Enhancing experiential and subjective qualities of discrete structure representations with aesthetic...

ARTICLE IN PRESS

Journal ofVisual Languages & ComputingJournal of Visual Languages and Computing

16 (2005) 406–427

1045-926X/$

doi:10.1016/j

E-mail ad

www.elsevier.com/locate/jvlc

Enhancing experiential and subjective qualities ofdiscrete structure representations with

aesthetic computing

Paul Fishwick

Department of Computer and Information Science, University of Florida, Room E301, CSE Building,

Gainesville, FL 32611 6120, USA

Received 16 December 2004; accepted 10 January 2005

Abstract

The task of visualization, as it applies to computing, includes by default the notion of

pluralism and perspectivism since there is an explicit attempt at representing one, often textual,

interface in terms of a more graphical one. This desire for alternate, subjective perspectives is

consistent with art theory and practice, and even though rigor and formalism generally mean

different things to artists and computer scientists, there is room for collaboration and

connection by applying artistic aesthetics to computing, while maintaining that which makes

computing a viable, usable field. This new area is called aesthetic computing. Within this area,

there is an attempt to balance qualitative with quantitative representational aspects of visual

computing, recognizing that aesthetics creates a dimension that is consistent with supporting

numerous visual perspectives. We introduce one aspect of aesthetic computing, with specific

examples from our research and teaching to illustrate the potential and possibilities associated

with alternate representations of discrete structures such as finite state automata and a data

flow network. We limit ourselves, and our methodology, to model notations with components

that bear a largely symbolic connection to what they represent, thus providing greater degrees

of representational freedom. We show that by exploring aesthetics, we surface some important

philosophical and cultural questions regarding notation, which turn out to be at least as

- see front matter r 2005 Elsevier Ltd. All rights reserved.

.jvlc.2005.01.001

dress: [email protected].

www.elsevier.com/locate/jvlc

ARTICLE IN PRESS

P. Fishwick / Journal of Visual Languages and Computing 16 (2005) 406–427 407

important as the algorithmic and procedural means of achieving customized model

component representations.

r 2005 Elsevier Ltd. All rights reserved.

Keywords: Discrete structure; Modeling; Aesthetics; Customization

1. Introduction

There is a saying that goes something like this: I would rather see one sunset

through a thousand eyes than a thousand sunsets. This suggests that we can learnabout a thing by experiencing multiple perspectives, with different visualizations,materials and ways of crafting. This approach is generally borne out when studying apiece of knowledge that we are trying to acquire. If one wishes to delve into thetheory of groups in algebra, one is more likely to understand the material by readingdifferent books on the subject, by doing several examples, and by asking differentpeople from different backgrounds about their perspectives on groups. We wouldwant to see an example of a group, and its effects, in multiple ways: textually andgraphically. The ability to generate and explore multiple perspectives is closelyassociated with aesthetics [1].

The topic of Aesthetic Computing [2] relies on multiple perspectives fordifferent aspects of computing, from the mathematical foundations of computing(theory, discrete structures, program and data structures, database schemata,networks) to human–computer interaction and the topic of developing morediverse and customized [3,4], or personal, human interfaces. In our researchand teaching, we have focused on the representation of discrete structuresfound in the fields of computer simulation and software engineering. AestheticComputing is defined as the application of art theory and practice to computing.The underlying assumption is that by employing a diverse array of aestheticsand natural or artificial cultural artifacts, that we permit an exploration of differentviews for representation within computing. Before proceeding further, we shouldaddress what might be the first question for the Computer Scientist: namely, whythis area is new given that we have always had aesthetics in both mathematicsand computing.

To answer this requires some background on aesthetics, for the term has meantdifferent things to different people depending on when it was used. In ancientGreece, aesthetics were synonymous with the principle of mimetics—that art wasbest when it resembled a target object. During the same time period, Plato inventedthe forms [5], which stood at the top of a hierarchy, representing the pinnacle ofknowledge. Leaping forward to the 18th century, Kant [6] reinvigorated a kind ofPlatonic view of aesthetics with the idea that disinterest and a cognitive reflection,and not so much visual appearance, were the key aspects of aesthetics. Today, themajority of aestheticians would agree that art employs a balance of mind and matter,of cognition and body. Moreover, aesthetics and emotion have always been closely

ARTICLE IN PRESS

P. Fishwick / Journal of Visual Languages and Computing 16 (2005) 406–427408

aligned [7] even though the philosophy of expressionism relies on emotion as adefining characteristic of art [8].

Returning to our topic, mathematics has often been treated as a creature of themind, with the emphasis being less on the senses and more on cognitive activity.Thus, an elegant proof entails a minimal number of sentential connectives, but manymathematicians would not view notation in the same light; however, there isevidence that mathematicians create significant visualization, at least internally,when posing and solving problems [9]. In a Platonic fashion, the notation is said toexist simply to shed light on the greater truth. And yet, the phenomenological aspectsof mathematics (i.e., what happens when one solves a system) has a strongconnection with art [10]. For example, the theory of knots can yield marblemanifolds, and while these sculptures do not represent the underlying topologicallyfounded notation, they do a remarkable job of appealing to our senses, andproviding a few more eyes through which we view the sunset.

With regard to computing, which can be seen as a significant outgrowth ofmathematics, Knuth has written of the importance of the material aspects ofaesthetics within his thrust area of literate programming [11]. Much of the reasonfor his creation of the TEX typesetting language, resulting in the popularLATEX spinoff, seems to rest on his definition of aesthetics in style andpresentation, which ventures beyond the purely mental into the material realm.For example, in the Metafont book [12], we find the following in the preface: ‘‘Typedesign can be hazardous to your other interests. Once you get hooked, you willdevelop intense feelings about letterforms; the medium will intrude on the messagesthat you read.’’ Knuth was one of the first to make this connection within ComputerScience; however this kind of thinking was also at the heart of Marshall McLuhan’sdictum ‘‘The medium is the message’’ [13]. McLuhan’s main thesis was thatinformation was not only a matter of denotation, but that connotative effectsof the material are equally as essential and influential for the task of humancommunication.

Recently, there has been significant research in visualization for both softwarenotation as well as execution [14–18]. This work builds on top of the almostubiquitous flowchart graphics so common in the 1960s and 1970s, but creates asubstantial discipline out of it. However, the range of aesthetics and styles inrepresenting software and execution results is rarely tied to art by allowing artisticpractice, or an associated focus on customization and personalization, to play amajor role. The recognition of domain-specific visual languages as a distinct area ismore recent, with workshop proceedings serving as the major dissemination outlets.Esser and Janneck [19] provide a conceptual treatment based on a definition ofdomain-specific languages as being ‘‘tailored to a particular problem domain.’’ ForAesthetic Computing, there is a distinct connection with this approach, except thatthe word problem can be replaced by cultural. The more abstract the formalism, thegreater the diversity possible in choosing cultural iconography to notate the formalelements. Aesthetics takes on the role of an extra dimension in the area ofInformation Visualization [20], with the possibility to represent graph structure, forinstance, with markedly different objects and styles. In doing so, the role of

ARTICLE IN PRESS


aesthetics in visual computing is to allow for design variety, and to draw the user in

toward the design, while providing motivation for ‘‘living with it’’ [21].To summarize the role of art theory and practice in computing: (1) in

mathematics, the emphasis has generally been on representation within the solutionspace rather than the problem space, and (2) in computing, while diagrammaticmethods are common in both problem and solution spaces, the effects of artisticapproaches are less often conveyed. Therefore, there is considerable room forexploration by creating numerous bridges between art and computing.

Our purpose is to define aesthetic computing, show examples, and discuss the keyissues that center on the technique. The paper is meant as an introduction toaesthetic computing, and its relation to visual computing, with other publications[4,22] serving to define the algorithms and technical approaches required to achievethis goal. The notion is that problems in visual computing are not solely of atechnical nature. The thesis is that we may, as a visual computing community, needto examine non-technical, aesthetic, cultural and philosophical assumptions aboutthe way that we represent and notate software and other formal structures.Returning to the issue of the word ‘‘aesthetic,’’ the goal of aesthetic computing—aswe practice it in our teaching and research—is not to make the presumption thatthere exists an optimal aesthetic or style given a species of human interface problem,but to underscore the importance of allowing users to choose their own styles, andthus, apply their own aesthetics to the representation of formal structures. Thisextended view of aesthetics [23,24] as a matter of choice is influenced by language andthe arts, and not only by an idealist perspective that considers ‘‘aesthetic’’ to be acomputed artifact resulting from pre-specified design criteria. In some cases,indeed, a circle may be more appropriate than a rectangle, or the applicationof a flow analogy may be more appropriate than use of an agent-based analogy, butthere is considerable room to allow aesthetic computing to be driven bycustomization and personalization in much the same way that these qualities driveproduct design [25,26]. This approach, then, suggests that formal structures are

products and subject to the same issues of representational convenience and choiceprovided by vendors of physical products. Moreover, we are attempting to promotea situation where the larger definition of ‘‘aesthetic’’ as used in fine art, is one equallyapplicable to computing. In particular, within art, the concept of ‘‘aesthetic’’ isclosely aligned with artistic style, genre, or theme in addition to its otherconnotations [23].

Some of the new directions that we might take, suggested by aesthetic computing,are created by questioning assumptions. we proceed by starting with a simpleillustration of aesthetic computing, and then ask key questions identified through theclassic questions What?, Why?, and How?. We then close the discussion with a briefdiscussion of another example, and summarize where we are in this area, and wherewe might venture in the future. The examples and explanations are oriented towardone part of aesthetic computing that interests our research lab, rather than trying toprovide more general coverage [2]. This part centers on representing formal elementsin computing and mathematics, with a concentration on elements used in computersimulation and programming.

ARTICLE IN PRESS


2. Example: representing a discrete structure

In computing, the finite state machine (FSM) is a ubiquitous discrete structure,found in lexical scanners for language parsers and scripting languages, and inbehavior encodings for artificial agents in interactive games. Let us consider thefollowing Moore machine M:

�
M ¼ hQ; I ;O; d; li; � Q ¼ fS1;S2;S3g; � d : Q � I ! Q; dðSi; 0Þ ¼ Si for n 2 f1; 2; 3g; dðS1; 1Þ ¼ S2; dðS2; 1Þ ¼
S3; dðS3; 1Þ ¼ S2;
� I ¼ f0; 1g; � l : Q ! O; lðSnÞ ¼ Sn; n 2 f1; 2; 3g:
The machine has 3 states (S1; S2; and S3) with an input value of 1 achieving achange in state. An input of 0 leaves the machine in the same state. The machineoscillates between two states S2 and S3 after it gets a jump start from S1; and asubsequent stream of ones. This definition is compact and has a typographicpresentation, which is amenable to most input and output devices. It has the distinctadvantage of having been formulated centuries ago, first with wooden matrices, andthen with metal, and more recently with computer typesetting hardware andsoftware. An important aspect to visual computing is that we need not toss out onerepresentation to make way for another. We appreciate the power of the typographicnotation, while at the same time recognizing that notation is largely at the mercy ofwhatever technologies exist to support it. Without the technology for making paperor parchment, one is limited to scratching marks on minimally processed surfaces.

Now consider Figs. 1(a)–(d), which illustrate four other views of this particularFSM. Each of these views can be considered graph visualizations, as M is a directedgraph. The illustrations are not meant to be complete diagrams with all necessaryinformation required to formally specify the static FSM structure, but rathersnapshots of interactive interfaces suggesting alternate presentations. Thus, oneshould not expect these figures to carry the same static information content as thedynamic, interactive counterpart which cannot be inserted into a print medium.Transition inputs of M are assumed to labeled with 1 with reflexive transitionstriggered by a 0. Given these assumptions and this class of machine, consider Fig.1(a), representing a 2D diagram of M. In a related logic-based diagrammaticapplication, Shin [27] demonstrates that iconography can be every bit as sound ascomplete as the typographic presentation. One infers that the reason why thediagram is not as prevalent as type is due primarily to economy: as displays becomecheaper and more plentiful, the reasons for using diagrams, and more generallyvisualizations, become less questionable and increasingly justifiable. Such asentiment, or perhaps leap of faith, needs little encouragement within the computervisualization communities. Still, we gain greater recognition that our modes ofpresentation are governed by the economy of materials and labor. This suggests thatas these economies shift, that our notations need to evolve, and that we have the

ARTICLE IN PRESS

Fig. 1. Four different visual perspectives (i.e. presentations) of machine M. (a) 2D diagram, (b) spheres,

(c) fluid flow metaphor and (d) agent metaphor.


opportunity of exploring alternate evolutionary paths, while not discarding existingrepresentations. Despite the apparent familiarity with Fig. 1(a), there is a significantamount of analogy being used, from a learned knowledge that a circular boundarydemarcates a ‘‘state’’ and that a curved line with an arrow on the end of it has aspecific meaning with regard to a transition starting at one state and terminating atanother. There is no inherent reason, for example, why such arrows could notinstead suggest that data or control information actually be sent beyond the boundsof the circle at the arrow’s head. The analogical mapping is learned, which isdiscussed at more depth in Section 5. Fig. 1(b) is made from 3D primitives, echoingthe sort of research studies by Lieberman [28] (i.e., program execution) and Najork[29] (i.e., program structure). In particular, Fig. 1(b) is similar to Najork’s CUBEelements except that several primitives are used (i.e., sphere, cylinder, and cone).Spheres represent states, in Fig. 1(b), and cylinders and cones combine to form 3Darrows. Apart from being the first 3D programming language, CUBE also

ARTICLE IN PRESS


introduced new programming concepts (i.e., combined data flow/logic capability andstatic polymorphic typing) to the evolving area of visual programming. As withShin’s work, Najork is concerned with issues of soundness and completeness withinthe visual framework, ensuring that formal attributes of textual languages are carriedover into the visual domain—in this case one with 3D primitives.

Fig. 1(c) goes a step further than Fig. 1(b) by employing analogy and metaphor,allowing us to take advantage of our knowledge of fluids, pipes, and containers inrepresenting the FSM operation. As with any metaphor, we must carefully qualify it,taking from it items that we wish to transfer while ignoring, or abstracting away,other items. For example, even though the FSM uses discrete inputs to drive statechange, the Virtual Reality Modeling Language (VRML) animation associated withFig. 1(c) shows a continuous fluid animation to capture this discrete change: the fluidnature of water is used to capture the idea of transportation, but we are to ignore

speed or time needed for fluid transfer. Hesse [30] draws attention to this distinction,with definitions of positive, negative and neutral analogies. Positive analogies areattributes that carry through from source to target, like the directionality inherentwith fluid flow. Negative analogies refer to attributes that we are to ignore whenforming the connection. Even though, this could be seen as an inherent weakness ofmetaphors in general, the natural affinity humans have for analogy and metaphorallow us to work past these constraints. That is, the medicine of metaphor may havesome negative side effects, but its overall benefits outweigh these effects, and aresubstantial [31,32]. The ability to interpret Figs. 1(c) and (d) as discrete transitions islearned, and represents an arbitrary convention that can be associated with the visualencoding. As long as it is made clear that FSMs, for a particular group advocatingthis visual style, are defined in a consistent manner with the semantics of discretetransition, there seems to be no significant issue. There are numerous cases of 2Darrows being used by different visual dynamic models [33] to define both continuousand discrete transitions, so even the meaning of a simple arrow is by no meansobvious or natural.

Fig. 1(d) is an agent-based metaphor, where gazebo-like structures represent statesand a woman walks along a lamp-lit pathway to move from one state to another.Internal lamps within each gazebo light up when that state is active. This examplebrings out the idea that abstraction need not imply material or visual abstraction,since we can create human body models to do our bidding within formal structures,and that these non-minimal, visual structures are every bit as abstract as theequivalent typographical presentation. The concept of abstraction is also integral tothe non-visual, referential attributes of an object. The abstraction is indicated by theone-to-many mapping associated with an element’s symbolic reference regarding itsrole as a state or a transition. Some issues surrounding the use of these models arecovered in Section 4.

Even though Figs. 1(b)–(d) are 3D, aesthetic computing as it pertains torepresentation is not solely focused on 3D. Instead, it is focused on customizationand the ability to easily explore a pluralistic set of forms. However, 3D seems to offermore opportunity in this regard than 2D, and so our models and methods tend to beoriented toward 3D models and animation. For example, with 3D, one may not only

ARTICLE IN PRESS


build familiar structures, but ones that have texture, materiality, and lighting effects.The use of 3D also allows the user to adopt a first person viewpoint with respect tothe representation, moving through it, becoming one with it [21], and thuspersonalizing the experience.

3. What?

What is aesthetic computing? Some of the context for the area can be found inmultimodels [33], which stress the idea of heterogeneity not only in formal modelcontent, and coupling, but in a plethora of presentations. In 1999, our researchgroup began a project called RUBE [34], to allow dynamic models (i.e., modelsspecifying the temporal dynamics of systems) to be simulated over multipleabstraction levels, and with arbitrary representations. The motivation for the latterconcept was derived in part by a new set of curricula we created called Digital Arts

and Sciences (DAS) [35]. DAS bridges two colleges, Fine Arts and Engineering, withthe idea of producing students who have a hybrid knowledge base allowing them tobecome ‘‘digital Leonardos,’’ but maintaining an academic, rather than vocational,orientation.

Some initial concepts in aesthetic computing were published [36,37], and aDagstuhl Seminar (cosponsored by Leonardo [38]) on the subject was convened insouthwestern Germany in July 2002 [39], with colleagues Roger Malina and ChristaSommerer. We had attendees representing a broad array of mathematicians,computer scientists, designers, and artists. A signed manifesto resulting from theworkshop will appear in Leonardo, and is included in the Appendix. An edited bookvolume is in progress [2].

Aesthetic Computing, as an emerging area, is broader than the examples shown inSection 2. As the field is defined by the application of art theory and practice tocomputing, researchers span a gamut from those with an interest in the nature ofaesthetics and semiotics, to interface and design specialists. The majority ofparticipants are concerned with aesthetics of the human–computer interface, somewith an interest in synthesis, and others with a more analytic orientation. Our work,as evidenced by the simple example in Section 1 is one aspect of aesthetic computing,and is driven by an interest in representing mathematical formalisms found incomputing: discrete structures, program, and data structures, and applied mathe-matics especially for dynamic systems typically employed for computer simulation.

4. Why?

Why do aesthetic computing? Consider a gedanken experiment where one is ableto conjure up any immersive holographic scene with advanced technological ease.Houses and landscapes, in such an environment, would be as easy to synthesize ascircles and squares in today’s computing environments, Two questions arise: (1)what could we do with such technology in terms of representing formal computing

ARTICLE IN PRESS


structure and behavior, and (2) what issues arise as a result? A conservative attitudemay be at the core of thinking that with this advanced technology, we would still useonly typography, draw circles, or mold spheres, but this seems unlikely when weconsider technology trends from ancient to modern times. The art of representationis certainly constrained by technology, but it is worthwhile considering the role ofabstraction, since this is at the heart of several issues regarding aesthetic computing.Are we in violation of principles of economy and abstraction when artisticallyrendering formalisms? Is it always better to tend toward the minimal when doingcomputing?

Let us try to address these questions by first considering that abstraction is a formof information reduction. One can take an object for ‘‘tree’’ and imagine that theobject refers to the set of all trees that exist in the world. Mathematical concepts ofnumber, set theory, and algebra all leverage abstract thinking. However, this sort ofabstraction is different than material reduction associated with using a circle insteadof a sphere, or a sphere instead of a house. As long as the object in question can bevisually identified as a reference for a target object, we are likely to employ whatevereconomical tools are available to us—which is why we scratched marks in clay [40]rather than synthesizing more substantial objects in a Holodeck1 environment.However, the tendency for mathematicians and computer scientists to eschewsubstantial forms in lieu of small marks is strong, and has a long history. There maybe a feeling that economy of thought requires economy of representation down tothe very atoms and bits used to capture what an object looks or sounds like. Even theera of abstract art seems to suggest that we employ minimal presentational forms,until we remember that the use of abstraction is at least as strong within thesurrealist genre [41,42], and in that movement, objects are not depicted with simplegeometric forms.

Abstraction may be one of the more salient aspects of aesthetic computing whencritically analyzing it, but there are more facets of motivation and agenda items onthe list of issues. Regarding motivations beyond the mental modeling exercise, onecan always point to the age-old desire to better connect art with science. Scienceaffects art, but in what ways does art affect science? For computing, new algorithmsadd to the painter’s palette or the sculptor’s hand tools, but aesthetic computingexamines the converse situation where art affects computing. The issue arising fromthis thought is whether we can maintain qualities we consider important incomputing as we mix artistic elements inside the computing barrel? After all, doesnot art have its own motivations and goals, with some of these being in directopposition-to or conflict-with computing? For answering the first part of thisquestion, it is worthwhile noting that the Greeks referred to art, craft, andtechnology under the same name: techne. It is only within the past two hundredyears that we have come to differentiate fine art from technological artifacts. It ispossible, and achievable, to craft computing formalisms with an eye to aesthetics,without sacrificing rigor, completeness, or soundness.

1The Holodeck is a fictional environment that first appeared in the popular science fiction show, Star

Trek: The Next Generation, produced by Paramount Pictures.

ARTICLE IN PRESS


One of the aspects of artistic practice, which makes it attractive to computing, is itsfocus on the individual and the cultural. Personalization and customization, despitevalid privacy issues associated with them, have become important attributes ofconstructivism [43] in education, and manufacturing [44] in the commercial sector.The idea of having many ways to represent something is usually viewed as a highlypositive concept, and this concept is quite consistent with the goals of art. Issues canarise based on who is doing the fine-tuning of the representation and what sorts ofinformation they have at their disposal, some of which may be deemed too personalfor some individuals. Like the issue of abstraction, the issue of standardization jumpsto the forefront in discussions of customization. How can we standardize notation ifeveryone fosters a Tower of Babel? The first answer to this is that we need noteliminate one representation in favor of another. We can mix and blend notationstogether, or use one notation when in one cultural setting, and another notation for adifferent group, or for a different purpose, such as learning rather thancommunicating. At least within the literature on dynamic systems [33], there are anumber of different visual modeling techniques, and yet we can all use the existingand powerful textual notation that has wide coverage and is well-understood bythose having studied grade school mathematics. This suggests that to explore newrepresentations, we should not discard existing, functional ones, instead we canaugment representations to allow for greater flexibility in representation.

Fig. 1(d) surfaces an issue regarding the variety of aesthetics and usability. If we lookat the lampposts that adorn the sides of the walkways, are these confusing orunnecessary? Decoration is consistent with many forms of art, and as long as theencapsulating forms are clearly identifiable with their symbolic references intact, thenwe should feel free to explore structures such as lamp-lined walkways. One needs toknow the rules by which models are made, and references defined. If all state transitionshave lamp-lined streets, and we are told so, then there may not be a problem; however,if some lights are different colors, we might inquire as to the information content beingtransmitted. We need to preserve the idea of rule, which could be seen as the definingcharacteristic of formalism. Issues of usability should be central to our task, but neednot dominate it to the extent that we revert to minimalist forms. We are not attemptingto find one holy grail for presentation, but rather to celebrate a large diversity ofpersonalized, subjective forms that can coexist with standards.

5. How?

5.1. Overview

How does one represent formal structure with aesthetic computing? We can lookat this from a theoretical perspective first, and then follow with a methodology andbrief discussion of our software implementation.

�
Representation: A review of traditional notation for mathematics and comput-ing, with examples drawn from algebra, discrete structures, programming, and

ARTICLE IN PRESS


dynamic systems. At this stage, students learn to begin with what they know, usingthe text-based notation that is familiar to them.

�
Mapping: A slew of techniques for mapping elements and systems to differentrepresentations, usually in target domains such as art, theatre, architecture, andmachines. We start with creativity exercises and games, and then move ontocomputational approaches for analogy and metaphor.
�
Crafting: The actual making of the aesthetic computing target object—forexample, a representation of a data flow network. This involves a discussion ofdifferent tools for 2D and 3D graphics, as well as sound editing. We discuss modeljuxtaposition techniques so that students can choose how to combine or blend amodel with the system being modeled. Automated layout techniques, whereappropriate, are presented.
�
Assessment: Gauging the quality of the product, ensuring that it meetsrequirements for items such as completeness, consistency, scalability, usability,and aesthetics.
5.2. Method

The method of aesthetic computing as practiced by our research group is oneof personalized representation, where one of a variety of ‘‘themes’’ is chosen. Forexample, Figs. 1(a)–(d), each represent a presentational theme for the sameunderlying FSM defined earlier. The problem of how to achieve this representa-tion is based on what is known about the source and target content for a formalstructure. For example, the source content for the FSM is defined as the FSMitself without any particular presentation. From a purely theoretical perspective,this is impossible, but it can be practically achieved within the context of theSemantic Web [45] by using eXtensible Markup Language (XML) applicationlanguages. So, the content for the FSM is defined in MXL (refer Section 5.3) and thecontent for the FSM presented in Fig. 1(b) is dictated by the ‘‘primitive’’ theme, andencoded in eXtensible 3D (X3D). The personalized representation problem is thus ananalogical mapping that is framed within the Semantic Web, and XMLtransformations.

In a manner similar to the Structure Mapping Engine (SME) approach of Gentneret al. [46,47], the conceptual (i.e., class) mapping between the themes is illustrated inFig. 2. Even though we have implemented only style sheets for a series oftransformations of certain XML documents, we are working presently on a moreautomated solution to the mapping problem. The diagram shows how concepts inthe ontology for an FSM can be mapped to presentation styles based on a diagram, apipe network, and a specialized landscape. This ontological excerpt is greatlysimplified since an actual ontology for an FSM would also include the systemdefinition for the interface (input, output) in addition to start state, the set of allpossible states, inputs, outputs, and so forth. The concept relation ipo means ‘‘ispart of’’ and conn:2 means ‘‘connects’’ with an existential quantity constraint of‘‘connecting two of.’’

ARTICLE IN PRESS

FSM

ipoipo

conn:2

State Trasition

DIAGRAM

ipoipo

conn:2

Circle Curve

PIPE_NETWORK

ipoipo

conn:2

Tank Pipe

LANDSCAPE

ipoipo

conn:2

Gazebo Path

Fig. 2. Analogical concept mapping for a FSM.


The formation of the ontology and a concept map is only part of the method. Thecomplete method is currently defined as follows:

(1)
State the representation problem by specifying either, or both, of (a) the source(b) the target. Example: Represent the 3 state FSM in Section 2 in MathML.
(2)
For source-only problems, explore possible mapping paths from source tocandidate targets. Example: (a) using a lexical similarity measure based onWordnet [48], identify a mapping between ‘‘states’’ and ‘‘tanks.’’ Using this link,explore consistent semantic link candidates for ‘‘transition’’; (b) use an XSLTstyle sheet transformation to consider mapping from MathML to MXL; (c) useLakoff’s conceptual metaphor database [49] to assist in relating the concept of‘‘state’’ to the concept of ‘‘location,’’ resulting ultimately in state as a locationwithin a ‘‘boundary.’’
(3)
Build or reconstruct a comprehensive set of ontologies. Example: build anontology for the FSM from scratch, but link with a pre-existing ontology for pipenetworks through an ontology merging operation.
(4)
Establish a consistent and complete mapping for five potential layers ofabstraction and aggregation: levels: (a) ontologies, (b) schemata, (c) models, (d)lexical components, and (e) presentation. Example: map levels: OWL ontologylanguage, XML schema, MXL, and word phrase levels, and then explorealternate modes of human interaction to visualize the words with X3D modelsfor tanks and pipes.
(5)
Represent the source content as a target using the mapping and presentation.Example: execute the mapping by enabling the representation to the pipe/tankrepresentation in a 3D model viewer.
Step 1 is to identify the representation problem. A typical problem is one where thesource is known and a personalized representation, and presently unknown target, isdesired. Step 2 is a browsing and exploration phase, involving search, mapping, anda variety of tools for lexical similarity. Step 3 takes the result of Step 2 to formontologies and Step 4 establishes clear linkages at five levels: ontology, schema,model, lexical, and presentation.

ARTICLE IN PRESS


Within the semantic web, these correspond roughly to ontology design, XMLgrammar or schema, XML document verified by that schema, and individual XMLelements with qualifying attributes, and possible Scalable Vector Graphics (SVG) orX3D presentations. Step 5 represents a tool that yields the representation as anexecutable procedure. These steps are not currently automated, but instead, are usedas guidelines for the process. In particular, Steps 2–4 are manually intensive, and weare currently investigating ways in which a subset of these steps can be at least semi-automated.

5.3. Implementation

We have built a software system, called RUBE, which facilitates model-building inaesthetic computing and computer simulation. With regard to Section 5.2, RUBEassumes that one already has performed the mappings and has both the source and

targets clearly identified. ‘‘Themes’’ are used to allow different 3D visualizationsbased on a limited set of analogies. Fig. 3 is a visual snapshot of the Blender-based[50] RUBE interface, showing a three-function data flow network modeling anaircraft reconnaissance mission scenario.

Currently, the source formalisms for RUBE are those used in modeling dynamicsystems: finite state automata, data flow networks, Petri nets, ordinary differentialequations, and System Dynamics graphs. The approach is to allow the modeler touse these tools, while specifying what each graphical object means. The meaning isspecified with an XML language called Multimodeling eXchange Language (MXL).

Fig. 3. The RUBE user interface.

ARTICLE IN PRESS


MXL captures the topology of a system without specifying presentation, and istranslated into a lower ‘‘assembly’’ layer Dynamics eXchange Language (DXL)elements, which are then translated into target code, currently Javascript, with Javaand Python planned for future support. While the graphical scene is being authored,the modeler also specifies which code components go with each dynamic modelcomponent. For example, in an FSM, the modeler wants a chunk of code to beexecuted for any given state, so when a graphical object is identified as an FSM

State, there is also a tool for identifying the function that goes along with this. Ifthere is to be any real-time scene graph manipulation as a result of code execution,the modeler must provide this code. Our research group is attempting to ease themodeling burden by deriving a set of primitive functions without forcing the authorsto create code for each component. SodiPodi is fairly easy to use, but Blender has asteeper learning curve. 3D authoring is getting cheaper and faster, but we still cannotsnap our fingers to create, position, and texture objects in space.

Fig. 4 displays a waveguide model for producing sound. It is a ratherstraightforward figure, and we can easily discern connections between the elements,such as the bandpass filters (BP) and delay lines. The diagram serves the purpose ofbeing usable. Now, let us consider another, related figure.

Fig. 5 is a single frame from an interactive 3D representation contains the sametopology as the diagram in Fig. 4. While it was constructed with Maya, and notBlender using the new RUBE interface, the model serves as another sample target forthe sort of geometrical structure that should be possible in the near future. It is astudent project, that meets the above assessment guidelines, while providing for anorganic aesthetic for a network. The modeled network is composed of eight nodes,each of which performs a specific task. The interaction node can be seen at the baseof Fig. 5(d), as it accepts input from below. The digital signal passes up through theinteraction node, which passes its data into a bandpass (BP) filter. Then, from thefilter, there are six pipelines (transport delays), which feed back into the interactionnode. The feedback from the system defines banded waveguides [51], which achievethe simulation of instruments such as bowed bar percussion, glass harmonica, andthe Tibetan prayer bowl. We have not yet passed this model through the RUBE

Fig. 4. Diagrammatic representation of the banded waveguide model.

ARTICLE IN PRESS

Fig. 5. Four views of a physical wave propagation model for real-time sound synthesis (Artist: Joella

Walz): (a) Closeup of a functional node, (b) another closeup, (c) top view of complete network and (d) side

view of network.


system, but this is planned. If one compares the two figures, one first realizes thatwhile Fig. 4 is ideal for presentation on paper media; Fig. 5 requires a more complex,interactive interface. Thus, if paper is the target medium and the task is purely one ofusability, Fig. 4 is clearly superior. Regarding the two figures, though, as differentinterface projections, the first note is that ideally both would be part of the sameinterface and that it would be possible to move seamlessly from one to the otherthrough integration and interaction. However, even provided this latitude, one mightwonder as to the purpose of Fig. 5. Fig. 5 provides an integrative interface with apersonalized representation, ‘‘drawing in’’ the viewer to reflect upon its structure. Inthis sense, the organic aesthetics present within Fig. 5’s interface individualizes andpersonalizes the interface experience. If, as Norman [26] and Jordan [25] point out,that design is as much about subjective quality as usability, we need to pay moreattention to the affective state of the user, and hence the sphere of aesthetics that theuser finds influential. This is useful for teaching novices, but also may pave the wayfor future interface design.

The discipline of semiotics, with its definition of symbol, tells us that the choice ofgeometry when representing a function is conventional and arbitrary. It is no more

ARTICLE IN PRESS


natural to represent a function as a square or rectangle than it is to use the craftedobject within the network shown in Fig. 5. Existing diagrammatic presentation ofmodels such as FSMs appear to be meaningful only because of the abundance oftechnical literature that has employed them. This employment is due, not to theidealized nature of diagrams, but to the inherent cost in creating other forms. Themain issue is one of economy of the modeling process, and as suggested in [19], thedearth of tools needed to achieve customization and cultural choice in a timelymanner. Free tools such as Blender and RUBE should assist modelers to craft theirsoftware and formal structures. If visual computing is to hope for progress along thelines of representation, new forms must be considered based on what notationalmethods are now made affordable by improved tools and technologies. Ideally,integrative interfaces combining multiple 2D and 3D modes, will prove to enrich theuser’s experience.

6. Empirical studies

Integrating visual modes suggested by Fig. 5 may well increase apparent usability

[52,53]. To test this and related hypotheses about the efficacy of aesthetic computing,over the past three years, we performed three studies to try to empirically assess theaesthetic computing technique. Two of the studies were held within the AestheticComputing course, and one of the studies in a Computer Simulation class. Thesestudies are described in some detail in an unpublished manuscript [54], and highlightsare presented here.

The author has taught three classes in Aesthetic Computing at the University ofFlorida, in each of the last Spring semesters. The class is primarily a special topics

class with some lecture, deliverables, student talks, and invited speakers [55]. Thecourse has changed somewhat since the original one in 2000, but the key aspects haveremained constant. There is a combination of instructor lecture, student talks,invited lectures, and student projects. The lecture involves a coverage of language,semiotics, analogy and metaphor, in addition to methods meant to enhancecreativity. The lecture methods mirror the approach described in Section 5.2. Bothartists and computer scientists take the class. Students give talks on papers that havebeen published in areas that highlight these topics. Invited lecturers come fromoutside of Computer Science. The student projects are divided into two types: virtualand physical. Virtual projects are delivered in both 2D and 3D movies or interactivemedia. Physical projects are made from materials found in craft, hardware, and on-campus stores.

The studies were conducted by a third party and all student verbal and writtenresponses remain anonymous to the author. Each study was composed of (1) awritten survey, and (2) focus group questions. In one instance, the focus group wasvideo-taped and transcribed, and then answers could subsequently be coded to teaseout quantiatively meaningful results. A sample Table 1 summarizes the responses toa question about the usefulness of the aesthetic method for other areas in computerscience. The response to this question is very similar to the response to the same

ARTICLE IN PRESS

Table 1

Spring 2002: usefulness of the aesthetic method

In what ways can aesthetic methods be used to illustrate, teach, learn, or communicate other areas in

computer science?

9 Data structures/algorithms/OO-programming

1 OS/networks

1 Database

2 Digital logic/computer architecture

0 Discrete math/numerical analysis

0 Graphic user interface

1 Artificial intelligence

6 Useful in any area

12 Other

5 Better understanding of difficult concepts

0 More interesting

6 Easier to visualize/illustrate

8 Other

2 Do not know/no response

2 Negative response


question by the Fall 2001 class on modeling and simulation. About two-thirds of thestudents listed a specific topic in which it would be useful, and about half gavegeneral responses. Here is a sample response that emphasizes the communicationaspects of the aesthetic method:

The aesthetic methods strength lies in the ability to communicate concepts toparties that have disparate backgrounds or no prior knowledge of the conceptbeing explained. This means that aesthetic methods would be excellent forinstructing novices or communicating concepts across disciplines or with foreignstudents.

Overall, and apart from the general interdisciplinary goals, our studies on studentsin the Simulation and Aesthetic Computing classes demonstrated the followingresults:

�
Students find the ability to customize and personalize a positive attribute. This isconsistent with the constructivist approach to learning [43] and the ‘‘multipleintelligences’’ argument of Gardner [56].
�
Students enjoyed the technique, providing incentive for employing it. This can beseen in a high preference for the methodology. Learning and preference appear tobe highly correlated: those who enjoy a task are more likely to learn the conceptsthat form the basis for the task. Preference, as indicated by affective qualities ofmotivation and attitude [57], seems to correlate well with the ability to be creative,and creativity has a strong connection with aesthetics [58,59].

ARTICLE IN PRESS


�
Students found that the technique improved their communication of technicalconcepts to others. This can be especially important in teams with diverse sets ofbackgrounds and interests, such as in the gaming and cinematic effects industries.And, increasingly, simulation packages are taking on more 3D, ‘‘game like’’characteristics such as three dimensions, lighting conditions, textures, digitalactors, and the ability to be engaged within the modeling environment (referpackages such as Arena, Automod, and Taylor discussed in [60]).
7. Conclusions

Even though we are constructing an initiative and environment conducive topluralistic representation of formal mathematical and computing structures, it isnatural to question whether anyone will use such a facility even if it is made efficient.Apart from our own inherent optimism, which drives our research, we see a positiveset of trends that suggest that as a technologically based society, we are moving in adirection consistent with the goals of aesthetic computing:

�
Notation for mathematics and computing, which encompasses how we viewprogram and data structures, is historically constrained by technology andeconomy. We move in a direction that is economical, avoiding complex, expensiveapproaches. The areas of visualization within computing, including those ofinformation, software, and science, are still new, and made possible through earlywork in the 1960s when computer graphics got its jump start. The formation, inthe late 1970s and 1980s, of the graphical user interface was a major step in, noteliminating text, but augmenting it to achieve more humane connections to code.It is not too surprising to imagine that this movement will affect structures notonly in computing, but also in mathematics.
�
Ever since the early Greek confluence of art and science, in the spirit of techne, thetwo areas have split and separated. Efforts in mass production benefited society atlarge, but caused another wedge to increase the size of the split; however, it is ironicthat mass customization builds directly on mass production, and so is making iteconomically feasible to personalize and customize objects. The skinz movement isone example of this in computing; however, at least within the area of dynamicsystems, there is a half-century of developing new model structures, of permittingcultural groups to flourish within technical societies. It is possible to have standardsand pluralism at the same time, permitting multiple metaphors to coexist.
An impediment to progress in representation is the inertia associated with ourpresent biases and cultures. Mathematicians and computer scientists have beentrained to think non-visually, and state of the art programming languages are stilltextual. UML [61] seems to be thriving, but it is used mainly as a requirementsspecification tool, not a language for creating code from diagrammatic forms. Untilwe get to that stage with UML or other contenders, visual computing paradigms willremain in their infancy. We have started one thread of research around using RUBE

ARTICLE IN PRESS


for programming, and hope to have results within the year. If the field of computingcan be thought of as being slow to adopt visual programming languages, the field ofmathematics proceeds at snail’s pace. We have relied, successfully, on textualnotation for many centuries but it is time to explore visual approaches, catalyzed bythe work done in the visual computing community.

The issue of form vs. function is ever present in visual computing as it is in art.Form refers to the structure and the material aesthetics of a set of objects, whereasfunction refers to pure utility. Is the goal of aesthetic computing to allow morevisually expressive forms, or to leverage artistic practice to yield ‘‘natural’’ formsthat seem to automatically suggest specific functions? Both goals are important sinceit is not really possible to separate form from function: one of the functions ofmodeling is to engage the user, by stressing design preference even over minorfunctional inconveniences, and forms are meaningless if they cannot be directlyassociated with functions and their symbolic references learned in a reasonableamount of time. Frank Lloyd Wright [62] echoes this logic in his quote ‘‘Formfollows function—that has been misunderstood. Form and function should be one,joined in a spiritual union.’’ Tractinsky [63] suggests that aesthetics can haveempirically definable positive results on usability and functionality. Increasing thediversity of cultural forms provides more choice to users, while seeking forms thathave natural functions is problematic since symbolic reference is conventional, andthus, not automatic. There is nothing natural about an arrow and two circles: one isrequired to learn these symbols and their meanings. However, the arrow whererecognized as such, can suggest motion and direction. But, the same can be said ofpipes and paths with moving agents. These objects require cultural associations towork effectively as symbols. Ultimately, the ability to manifest new domains forabstract concepts, while economically leveraging culturally specific artifacts,provides us with greater freedom and more choice. A focus on aesthetics assists usin reaching this goal.

Acknowledgements

The author would like to thank the National Science Foundation under GrantEIA-0119532 and the Air Force Research Laboratory under Grant F30602-01-1-05920119532, with special thanks to recent and current students working RUBE:Minho Park, Jinho Lee, Hyunju Shim, Nam Kyu Lim, Kristian Damkjer, JohnHays, and Joella Walz for her independent study. The author would also like tothank the student population of the Aesthetic Computing classes for their valuablefeedback and projects.

Appendix: Dagstuhl Manifesto

The following was drafted in July 2002, and is provided courtesy of Leonardo, 35(4),MIT Press, Cambridge, MA, 2003, p. 255. The application of computing to

ARTICLE IN PRESS


aesthetics, and the formation of art and design, has a long history, which resulted inthe emergence of computer art as a new art form in the 1960s, with the integration ofhardware, software, and cybernetics. We propose to look at the complementary areaof applying aesthetics to computing. Computing, and its mathematical foundations,have their own pre-existing aesthetics; however, there is currently a differencebetween the relative lack of variety of these aesthetics in contrast to art, which has along history containing of a multitude of historical genres and movements. We wishto strike a balance between cognitive and material aesthetics. Software as written intext or drawn with flow-charting may be considered elegant. But that is not to saythat the software could not be rephrased or represented given more advanced mediatechnologies that are available to us today, as compared with when printing was firstdeveloped. Such representation need not compromise the goals of abstraction, whichis a necessary but not sufficient condition for mathematics and computing, asmeaning, comprehension, and motivation may be enhanced if the presentation isguided by a pluralism of aesthetic choices and multiple sensory modalities.

Computer programs and mathematical structures have been traditionallypresented in traditional text-based notation even though, recently, substantialprogress has been made in areas such as software and information visualization toenable formal structures to be comprehended and experienced by larger and morediverse populations. And yet, even in these visualization approaches, there is atendency toward the mass-media approach of standardized design, rather than anapproach toward a more cultural, personal, and customized set of aesthetics. Thebenefits of these latter qualities are: (1) an emphasis on creativity and innovativeexploration of media for software and mathematical structures, (2) leveragingpersonalization and customization of computing structures at the group andindividual levels, and (3) enlarging the set of people who can use and understandcomputing. The computing professional gains flexibility in aesthetics, and associatedpsychological attributes such as improved mnemonics, comprehension, andmotivation. The artist gains the benefits associated with thinking of software, andunderlying mathematical structures, as subject material for making art. With thesebenefits in mind, we have created a new term Aesthetic Computing, which we defineas the application of art theory and practice to computing.

References

[1] H. Bredin, L. Santoro-Brienza, Philosophies of Art and Beauty: Introducing Aesthetics, Edinburgh

University Press, Edinburgh, 2000.

[2] P.A. Fishwick (Ed.), Aesthetic Computing, MIT Press, Cambridge, MA, 2005, in press.

[3] P.A. Fishwick, Exploring multiple visualization perspectives with aesthetic computing, in: A.F.

Cardenas, P. Mussio (Eds.), 2003 International Conference on Visual Languages and Computing

(VLC 2003), 2003, pp. 341–346.

[4] J. Hopkins, P.A. Fishwick, Exploiting an agent-based metaphor in software visualization using the

rube system, Journal of Visual Languages and Computing 14 (2003) 97–117.

[5] A. Bloom, The Republic of Plato, second ed., Basic Books, New York, 1968.

[6] I. Kant, The Critique of Judgement, Clarendon Press, Oxford, 1790, transl. by James Creed

Meredith, 1952.

ARTICLE IN PRESS


[7] W.A. Shibles, Emotion in Aesthetics, Kluwer Academic Publishers, Dordrecht, 1995.

[8] L. Tolstoy, R.G. Collingwood, The expression theory: Art and feeling (Section 1), in: J.M. Thompson

(Ed.), 20th Century Theories of Art, Carleton University Press, 1990, pp. 1–56.

[9] J. Hadamard, The Psychology of Invention in the Mathematical Field, Princeton University Press,

Princeton, NJ, 1945.

[10] M. Emmer (Ed.), The Visual Mind: Art and Mathematics, MIT Press, Cambridge, MA, 1993.

[11] D.E. Knuth, Literate Programming, CSLI Lecture Notes, vol. 17, 1992.

[12] D.E. Knuth, The METAFONT Book, Addison-Wesley, Reading, MA, 1986.

[13] M. McLuhan, Understanding Media: the Extensions of Man, MIT Press, Cambridge, MA, 1994.

[14] M.H. Brown, Algorithm Animation, MIT Press, Cambridge, MA, 1987.

[15] N. Shu, Visual Programming, Van Nostrand Reinhold, New York, 1988.

[16] S.-K. Chang (Ed.), Visual Languages and Visual Programming, Plenum Press, New York, 1990.

[17] J. Stasko, J. Domingue, M.H. Brown, B.A. Price (Eds.), Software Visualization: Programming as a

Multimedia Experience, MIT Press, Cambridge, MA, 1998.

[18] S. Diehl (Ed.), Software Visualization, Springer, Berlin, 2001, lNCS 2269.

[19] R. Esser, J.W. Janneck, A Framework for Defining Domain-Specific Visual Languages, in:

Workshop on Domain Specific Visual Languages, ACM Conference on Object-Oriented Program-

ming, Systems, Languages and Applications (OOPSLA-2001), 2001.

[20] S.K. Card, J. Mackinlay, B. Shneiderman (Eds.), Readings in Information Visualization: Using

Vision to Think, Morgan Kaufmann, Los Altos, CA, 1999.

[21] D.N. Snowdon, E.F. Churchill, E. Frecon (Eds.), Inhabited Information Spaces: Living with your

Data, Springer, Berlin, 2004.

[22] T. Kim, J. Lee, P.A. Fishwick, A two-stage modeling and simulation process for web-based modeling

and simulation, ACM Transactions on Modeling and Simulation 12 (3) (2002) 230–248.

[23] M. Kelly (Ed.), Encyclopedia of Aesthetics (4 vols.), Oxford University Press, Oxford, 1998.

[24] A.C. Sukla (Ed.), Art and Representation: Contributions to Contemporary Aesthetics, Praeger

Publishers, New York, 2000.

[25] P.W. Jordan, Designing Pleasurable Products: An Introduction to the New Human Factors,

Routledge Press, London, 2002.

[26] D.A. Norman, Emotional Design: Why We Love (or Hate) Everyday Things, Basic Books, New

York, 2004.

[27] S.-J. Shin, The Iconic Logic of Peirce’s Graphs, MIT Press, Cambridge, MA, 2002.

[28] H. Lieberman, A three-dimensional representation for program execution, in: E.P. Glinert (Ed.),

Visual Programming Environments: Applications and Issues, IEEE Press, New York, 1991.

[29] M. Najork, Programming in three dimensions, Journal of Visual Languages and Computing 7 (2)

(1996) 219–242.

[30] M. Hesse, Models and Analogy in Science, University of Notre Dame Press, Notre Dame, IN, 1966.

[31] G. Lakoff, M. Johnson, Metaphors We Live By, University of Chicago Press, Chicago, 1980.

[32] G. Lakoff, Women, Fire and Dangerous Things: what categories reveal about the mind, University of

Chicago Press, Chicago, 1987.

[33] P.A. Fishwick, Simulation Model Design and Execution: Building Digital Worlds, Prentice-Hall,

Englewood Cliffs, NJ, 1995.

[34] J.F. Hopkins, P.A. Fishwick, Synthetic human agents for modeling and simulation, Proceedings of

the IEEE 89 (2001) 131–147.

[35] Digital arts and sciences (2003). URL http://www.cise.ufl.edu/dept/das/das.html

[36] P.A. Fishwick, Aesthetic programming: crafting personalized software, Leonardo 35 (2002) 383–390.

[37] P.A. Fishwick, Aesthetic computing: making artistic mathematics and software, YLEM Journal:

Artists Using Science & Technology 10 (2002) 6–11, Special Issue on Art and Programming.

[38] R. Malina, The stone age of the digital arts, Leonardo 35 (5) (2002) 463–465.

[39] Aesthetic computing seminar (2002). URL http://www.dagstuhl.de/02291/

[40] D. Schmandt-Besserat, How Writing Came About, University of Texas Press, Austin, TX,

1997.

[41] W. Bohn, The Rise of Surrealism, State University of New York Press, New York, 2002.

http://www.cise.ufl.edu/dept/das/das.html

http://www.dagstuhl.de/02291/

ARTICLE IN PRESS


[42] R. Arnheim, Toward a Psychology of Art: Collected Essays, University of California Press, Berkeley,

CA, 1966.

[43] S. Papert, Mindstorms: Children, Computers, and Powerful Ideas, second ed., Basic Books,

New York, 1999.

[44] B.J. Pine, Mass Customization: The New Frontier in Business Competition, Harvard Business School

Press, 1993.

[45] T. Berners-Lee, J. Hendler, O. Lassila, The Semantic Web, Scientific American, May 2001.

[46] D. Gentner, A. Stevens, The Mental Models, Lawrence Erlbaum Associates, Hillsdale, NJ, 1983.

[47] D. Gentner, D.R. Gentner, Flowing waters or teeming crowds: mental models of electricity, in: D.

Gentner, A.L. Stevens (Eds.), Mental Models, Lawrence Erlbaum Associates, Hillsdale, NJ, 1983,

pp. 99–129.

[48] C. Fellbaum (Ed.), Wordnet: An Electronic Lexical Database, MIT Press, Cambridge, MA, 1998.

[49] Conceptual metaphor collection. URL http://cogsci.berkeley.edu/lakoff/

[50] Blender, 2003. URL http://www.blender.org

[51] G. Essl, Physical wave propagation modeling for real-time synthesis of natural sounds, Ph.D. thesis,

Princeton University, November 2002.

[52] M. Kurosu, K. Kashimura, Apparent usability vs. inherent usability: experimental analysis on the

determinants of the apparent usability, in: 1995 Proceedings of the Conference on Human Factors in

Computing Systems, Association of Computing Machinery (ACM), New York, 1995, pp. 292–293.

[53] N. Tractinsky, Aesthetics and Apparent Usability: Empirically Assessing Cultural and Methodolo-

gical Issues, in: 1997 Proceedings of the Conference on Human Factors in Computing Systems,

Association of Computing Machinery (ACM), New York, 1997, pp. 115–122.

[54] P.A. Fishwick, T. Davis, J. Douglas, Using aesthetic computing as a method for customizing model

structure: an empirical study, ACM Transactions on Modeling and Simulation, submitted for

publication.

[55] Aesthetic computing, 2003. URL http://www.cise.ufl.edu/fishwick/aescomputing

[56] H. Gardner, Multiple Intelligences: The Theory in Practice, Basic Books, New York, 1993.

[57] R.S. Nickerson, Enhancing Creativity, in: R.J. Sternberg (Ed.), Handbook of Creativity, Cambridge

University Press, Cambridge, 1999, pp. 392–430.

[58] H.E. Gardner, Art, Mind, and Brain: A Cognitive Approach to Creativity, Basic Books, New York,

1984.

[59] L. Candy, E.A. Edmonds, Explorations in Art and Technology, Springer, Berlin, 2002.

[60] A.M. Law, W.D. Kelton, Simulation Modeling and Analysis, third ed., McGraw-Hill, New York,

2000.

[61] J. Rumbaugh, I. Jacobson, G. Booch, The Unified Modeling Language Reference Manual, Addison-

Wesley, Reading, MA, 1999.

[62] Frank Lloyd Wright. URL http://www.cmgww.com/historic/flw/quotes.html

[63] N. Tractinsky, A. Shoval-Katz, D. Ikar, What is beautiful is usable, Interacting with Computers 13

(2000) 127–145.

http://cogsci.berkeley.edu/lakoff/

http://www.blender.org

http://www.cise.ufl.edu/fishwick/aescomputing

http://www.cmgww.com/historic/flw/quotes.html

Enhancing experiential and subjective qualities of discrete structure representations with aesthetic...

Documents

Transcript of Enhancing experiential and subjective qualities of discrete structure representations with aesthetic...