About | Project Members | Research Assistants | Contact | Posting FAQ | Credits

Ben Fry, Valence (2001)

Research Report by Brooke Belisle
(created 05/21/07; version 1.0)
[Status: Draft]

Related Categories:

Original Object for Study description

Valence is a software program written by Ben Fry to dynamically render complex information as a visual, three-dimensional, and relational representation. It has been produced and installed in multiple versions that take various inputs. The original version, which ‘reads’ novels, was installed in 2001 at Ars Electronica in Austria and appeared in the film Minority Report. The latest version, which visualizes genetic information, was installed in 2002 at the Whitney Biennial in New York and appeared in the film Hulk. Images and quick time movies of various instantiations of Valence can be viewed at Fry’s website.

Benjamin Fry describes Valence as “a set of software sketches about building representations that explore the structures and relationships inside very large sets of information.” [1] The first version of Valence explored the relationships between all the words that constitute Mark Twain’s novel The Innocents Abroad. Each word of the novel is input into the Valence software, presumably through a program that pulls each word, one by one, from a digital file of the book’s text. Valence then visually renders each word, and the developing relationships between words, in a virtual, three-dimensional space. Fry describes this as an act of reading, performed by the software:

The program reads the book in a linear fashion, dynamically adding each word into three-dimensional space. The more frequently particular words are found, they make their way towards the outside (so that they can be more easily seen), subsequently pushing less commonly used words to the center. Each time two words are found adjacent in the text, they experience a force of attraction that moves them closer together in the visual model. [1]

Valence generates a visual model of the text in which words orbit and cluster in a roughly spherical organization, like an astronomical diagram. The words move and change size, and new lines of connection sketch different constellations as the source text is continually processed. In 2001 a version of Valence using German books as input was installed at Ars Electronica in Austria. [2] Users were able to interact with the projection using a custom interface; they could “zoom into the center of the space to look around, or rotate the piece to view other locations.”[3, p68] With this interaction, the user performs an act of ‘reading’ distinct from the program’s initial processing and rendering of the text.

While Valence looks like information, and may be said to ‘read’ in the way that it processes and renders its source data, the information Valence represents is not necessarily legible to the user who would read it. Valence reads only in the most visual sense, exploring “structures” and “relationships” found at the visual level of the text. It tallies the number of times a word appears rather than evaluating a word’s syntactical weight or semantic importance; it registers the visual proximity of words in prose rather than interpreting their syntactical or semantic relationships. Valence’s rendering of Innocents Abroad seems to offer a statistical rather than a literary interpretation; if it offers anything of literary interest it may be in its representation of American English in 1869, or of Mark Twain’s prodigious and unusual vocabulary. A reader may glean very little about the novel by interacting with it as the interface Valence offers. In the film Minority Report, Valence has a brief cameo as the futuristic interface on the lead character’s home computer. Without offering any detail, the film uses Valence as an image of interface, suggesting, I would argue, a way in which Valence coincides with a somewhat cinematic imagination of information.

Fry admits that Valence is limited as a reading interface in the traditional sense of reading text. He was interested in “useful” results and frustrated that one could not really do anything with the version of Innocents Abroad Valence produced. Finding that books provided an “imperfect” example he went on to develop alternate versions or inputs for the program. Fry describes a particularly successful iteration of Valence that explored user traffic on websites, tracking the number of visits to any given URL and the paths users followed between sites. He claims that this modeling of actual use suggested an optimal architecture which could then be compared to the actual design of the website and used to improve it. At the other extreme, he describes a failed experiment in which he set up Valence to render live keyboard input from users. He found that the random input, often strings of letters rather than words, was too erratic to provide any coherent or compelling visual shape. At some point in his research, Fry decided to set aside text altogether:

At this point I was uncomfortable with the use of text, because it seemed to fall into a common trap that exists when using text. It is easy for users to quickly latch onto a line of text, people seem to find text familiar (in a potentially unfamiliar, abstract space) and stimulating (when executed in 3D). It was easy, then, for observers to disregard the remaining visuals and be content with just text, because it was more concrete than anything else in the composition. Because I wanted to create forms that were evocative of the data, without simply relying on text to narrate the data, I temporarily discontinued the use of text in an attempt to return to the consideration of visual form. [3, p73]

Fry seems to have been frustrated by user’s attempts to “read” Valence as a text, attending to the words that it reiterated rather than to the visual and spatial relationships it rendered. He distinguishes his effort to “create forms that were evocative of the data” from the way in which text seemed to “narrate the data,” suggesting that because he wanted Valence to ‘show’ rather than ‘tell’ he was forced to abandon text. While Valence ‘red’ text for its visual or spatial context rather than its semantic value, and rendered words as images more than signs, users seemed to attribute semantic value to the words they encountered in the image, constructing the kinds of syntactic relationships that the program itself had ignored. This effort to ‘read’ Valence could have been valued as its contribution to or experimentation with textual form, but, instead, this textual appeal was viewed as a “trap” or a crutch, something not only less challenging but less interesting than the purely visual form and the meanings it offered. People would not learn to read Valence, it seemed, until they were prevented from reading it through the more familiar model of textual reading.

For the 2002 Whitney Biennial Fry created the next, and currently last version of Valence, Genome Valence. [4] Genome Valence uses a revised visual display, with colored ribbons connecting information in more complex paths than the geometrical arcs of the first version. Mapping genetic information seems particularly suited to the program, which was initially developed according to metaphors patterned after organic life.

Research Context:
Fry developed the first version of Valence in 1999, as part of his master’s thesis project at MIT. A student in John Maeda’s Aesthetics + Computation group, Fry produced a thesis titled Organic Information Design, downloadable from his website. [2] In this thesis, he contextualized Valence within a history of visualizing dynamic systems that includes StarLogo and the Game of Life. He explains how the algorithyms Valence uses are modeled after life processes such as metabolism and homeostasis; the system enacts rules based on organic behavior. He images data as “nutrients fed to the system” and inputs as relating according to an organic logic of nodes and branches. [2, pp48,66] Fry developed a second, less known version of the software which he called Anemone. Anemone visualizes website usage as the growth, branching, and development of an organic structure like a tree. Unused areas “wither” while heavily used areas thicken and “blossom.” Unlike Valence, Anemone is a two-dimensional interface. It way seem less impressive, but Fry actually argues that it may be more legible, since when projected on a three-dimensional screen, the depth of a three-dimensional representation is actually more confusing than it is informative. While Valence has received more attention and acclaim, Anemone offers an equally, if not more, compelling achievement of Fry’s goal: to represent complex and dynamic information as visual form.

Fry does not seem to have plans for creating new reading interfaces in the future. In his thesis, he speculated that the visualizations Valence and Anemone offer would be most promising for research areas such as genomics, economics, audio processing, game theory, and software architectures. [2, p90] It is likely that his contributions will be taken up by others in those fields, and others.

Technical Analysis:
The full version of Valence was created with C++, Perl, and OpenGL. It can be viewed on a computer monitor and interacted with through a standard PC interface, but it is usually installed as a digitally projected image, which users can interact with using a custom input device. An online applet representing the Innocents Abroad version of Valence was created using Processing, an Open Source programming language created by Ben Fry and Casey Raes. [5]

Evaluation of Opportunities/Limitations for the Transliteracies Topic:
Valence is perhaps less interesting for the way it ‘reads’ literature than for its ability to render the spatial and visual relations of data sets in which these kinds of relations are inherently more meaningful. For instance, it is likely that the path a user takes between two pages of a website will yield meaningful information about the website and the ways websites are used; but it seems less likely that the proximity of words on a page will reveal meaningful insights about a novel or the way we read novels. Fry’s work to visually represent complex data emphasizes the limitations of textual approaches to ‘reading’ and reaches toward alternative, more graphical forms that would redefine what reading can be.

Resources for Further Study:

  Kimberly Knight2, 05.21.07

Comments are closed.