Games of Inquiry for Collaborative Concept Structuring

Google’s project to digitize five of the world's greatest libraries will dramatically extend their search engine reach in the future. Current search-engine philosophy, which asserts that "any search starts with a question to be answered," will need to be advanced in terms of Peirce's philosophy: "Any inquiry begins by creating an hypothesis to be tested or with abduction." As conceptual structures researchers prepare to meet access challenges in the world of large Internet knowledge stores, they have a solid foundation in Peirce's theorized stages of inquiry: abduction, deduction, and induction. To indicate how conceptual structures tools must augment collaborative, Internet-based inquiry, we imagine a future scenario in the context of a user-centered testbed, where Peirce scholars apply Peirce's pragmatic theory in their complex manuscript reconstruction work. We suggest that games of inquiry can be developed to formalize user collaboration and technology needs, for improved specification of tool requirements in the testbed context. 1 Scenario Context and Characters Imagine some particular day in the future, when the technology to support collaborative inquiry has advanced through rapid-prototyping in user-centered testbeds, so that researchers can work together efficiently, no matter where they are located and no matter where their primary data is safely stored in library archives. In this particular scenario, the researchers are Peirce scholars and the data of their research are Peirce's manuscripts archived at Harvard's Houghton Library. These scholars are remotely located from one another, so must use the Internet to collaborate in their work to transcribe the digitized images of Peirce's handwritten manuscripts into digital text and to catalog a coherent sequence of Peirce's manuscript pages. Transcription involves deciphering what Peirce wrote, which becomes more difficult as his handwriting deteriorated in his declining years when he suffered from cancer and other ailments. During the last seventeen years of his life (1897-1914), Peirce produced his most intensive theoretical work, on semiotic, pragmatism, and his system of Existential Graphs (EG) as a notation for the study of logic [see 1 and 2]. Amounting to some 40,000 pages, very few of these manuscripts have ever been published. And very few scholars have ever studied these pages, which their Houghton Library curator estimates will survive no more than a few dozen more years (due to their high-acid paper). By 2005, scholars and technologists realized that Internet-based technology had advanced to make possible the reconstruction and representation of Peirce's polymathic corpus (ranging from religion, philosophy, and cosmology, to computer science, mathematics, logic, geophysics, and many others fields). Five, especially qualified scholars joined to pursue some of that reconstruction, collaboratively, in the Peirce Online Resource Testbed (PORT) project [3]. Jay Zeman, Professor Emeritus of Philosophy and specialist in Peirce's EG, lives in Florida; Bob Burch, Professor of Philosophy at Texas A&M is also an expert in Peirce's EG; Peter Øhrstrøm, Professor of Communication at the University of Aalborg in Denmark is an specialist in modal logic (including Peirce's "gamma graphs"); Robert Marty is Professor Emeritus at the University of Perpignon in France, a specialist in Peirce's "semiotic logic"; and Frithjof Dau, in Germany, completed his doctoral degree studying concept graphs in 2002, and a treatise on Peirce's EG in 2005. They joined PORT's collaboratory testbed, in partnership with technologists, to help develop imaging and transcribing methods and tools that would improve the quality and efficiency of their work. They are particularly motivated to create better access to the manuscript data because of inadequacies and misrepresentations in previous editions of Peirce's writings. The Collected Papers of Charles Sanders Peirce (referred to as "CP"), an eight-volume edition produced more than fifty years ago1, remains (in spite of its many deficiencies) the most complete print resource of his work2. On this particular imaginary day in the future, the scholars are engaged in perhaps the most difficult detective problem during the initial stage of 1 The misnamed Collected Papers of Charles Sanders Peirce is the only large-scale and purportedly comprehensive edition of Peirce's writings. Published in six volumes, from 1931 to 1935, and enlarged by two volumes in 1958, its topical selections omit Peirce's writings on science and mathematics almost entirely. Although, it contains nearly 150 selections from his unpublished manuscripts, only one-fifth of those selections consist of complete manuscripts, and many of them are inaccurately dated; parts of some manuscripts appear in up to three of the eight volumes, and at least one series of papers is scattered throughout seven of the eight. Most manuscripts appear in excerpts (with only rare indication of how much has been left out), and different parts of the same manuscript appear in different volumes (because of the topical order), and in some cases parts of different manuscripts are grafted together (without mention), sometimes consisting of writings composed more than three decades apart. As a scholarly tool, this edition is unreliable and often frustrating [See 4 for detailed discussion]. 2 A textually and chronologically reliable print edition of Peirce's writings was begun in the late 1980s (Writings of Charles S. Peirce: A Chronological Edition) [5]. As of 2005, only six of the originally projected thirty volumes have been published, and even if completed, that edition would present less than one-third of Peirce's entire work. More significantly, in its paper-print format, cost of production may prohibit representation of Peirce's late manuscripts, which are characterized by text enclosed in graphical figures, graphics embedded in text, text contoured around graphics, whole pages of graphics with no text at all, and graphical figures with as many as four colors. reconstructing Peirce's corpus. As in the initial stage of scientific investigations, the scholars must prepare the raw data (in the form of digitized manuscript page images) to serve as reliable data for interpretation, by calibrating (ordering the pages in some sequence and grouping the sequences accurately). For scholarly interpretation of his ideas as they evolved, the page order should be chronological, as they were written by Peirce. Unfortunately, achieving that ideal may never be entirely possible, since only about one third of the corpus was originally dated by Peirce. At least, the order of each manuscript should be as close to Peirce's own apparent compositional order as possible. Even that goal is challenging, since some 10,000 pages in the Harvard collection3 are considered "lost." These pages are present in the collection, but not properly placed in any manuscript sequence. Many more pages have been tentatively placed in supplementary folders, and labeled, for example, "145(s)," to indicate pages thought to belong to MS 145. In other cases, if scholars have not yet considered some group of "lost pages," which seem to belong together, they are simply labeled with a manuscript number. Such is the case with manuscript "507," a folder of nine pages, listed in the "Robin catalogue" of the Houghton collection4 as undated and unnumbered. 2 A Manuscript Mystery We enter the scenario with the five scholars attempting to solve a "lost page" mystery, as an example of the complexity of work in the initial stage of reconstruction. The 32 pages of the folder labeled "145(s)" have been previously identified as possibly belonging with the folder labeled "145," which the Robin catalogue lists as a manuscript of 12 pages (undated and unnumbered). Quoting from the manuscript title page, the catalogue entry reads: "An Attempt to state systematically the Doctrine of the Census in Geometrical Topics or Topical Geometry, more commonly called 'Topologie' in German books; Being a Mathematical-Logical Recreation of C.S. Peirce following the lead of J.B. Listing's paper in the 'Gottinger Abhandlungen'" [6:17]. The scholars each verify this catalogue entry with what they find in the "145" folder (using Hoffmann's online version of the Robin catalogue and PORT's online manuscript data archive). They find that the association of "145" pages with the manuscript pages of "145(s)" is 3 When Peirce's papers (amounting to between eighty and one-hundred thousand pages) were given to Harvard, shortly after his death (1915), they were not properly stored for several years (apparently, even distributed as scratch paper in war-time paper shortages), during which they fell into disarray. The microfilm (high-contrast, black and white, produced in 1967) and photo copies from it cannot be used in place of the original pages, which must be carefully examined for any discriminating features that might be clues as to where they belong (such as color and width of ruled lines and handwritten script, type of paper, watermark, shape of a torn or cut edges, weathering effects, and so on). 4 Richard Robin's Annotated Catalogue of the Papers of Charles S. Peirce [6] provides the most valuable and comprehensive view of the disordered manuscripts (and their incomplete representation in the microfilm edition). See Michael Hoffmann's digital version of the Robin catalogue: http://www.iupui.edu/~peirce/robin/robin.htm. supported by empirical evidence they can see in the manuscript page images displayed on each of their screens simultaneously. The pages of "145" are in a notebook with lined pages that appear to match the pages of "145(s)." But they question whether the folder labeled "145(s)" (of 32 pages) should be designated "145," while the folder labeled "145" (of 12 pages) should be "145(s)," indicating that its fewer pages are supplementary material for what should be the primary material in the folder currently marked "145(s)"? Meanwhile, at least one page in the "507" folder (of unplaced pages) has been recognized as very similar in appearance and content to page number 22 in "145" (so close as to be considered possibly an earlier or later-written version of page 22). However, even though the "507" page has obviously been cut from a notebook, its cut edge does not match any cut fragment in the "145" or "145(s)" notebooks. And though its textual content appears to be nearly identical to that on page 22, the EGs across the top of the page are not identical. The "507" catalogue entry gives almost no clue: "Beta and gamma graphs, with algebraic translations. Rules of transformation" [6: 65]. Previously, the scholars had encountered the comparable case of a notebook catalogued and labeled as "464" and a folder of 15 pages labeled "464(s)" on notebook paper that looks exactly like that of the 35 pages of "464." In this case too, "464," but not "464(s)," has an entry in the Robin catalogue designating it as "Part 1 of the 3rd draft of 3rd Lecture" of his 1903 "Lowell Lecture" series. As with "145," the content of one page of "464(s)" has obviously been cut from a notebook, and it appears very similar to that of a page in the "464" notebook; but in this case the cut edge of the "464(s)" page matches exactly with a cut fragment in "464," and even the marginal text that has been cut through matches. So there is strong empirical evidence that "464" and "464(s)" belong together, while the proper designation of "145" and "145(s)," together with the mystery of the "507" page (which is extremely similar in content but does not empirically match page 22 in "145"), remains in doubt. Where does this page belong, and how should "145" and "145(s)" be properly designated in the catalogue of manuscript material? 3 Essential Collaboration Capabilities For the critical identification work in our future scenario of scholars at work, the digital images of the manuscript pages must be extremely accurate representations of the original pages. They must provide evidence not only of what Peirce rendered in graphical forms and in handwritten text but also of other bibliographic features (such as color, kind, and condition of the paper). The manuscript images must exhibit to all collaborating scholars the same critical features, no matter what may be the variations among the qualities of their computer display systems. Any visible feature on a page may be needed in their judgments as to where the page belongs, what was its significance to Peirce, and what insights it might convey. The mediating tools of their virtual collaboration context make it possible for the scholars to confer effectively in such minute judgments, in spite of the differences in display capabilities among their computer systems. 3.1 Consistent Image Quality An image-display mediator (originally developed as "AI-Trader," by Puder and Römer [7]), monitors the differences among the scholars' five different computer displays and matches them across all system factors (such as file-compression ratio, screen size and resolution, and color calibration). This automatically standardized image is the reference quality for all collaborators, and enables the mediator to notify the scholars when their views do not correspond. On request, the mediator negotiates among the collaborator's individual requests to establish the optimum perceived attribute values for any selected purpose. Matching views is especially important for collaboration on detailed features such as tone of paper (for page matching). Subtle qualities that are not immediately recognized as significant might be identified as crucial evidence by one scholar and then confirmed in collaborative interpretation. All scholars have access to reliably comparable views of the same artifact, while individually each is able to explore the primary data for potentially relevant features identified from the perspective of his particular background knowledge of Peirce's work. The mediation service avoids the problem of simply serving the lowest display system capabilities, and allows the scholars to intervene in the process at any point needed to resolve discrepancies because of differences in image quality among their displayed views of manuscript pages. The lowest quality display system of any scholar does not prevent those with higher quality systems from seeking and finding crucial evidence, and bringing it to the collaboration context as the object for consideration and discussion. The image-display mediator negotiates among complexly interrelated system variance factors (including resolution and color accuracy, compressed file quality, and speed of transmission) to serve all participants efficiently. Concurrently, the scholars can readily monitor this mediation process, as it reconciles "competing factors," to remain aware in viewing the images as the evidence of what Peirce rendered on paper. Any relative bibliographic invariance they identify in the evidence (such as a particular type of notebook paper or color of ink used in particular graphic forms) might then serve to establish meaningful segmentation, to be indexed for catalogue identifications that can be used in automatically searching for possible matching pages. 3.2 Concurrent Catalogue Documentation PORT's concept-based catalogue for the Peirce archive, which has evolved from the data of print finding aids, has been digitally linked to both transcribed text and digital images of the original manuscript pages. By means of the catalogue, each manuscript page's digital image and transcription is coded to all existent corresponding print edition entries of its content, and each image appears on screen with its actual or estimated date of composition, if known. The scholars can propose corrections to the catalogue entries reflecting any additions and alterations accumulated during their intensive transcription and calibration work. Variations among their transcriptions and calibrations are monitored automatically, by digital matching, and these points of disagreement are highlighted in the digital transcriptions. An ontology of the collection (or a comprehensive framework capable of relating the digitized archive data in a conceptual structure, for logical representation of inference relations) interrelates the catalogue entries on many conceptual levels. Special ontologies function as specialized database indexes (or views), representing the conceptual perspectives of particular disciplinary conventions with respect to the archive and any related data. These specialized indexes continue to evolve as each discipline evolves, keeping all views conceptually coordinated. Formal Concept Analysis [see 8] provides a graphical interface of lattices displaying the related views of catalogue entries in a relational database [see 9]. Graphical views (or visualizations) of this data-under-evolution enable disciplinary specialists to remain critically aware of the complex implications of each researcher's contribution (including their own) with respect to others in the scheme of inquiry as a whole. Other methods of knowledge representation and processing have made possible the essential functions of this collaborative resource evolution (see section 5 for scenario examples of the user-driven development process). 3.3 Essential Testbed Functions Scholars or researchers in any specific discipline can report their reconstructions of Peirce's complex arguments, as hypotheses, based on their interpretations of the imaged and transcribed manuscript content. Editors and other scholars or researchers are then able to track their own hypotheses as well as those of others, from the image evidence supporting them, through the systematic deduction of their implications for manuscript order and content, to the ongoing inductive testing by scholars or researchers who interpret and employ content of the reconstructed arguments. Not only is the order of pages in a manuscript frequently hypothetical because of lost or misplaced pages, but Peirce's complex style of composition increases the uncertainty of accurate reconstruction. He sometimes uses the same page in several versions of elaborated discussion, the course of which sometimes even doubles-back on itself to pick up an unexplored path. In our future scenario, each scholar in the testbed creates a map of the page-bypage course of a particular hypothetical Peircean "exploration." These computergenerated diagrams record each (possibly unique) reconstruction of Peirce's multipath arguments5. These "S-diagrams" can then be matched, by either human perception or computer-generated graph methods. Detected relative invariance among maps reveals agreements in sequence across all scholars' hypothetical reconstructions, and pin-points where controversies in reconstructions lie and further investigation should be pursued. As interpretations proceed and scholars' hypothetical reconstructions evolve, a "dynamic map" is generated, which continues to display the evolution of the collaborative effort (see section 5). Many tools (for 5 As early as 1986, one Peirce scholar (Shea Zellweger) mapped on paper the pattern of several Peirce manuscripts, creating an elaborate, page-by-page, "organic" figure of the course of his writings (see 1 for examples). database, document, and knowledge-based management, search and retrieval, knowledge acquisition, interlingua for both natural language translation and system integration, knowledge-based communication services support and discourse management) have been integrated and customized to serve the scholars’ reporting, tracking, and mapping needs (de Moor suggests an architecture [10:269-70]).