge about the Inca tunic. Images: tunic image is from the public DMA collection. The map in the upper left is from Wikimedia Commons: public domain. Remaining images from Shutterstock, Inc., standard license

Figure 5: extended concept map of knowledge about the Inca tunic. Images: tunic image is from the public DMA collection. Map in the upper left is from Wikimedia Commons: public domain. Remaining images from Shutterstock, Inc., standard license.

Figures 3 through 5 represent the beginnings of thinking as a system scientist about the tunic, with the first steps being to identify the components of the system, and subsequent steps being to clearly identify the relationships among those components. In figure 5, there are a few notable points: the tunic is at the center, all relations are specified, and in one instance a URL is denoted. Something similar to figure 5 could be encoded within the various languages of the Semantic Web and connected to the DMA’s database, although this is a future step. An open question is who creates the map: the curator, a teacher, or a student? The map (i.e., model) can be created by all three types of people.

The history of systems thinking of this sort began with Bertalanffy (1969) and is now so wide-ranging that the idea of “system” is pervasive in all STEM fields. Meadows (2008) captures systems thinking from Jay Forrester’s early work, where Forrester identified an approach where one began to model a system in terms of a “causal graph.” Forrester coined this engineering approach to systems, “System Dynamics,” not to be confused with the general idea of the dynamics of systems. A causal graph begins with concepts and relations that denote explicit causality. For example “if there is smoke, there is probably fire” could be diagrammed with an arrow from smoke to fire. From the causal graph, there are other stages, ultimately ending in a set of equations, specifying change of system state over time.

The next step in seeing the tunic from the lens of systems thinking is to map out the dynamic relations. We do this by focusing on verb-based relations in English. The diagram in figure 6 represents a finite state machine (FSM, 2016), as it is termed in computer science. Each state has a participle verb form indicating state. For example, to craft a tunic, we begin by shearing an animal from the camelid family, such as an alpaca. Thus, the system that indicates how the tunic is made can be seen as a sequence of activities (i.e., states) of different people in a sequence-based pipeline.

Figure 6: Four connected states comprising a finite state machine (FSM) for the tunic process. Images: Shutterstock, Inc., standard license

Figure 6: four connected states comprising a finite state machine (FSM) for the tunic process. Images: Shutterstock, Inc., standard license.

The arrows have a simple label of an asterisk (*), which means “next in sequence.” First, there is shearing of the alpaca, then spinning the wool, followed by dyeing some of the wool, ending in the weaving activity. Model engineering (Fishwick, 1989) represents the iterative process of developing models.

In our systems thinking evolution of models, we started with concept maps, such as the ones in figures 3 to 5, to arrive at figure 6. This evolution can involve two possible sub-paths: one can 1) derive the FSM from preexisting verb-based connections in the map or 2) start with a clean slate by focusing the design by starting with verbs. This difference amounts to having a different schema or frame when designing the models. Either we use derivation completely or begin our design with a set of process-based constraints. For figure 6, this FSM requires a frame that involves activities or states, listing them, and connecting them together as appropriate to accurately match a known process at some level of abstraction.

Figure 7 is a model that is slightly different than the one in figure 6. Figure 7 shows a set of functions (in white boxes), or processes, each of which takes input and produces output. The graph in this image is sometimes referred to as data flow and the FSM in figure 6 as control flow. This differentiation is because the flow in figure 6 represents flow among activities, whereas in figure 7, material flows through each node. Observe that control flow diagrams have states that are associated with images of people performing activities, whereas for data flow, the emphasis is on the flow of material (encoded as data for purposes of simulation).


Figure 7: a data flow graph that represents material flowing from left to right. Each node is a process, as indicated by a verb. Images: Shutterstock, Inc., standard license, with the exception of the S/Z image (public domain, Wikimedia Commons) and the tunic (courtesy of DMA).

Starting on the left of figure 7, an alpaca is sheared. In a more detailed model, there would be an arrow input to “shear,” but this is left out for simplifying the diagram. There are two outputs from shear: one going to the wool, which subsequently must be spun, and another representing the alpaca minus the sheared wool: the shorn alpaca. Spinning can be done in one of two directions (termed S versus Z ply). Diagrams such as figures 6 and 7 are common in computer science, where they are used to model software prior to code development. Modeling techniques that are similar in structure are used widely in other science and engineering disciplines. These are ways to think about the tunic from the standpoint or perspective of STEM. Having these diagrammatic explanations of the art represents a personalization, targeting the visitor’s STEM background or interest.

The reader may wonder when these diagrams are most appropriately used in a museum. Diagrams tend to convey information to the visitor quickly. Contrast this information delivery against using only English or using video media. An English-only description takes time to read, whereas diagrams are easily and quickly assimilated. Similarly, a video takes time to watch, and visitors may not have this time given their short object-oriented attention spans (Pitman, 2016).

The diagrams represent models, and each model type is a formal language. The language of figures 2 through 5 is the concept map, which is a model of knowledge. Figure 6 is a finite state machine, which is a model of an object’s behavior or dynamics. Figure 7 is a data flow model, which also models dynamics. By viewing these models, the visitor learns not only something specific about the Inca tunic, but also abstract, formally defined languages. Table 1 explains the model languages by name for figures 2 through 7. Models can express behaviors and information that may otherwise be obscured by natural language. For instance, the finite state machine can be used to model everything from alarm clocks to cooking recipes. Modeling concepts reflect skills that can easily be transferred to arbitrary museum objects. This transference gained through new language understanding is what is meant by systems thinking. The visitor learns about the languages of systems thinking, which is just as important as learning about the Inca since these languages are widely applicable for all objects in the museum’s collection.

Figures in paper Type Components Additional information
2, 3, 4, 5 Semantic network Concept, Relation Concept map, mind map, logic diagram
6 Finite state machine State, Event Control flow diagram
7 Data flow network Function, Data Process flow diagram

Table 1: models of the Inca tunic

5. Bridging informal with formal learning

We’ve seen that the Inca tunic can be seen from a STEM perspective by choosing to examine the tunic using a systems thinking lens. This lens represents a type of personalization where visitors who have a STEM background are introduced to a STEM way of thinking about the tunic. This way of thinking would augment, not replace, an art historical perspective. This systems approach leads to a broader question about informal learning in a museum context versus formal learning in schools. What is the difference? One view is that the museum exists as an external physical resource for schools. While this has always been true for museums, this type of bridge is relatively weak. A deeper approach to personalization brings what one learns in schools (formal subjects) inside of the museum, while the visitor is attending to the object. This approach encourages thinking of museums and schools as having a wide, bidirectional conduit rather than viewing museums as purely destinations for informal learning field trips.

Learning through objects (Turkle, 2007; MacGregor, 2011; Chatterjee & Hannan, 2015) is fundamentally different than learning through a discipline. Disciplines are typically vertically oriented. This verticality matches the intellectual architecture of school curricula and the physical buildings found at universities. Connections can, and are, made between disciplines, but the student’s primary exposure to knowledge is by way of discipline rather than through personal experience. A student goes into building X to learn discipline X. However, museums contain objects, not disciplines. An object can be interpreted through a variety of lenses or disciplinary angles even though a museum may align itself with a particular one. Figure 8 displays two philosophical approaches to bridging a museum collection with formal learning. We encourage the method on the right, where a museum’s natural disciplinary alignment creates a priority for information about a discipline denoted in orange, and yet there remains a one-to-many relationship between object and discipline. This mapping is simplified since some museums explicitly combine science and art (e.g., the Exploratorium). Figure 8 illustrates that STEAM and the bridge from informal-to-formal education are sub-types of deep personalization as engaged by the museum visitor.

Figure 8: The traditional vs. integrative object-discipline mapping for museums

Figure 8: traditional versus integrative object-discipline mapping for museums.

There are two sub-figures in figure 8. On both sides, we have museums and disciplines mentioned. Museums contain objects, and disciplines are subjects. Objects have a one-to-many mapping to disciplines. The traditional approach is to map science (in a science museum) or art (in an art museum) to their component disciplines. For example, in a science museum, one is likely to learn about not only science, but also some aspects of engineering and mathematics. On the right side of figure 8, we have a goal of freeing the object to deliver information on many disciplines. The orange lines indicate natural disciplinary alignments. The blue lines show where connections can be made from type of museum to learned subject. Thus, it is possible to learn science in an art museum, and art in a science museum, since all objects are multifaceted.

The object contains information about virtually every subject since that the experience of the object is dependent upon the visitor’s personal profile. The goal of the museum is to encourage connections between the profile (i.e., what Internet-based companies such as Facebook and Google think of as “the model of the user”) and the object on display in an exhibit.

6. Architecture for deep personalization

Let’s discuss an architecture that can scale to any object. A museum visitor will require a profile. Complicated profiles can be gleaned by connecting a museum via a popular social media platform such as Facebook. However, our purpose to begin is to use STEAM as a categorical profiling mechanism. For example, a visitor can choose to know about art and engineering, in which case the visitor can be presented with a topic-selection mechanism, and they choose stEAm, with uppercase letters referring to desired subject matter content. As a pilot study, we will create a smartphone app and other phone Web connections to a collection of art donated by Joan Davidow within the Arts & Technology (ATEC) building (Davidow, 2016). The visitor will specify the profile by selecting which components of STEAM are desired on their phones. The main information on each artwork (title, artist, media) will be presented by default, and then the personalization occurs.

The visitor will be notified that they are near an artwork through the use of Bluetooth low-energy (BLE) beacons. Beacons have been explored, and experiences with them discussed, by several museums (e.g., the Brooklyn Museum). Our approach can build on Apple’s iBeacon as well as Eddystone of Google’s Physical Web initiative (Jenson et al., 2015). The smartphone, having Bluetooth-enabled communications, will then have an approximate idea of the visitor’s focus—which piece of art the visitor is near. The information presented is planned to be stored on a communal wiki. The wiki will include all available related knowledge, edited by the larger community. For the systems thinking diagrams (e.g., figures 2 through 7), these diagrams would be created on the wiki by someone who teaches a class in systems modeling. The class identification is placed on the Wiki, so there is a trace back to where the educational material is taught in a semester-long class. Material is expected within the university and outside as appropriate.

7. Broader implications for museums

This research identifies a nexus that is formed by three strands: 1) the emergence of Google’s Physical Web, 2) object-based learning and inquiry, and 3) deep personalization. Each strand arises naturally from recent trends. The Physical Web emerges from ubiquitous computing and the Internet of Things. Object-based learning is natural in a place that houses objects (i.e., museums) rather than disciplines (i.e., universities). Deep personalization takes its cue from theories of category and media consumption where topic or interest-based personalization has been the norm for at least a decade. The broader implications are that objects are not just things to collect and put on display. Objects provide ways of organizing and delivering multidisciplinary knowledge. Museums holding rare, and often unique, objects are in an ideal position to create broad communication and educational possibilities arising from this nexus.

The theme of “systems thinking in an art museum” conveys one story, whereas the broader story is that multiple-subject knowledge can be conveyed by any object in any type of museum. Museums evolved from the “cabinet of curiosity” (i.e., the kunstkammer) from the sixteenth through eighteenth centuries. The idea that a museum should be aligned specifically by discipline is a fairly recent phenomenon. For instance, the South Kensington Museum in London was split into two museums, the Victoria & Albert and the Science Museum, in 1893. Contrast this with the significantly earlier formation of disciplines, arising from the ancient Greeks (e.g., Plato’s trivium consisting of grammar, logic, and rhetoric). This is not to suggest that museums cease their fragmentation, as this occurs regularly in the academy at large, but rather that the objects contained within museum walls not be bound so tightly to a singular view that they lose their naturally diverse identities. Could the object-centered approach be a path toward seeing different views and different perspectives that surface naturally in the object-human relationship? If so, this approach represents a fresh means to connect disciplines in formal settings through bidirectional information traces to and from museums.

Museum staff profiles are as unique as the visitor profiles. Someone in the digital media department may need different information from an object than a curator. This difference can be made functional by extending the profiling capability when in the vicinity of an object. As with the recent trend of collaborative wiki editing in museums (i.e., edit-a-thons), curators might expand their domains to be curators of personal knowledge gleaned from different visitor profiles. The GLAM-Wiki initiative captures some of these editing practices. Curators will play key roles in deep personalization, while allowing for wider community editing and knowledge integration.

8. Mitigating risks

The first risk is quality control. Curators within museums will have natural concerns regarding quality control of community-generated information on objects. Some museums have created “Wikipedians in Residence” programs where content is permitted and encouraged from outside of the museum’s walls. This community spirit to content creation is necessary for deep personalization to succeed (Bowen, 2008). Wikipedians are motivated in several ways (Nov, 2007). If a wiki approach is used, and especially if multiple disciplines based on visitor topical profiles are instrumented, how are the curatorial departments to manage content? What about accuracy? For this reason, pilot studies are recommended so that the personalization approach is tested on one object or a small number of objects within a collection. Or the approach can be limited to one collection within one department. Accessing a permanent collection is desirable to encourage long-term knowledge acquisition gleaned from several disciplines. Another approach to mitigation is to separate the collaboratively edited wiki from the official object wiki page. The collaborative wiki would become a learning wiki linked from the main object Web page; the main object page undergoes more curatorial scrutiny, with the learning wiki being a mechanism for the community at large.

The second risk is credit assignment. This risk is best mitigated in two ways: to (1) encourage school administrators to “buy into” the idea of connecting to museums to encourage informal learning of multiple disciplines through objects, and (2) assign credit to instructors who create learning resources for objects. The instructors, their institutions, and their specific curriculum offerings should be included on the wiki page. For example, if Mary learns earth science while attending to the Benton painting, the earth science class number, name, and teacher should appear in the wiki. This explicit credit, along with the administration’s encouragement, should help motivate the content creators who can include created content citations on their professional resumes.

The third risk is user adoption. We are unsure how the system will be adopted, but plan to study user behavior. Some visitors may not wish to use technology, or they may not be in a learning predisposition while attending to objects. The adoption risk is not that different than the adoption of any technology within a museum, or external, setting. Mobile devices in the form of smartphones are being widely used. Another issue with adoption is related to self-directed learning. Some learners may require guidance prior to engaging in self-directed learning. This suggests a mitigation approach: having camps, exercises, or events where students use the smartphone application while inside a place of formal learning. Museums often have spaces for this type of exploratory piloting of technology.

9. Conclusion

During fall 2015, we had students create “systems interpretations” for DMA artwork, and we have recently created two Javascript-based smartphone apps that work with the Physical Web; however, the wiki development has just begun. Connections to formal learning classes have not yet been made. The personalization is being achieved by allowing the visitor to select specific subjects that interest them when experience a work. One of the major issues in the work is selecting the method of creating categories that visitors choose to customize their profiles. Web-based media aggregators use many types of topics to create reader profiles. Is a profile based on STEAM sufficient? STEAM is one way to categorize knowledge, but this way is influenced mainly by the STEM focus, with “A” being inserted for cultural diversity. Does STEAM reflect a global desire, or can other categorical systems be used? We are still investigating options and possibilities. A profile based on STEAM, at least within the United States, fits a recent “educational movement” toward a more diverse combination of discipline-specific interpretive approaches.

The five qualities of deep personalization are self-directed learning, topic profiles, diverse multidisciplinary breadth, community knowledge, and informal-to-formal knowledge mapping. Technology exists to support all five qualities. Deep personalization suggests museum object as Web interface. As such, technologies such as beacons and augmented reality provide visitors with a way to browse the emerging Physical Web—a web of knowledge centered on objects rather than on disciplines or through simulated pages on a screen.


I am most grateful to the anonymous reviewers of the original paper, and to the conference editors. They offered many valuable suggestions, and I have attempted to take these into account to produce a better, more concise argument. I would like to thank Bonnie Pitman (distinguished scholar in residence at UTD and former Eugene McDermott Director of the DMA) for our frequent discussions and collaborations on making connections with art. Thanks to Roger Malina (distinguished university chair, ATEC and professor of physics) for discussions around STEAM and Leonardo. I’d like to thank Rob Stein (former deputy director, DMA), who showed seminal interest in the project, with our shared goal of thinking using diagrams. Also, thanks to Shyam Oberoi (director of Technology and Digital Media, DMA) and Carlos Arroyo (senior software developer, DMA) for permitting my students and me access to the DMA Brain and its programming interface. Many thanks to Kimberly L. Jones (Ellen and Harry S. Parker III Assistant Curator for the Arts of the Americas) for her expertise in the ancient Americas, and in Inca culture specifically. The well-designed educational gallery located near the center of the Inca exhibit (DMAInca, 2016) was useful in understanding the processes and materials involved in Incan textile production. Also, thanks to Cassini Nazir (clinical associate professor, UTD) for his collaboration on issues of interaction design. Any errors made in this article are solely my responsibility.


APAH. (2012). Professional development, AP Art History: mapping concepts and themes in AP Art History. Curriculum module. The College Board, New York, NY. Consulted January 7, 2016. Available

APAH. (2015). AP Course and exam description; AP Art History, including the Curriculum Framework. Effective Fall 2015. The College Board, New York, NY.

Aroyo, L., & P. Gorgels. (2007). “Personalized Museum Experience: The Rijksmuseum Use Case.” In J. Trant & D. Bearman (eds.). Museums and the Web 2007: Proceedings, Toronto: Archives & Museum Informatics. Published March 1, 2007. Consulted January 6, 2016. Available

Berger, J. (1972). Ways of seeing. British Broadcasting Corporation and Penguin Books.

Bertalanffy, L. (1969). General systems theory: foundations, development, applications. George Braziller, Inc.

Bowen, J.P. (2008). “Wiki Software and facilities for museums.” In J. Trant & D. Bearman (eds.). Museums and the Web 2008: Proceedings, Toronto: Archives & Museum Informatics. Published March 31, 2008. Consulted January 8, 2016. Available

Bowen, J.P., & S. Filippini-Fantoni. (2004). “Personalization and the Web from a museum perspective.” In D. Bearman & J. Trant (eds.). Museums and the Web 2004: Selected Papers from an International Conference, Arlington, Virginia, USA, March 31–April 3 2004. Archives & Museum Informatics, 63–78.

Bray, B., & K. McClaskey. (2015). Making learning personal: the what, who, wow, where, and why. Corwin Publishing.

Chatterjee, H.J., & L. Hannan (eds.). (2015). Engaging the senses: object-based learning in higher education. Ashgate Publishing Company.

Cubert, R., & P. Fishwick. (1999). “Modeling the simulation execution process with digital objects.” In Proceedings of SPIE – The International Society for Optical Engineering, Vol. 3696: 2–22. Orlando, Florida.

Davidow. (2016). Davidow Collection from Joan Davidow. Consulted January 9, 2016. Available

Dewey, J. (1934). Art as experience. Republished in 2005. TarcherPerigee Publishing.

DMAInca. (2016). “Inca: Conquests of the Andes/Los Incas y las conquistas de los Andes.” May 15–November 15, 2015, Chilton II Gallery. Curated by Kimberly Jones, Dallas Museum of Art. Consulted January 7, 2016. Available

Duchamp, M. (1957). The creative act. Consulted February 1, 2016. Available

Dudley, S. (2013). Museum materialities: Objects, engagements, interpretations. Routledge Publishing.

Duncan, C. (1995). Civilizing rituals: Inside public art museums. Routledge Publishing.

Edson, M. (2014). “The challenge of scaling up: museums in an age of scale.” Building the future of education; Museums and the learning ecosystem. Center for the Future of Museums.

Falk, J.H. (1999). “Museums as institutions for personal learning.” Daedalus 128(30): 259–275. The MIT Press.

Falk, J.H., & L.D. Dierking. (1992). The museum experience. Left Coast Press.

Ferguson, E.S. (1992). Engineering and the mind’s eye. MIT Press.

Fishwick, P. (1989). “Studying how models evolve. An emphasis on simulation model engineering.” In Proceedings of the SCS Multiconference on AI and Simulation, Society for Modeling and Computer Simulation International 20(4): 74–79.

Fishwick, P. (1995). Simulation model design and execution: Building digital worlds. Prentice Hall.

Fishwick, P. (1998). “An architectural design for digital objects.” In WSC ’98. Proceedings of the 30th conference on winter simulation, 359–366. IEEE Computer Society Press.

Fishwick, P., B. Pitman, & C. Nazir. (2016). “Models of X.” University of Texas at Dallas. Consulted January 8, 2016. Available

FSM. (2016). “Finite state machine.” Wikipedia. Consulted January 7, 2016. Available

Gardner, H. (2008). Multiple intelligences: New horizons in theory and practice. Reprint edition. Basic Books.

Greenberg, I. (2008). Processing: Creative coding and computational art. APress Publishing.

Horowitz, A. (2013). On looking: Eleven walks with expert eyes. Simon & Schuster.

James, W. (1909). A pluralistic universe. Hibbert lectures at Manchester College on the present situation in philosophy. Project Gutenberg, 2004. Produced by F. Urbanski, D. Starner, N. Hayes, and the Online Distributed Proofreading Team. Consulted February 29, 2016. Available

Jenson, S., R. Want, B.N. Schilit, & R. Kravets. (2015). “Building an on-ramp for the Internet of Things.” In IoT-Sys 2015, May 18, 2015, Florence, Italy, Association for Computing Machinery.

Leddy, T. (2012). The extraordinary in the ordinary: The aesthetics of everyday life. Broadview Press.

Lord, B. (2007). “What is museum-based Learning?” In B. Lord (ed.). The manual of museum learning. Altamira Press, a division of Rowman & Littlefield Publishers.

MacGregor, N. (2011). A history of the world in 100 objects. Penguin Books.

Martinez, M. (2002). “Designing learning objects to mass customize and personalize learning.” In D. Wiley (ed.). The instructional use of learning objects. Consulted January 7, 2016. Available

Meadows, D.H. (2008). Thinking in systems: A primer. Chelsea Green Publishing.

Muller, L. (2007). “The return of the wonderful: Monanisms and the undisciplined objects of media art.” In Studies in material thinking, Vol. 12 (Material thinking of display). Consulted January 28, 2016. Available

Nov, O. (2007). “What motivates Wikipedians?” Communications of the ACM 50(11): 60–64.

Novak, J.D., & D.B. Gowin. (1984). Learning how to learn. New York and Cambridge, UK: Cambridge University Press.

Pariser, E. (2012). The filter bubble: How the new personalized Web is changing what we read and how we think. Penguin Books.

Pine. B.J. (1992). Mass customization: The new frontier in business competition. Harvard Business Review Press.

Pitman, B. (2016). Personal communication.

Pitman, B., & E. Hirzy. (2010). Ignite the power of art: Advancing visitor engagement in museums. Dallas Museum of Art.

Rao, S., P. Edmonson, B. Baer, & A. Weinard. (2015). “Practical ontology: collaborating and communicating with concept maps.” MW2015: Museums and the Web 2015. Published January 31, 2015. Consulted January 7, 2016. Available

Simon, N. (2010). The participatory museum. Museum 2.0 Publishing.

Sparacino, F. (2002). “The museum wearable: real-time sensor-driven understanding of visitors’ intersts for personalized visually-augmented museum experiences.” MW2002: Museums and the Web 2002. Consulted January 6, 2016. Available

Stein, R. (2016). Personal communication.

Stein, R., & B. Wyman. (2014). “Seeing the Forest and the Trees: How Engagement Analytics Can Help Museums Connect to Audiences at Scale.” MW2014: Museums and the Web 2014. Published February 1, 2014. Consulted January 6, 2016. Available

Stock, P., & R.J.F. Burton. (2011). “Defining terms for integrated (multi-inter-trans-disciplinary) sustainability research.” Sustainability 2011 3: 1090–1113.

Turkle, S. (2007). Evocative objects: Things we think with. MIT Press.

Turkle, S., & S. Papert. (1990). “Epistemological pluralism and the revaluation of the concrete.” Journal of mathematical behavior 11(1): 3–33.

Villaespesa, E., & T. Tasich. (2012). “Making sense of numbers: a journey of spreading the analytics culture at Tate.” MW2012: Museums and the Web 2012. Published April 7, 2012. Consulted February 1, 2016. Available

Walkington, C.A. (2013). “Using adaptive learning technologies to personalize instruction to student interests: the impact of relevant contexts on performance and learning outcomes.” Journal of educational psychology 105(4): 932–945.

WF (Wallace Foundation). (2001). “Services to people: challenges and rewards. how museums can become more visitor-centered.” Published by the Wallace Foundation. Available

XChange (Nasher XChange). (2013). Nasher XChange. Nasher Sculpture Center.

Cite as:
Fishwick, Paul. "Deep personalization: A case study of systems thinking within an art museum." MW2016: Museums and the Web 2016. Published January 12, 2016. Consulted .