| Learning and Computing | Education | Computing | Psychology | Artificial Intelligence |


Case study has long been recognized as a primary tool for exploration and theory development in the behavioral and social sciences. Given that case studies -- in all their particularity and complexity -- can not be replicated, their most productive scientific use requires that case study materials be maximally available for secondary analysis. The enhancement of recording technologies during recent decades has made significant strides in permitting such secondary analysis. Comparable advances in the area of interpretation are required; such appear within our reach, if not within our immediate grasp, through the well directed exploitation of computer technologies.

Exploiting Computing Technology for Secondary Analysis

The classic example of case studies used for the illumination of differently based interpretations is Robert White's "Lives in Progress". White provided interpretations of three of his case studies based on the biological, the psychodynamic, and the social and cultural views of man. That work is a classic for psychological education, primarily because it so well illustrates the theories. What is less well known is that White's remaining studies exist in archives at Radcliffe's Murray Research Center. It is possible at this time, with sufficient knowledge, access, and support, to construct a new thing, a database of cases (call it a DataCase) organized around White's interpretations in his classic text but more thoroughly undergirding (and perhaps contradicting, even disconfirming) the notions published in his text. The first notion of a datacase then is that it represents a textual layer, interpretive in character and theoretically rich, supported by case detail both as exemplifications of notions and as existence proofs for phenomena. White's work can be made more available for secondary analysis through its embodiment in a DataCase. His work is important and exemplary in its commitment to multiple interpretations.

The on-line textual layer of the White Datacase could be created by scanning and re-structuring the text of White's book. The construction of the supporting layer, based on more extensive case study materials, would depend upon access to the archive of those cases [1] The best candidate software concept for organizing such ill-formed records and relations into an information processing system is hypertext. A hypertext datacase will unify the database and text processing functions of information processing systems into a medium for supporting scholarly investigation of data-rich fields, such as the behavioral and social sciences epitomize.

If DataCases also include the construction of epistemological models based on behavioral observation, they will exploit more fully the interpretive potential of computing. That is, simulation of processes comprise a third layer in a datacase. It is not possible to say, beforehand, that historically based studies are suitable for use as a foundation for the construction of epistemological models. White' studies were not collected with such an end in mind. Lawler's studies of his children's development were designed with such an intention, and their analysis has already led to some modelling studies [2]. One primary focus of this project will be the complete exploitation of Lawler's case studies for the construction of a three-layer datacase.

The Three-Layer DataCase

Lawler's case corpora currently include studies of three children, ranging in age from 18 weeks through 8 years. Other work is underway. The preliminary, partial textual layer of a datacase for Lawler's studies exists as the text of two books, in machine readable form, covering the early development of two children. The supporting layer exists as three corpora, much of which can be brought on-line with existing technology and considerable effort. The modelling layer exists in kernel form as a series of interactive machine learning programs, coded in a list processing langauge. The coverage of the text by the models is not complete; such should be considered the normal case -- as it also should be considered the normal case that complete corpora are not covered by the text that purports to interpret their detail. A primary effort of this project will be the extension of this DataCase to cover the existing case corpora, including the study of the third child's developing knowledge of language and space and the functional modelling of existing and new interpretations where possible.

Special advantages of dealing with this material are at least two: first, the preferred position with respect to copyright (the principal investigator owns the material); second, as the designer and builder of this DataCase, the principal investigator will be in an ideal position to show other scholars how to exploit the material (both on-line and in the extended corpora) for their development of alternative interprepations.

Exploiting Scholars' Work for the Future

A second view of a datacase is that it comprises an encyclopedic summum of a scholar's life and work, organized for its optimal exploitation by future scholars. Piaget's extended corpus of writing and experimental work represents a major challenge to the future for its thorough absorption and assimilation by scholars of today and tomorrow. Approaching the lifework of a such productive genius could be done different ways. One might, for example, focus on Piaget's creativity as an experimenter and form a catalog of his empirical work [3]. Alternately, one might think of Piaget primarily as a theorist of mind and focus on theory development, begining a datacase around that theme [4].

As with the studies by White and Lawler, special arrangements with respect to available materials and copyrights are very important. The Principal Investigator will seek advice and guidance members of the Archives Piaget and colleagues and collaborators of Piaget.

The intention behind constructing the Piaget DataCase is to bring his work into a form most suitable for future exploitation by students of mind. Since the future is still unclear however, this implies a need for maintaining the current structure and organization of the Piagetian corpus while admitting the need for later free-form linking, indexing, component extraction, and the development of tools for re-structuring based on ideas of future scholars. Various future reorganizations will be possible because there need be no single Piagetian DataCase. Much as commercial systems are "backed up" by making copies that are kept over time, the Piagetian DataCase will be backed up -- but differently because it will be copied and modified many times for various future purposes. The tree of possible reorganizations of the Piagetian corpus will blossom broadly. Time will tell which are the primary branches and what limbs will die and be lopped off. It is important that there be a well established root and bole lest future variations generate more chaos than creative exploitation.

Articulating the Knowledge of a Developing Science

At the end of his productive years, Norbert Weiner was approached by a younger mathematician who urged him to bring his knowledge to bear in organizing his own work for its best use by future students. Weiner preferred to spend his time working on harder, more personally interesting problems. Minsky, once that younger mathematician, has agreed to join in this effort because the agenda is not so egocentric as construction of a monument to a man, it is rather for exploiting today's technology to create a new tool for our scientific community. The DataCase Minsky will begin simply -- it has already begun in a modest way. Minsky's Society of Mind [5] comprises a textual layer of accessible introduction to themes in the nature of knowledge and its embodiment in mind, such as we know it now. Some essays point to the past, to what we know already. Other essays point towards to the future, suggesting new ideas for further development and evaluation. The 300 published essays in Minsky's book -- a portion of the original essays available -- can be extended, both with essays originally omitted from the printed version and with new work. The organization of material, sequential and chunked into chapters in the book, can be supplemented in the machine readable form by dynamically compiled, user-modifiable indices. The notions represented by the essays of Minsky's text can be supported or undermined by additional materials -- from various sources -- constructed as the underlying layer of a DataCase. Computational models of ideas, some already constructed by Minsky's students, can serve today as exemplifications of ideas. Minsky now holds copyright and editorial control over the content of his hypertext Society of Mind. He can offer his students the opportunity to extend his original text; he can offer them the opportunity to improve his original text. They can offer him aid in developing a shared, publicly articulated view of intelligence. Some day Minsky will yield control over editing the Minsky DataCase to others. It will become community property explicitly, as it is now implicitly. Then, perhaps even before then, variations of the Minsky DataCase should be expected to occur and develop in their own ways. Multiple variations will be the norm, not the exception; those variations, because occurring in machine readable form, will be more easily specifiable through machine embodied comparison of the DataCase variants themselves.

Where does the datacase go beyond commentary and extension ? Down, to the detail of supporting materials; here is where the datacase earns its name as materials are linked into the structure of ideas to either support or disconfirm those ideas. Up, to simulations and alternate representations of the ideas and the evidence bearing on those ideas; models first will appear as demonstrations of notions appearing in the textual layer. Later, as more work is done on the themes of extending modelling to cover the notions of the textual layer, there will emerge a computational layer amounting to a functional redescription of the view of mind advanced by Minsky and his colleagues. Ultimately, the modelling layer will become a computational epistemology, one growing out of the human experience but open to more general characterizations of what mind is and might become.

The Interrelation of Multiple DataCases

These particular cases have not been advanced merely because they are convenient and thus possible to construct. The ideas and research have long been intimately related. Lawler's case studies were inspired by Minsky's observations on challenges facing the field of Artificial Intelligence (AI) in the mid-seventies. Neither the studies not the results are trivially derivitive. One problem with mid-seventies theories of mind is that they were underdetermined -- this is still true, in that one can represent functioning in many ways. Lawler's studies did not constrain theorizing in AI, nor was such their intention. They have however, suggested some non-obvious starting points for knowledge development and some mechanisms as effective which were not prominent in theory at the time.

Piaget and McCulloch, in an earlier generation, sensed the distant compatibility of biological and computational epistemologies. Minsky's Society of Mind began as a collaboration with Papert, for five years Piaget's mathematician. Symbol-oriented machine learning studies began in the sixties with the notion of procedural composition. Piagetian structuralism, requiring more complex structures to embody knowledge than the procedures and combined-primitives of the sixties, offered a richer vision of mind, in psychological terms, and thus appeared as an addressible, compatible challenge to early machine learning research: compatible because structural; addressible because computation offers a wealth of mechanisms for embodying structural development beyond simple composition. The challenge is profound; it has not yet been adequately met.

Piaget's case studies are an obvious inspiration for Lawler's studies. In contrast, however, note that Piaget's studies were based on the unified interpretation of case material based on three infant studies. Lawler's most thoroughly developed study was based on his second subject. (The work with the first subject could be taken as an extended, exploratory pilot study for the more detailed later study -- thus influential in shaping the work but not integral.) Lawler's further studies involve interpretive work still underway and going beyond the initial study significantly in the range of themes and issues addressed -- relating themes of language development with the Piagetian focus on knowledge about objects and space.

The intention of this project is to construct three primary DataCases, with the White DataCase as an important prototype. In this effort, the participation of Minsky and Lawler, scientists owning their own material, will be a significant advantage. Further, participating as designers and developers of the implemented DataCases, they will be the best situated persons to guide others in use of the facilities for secondary analysis, a primary focus of the effort. Their collaboration over the long term and into the future will undergird the attempt to interrelate their different DataCases and that of Piaget as well; it will be a major advantage in this effort to enjoy the participation of Cellerier, whose professional commitment has long been the adaptation of notions of Piagetian structuralism to the novelties of function-oriented psychologies. To the extent that other primary collaborators of Piaget, Minsky, and Lawler (such as Gruber, Inhelder, Papert, and Sinclair) are willing to invest their time and energy in the effort, that will also support the effort to deal with these issues and information in the most profound way possible.

Purdue University

Publication notes:

Text notes:

  1. The principal Investigator has requested an appointment as a visiting scholar at the Murray Research Center for the summer of 1991 to begin this work.
  2. Lawler's studies are respected in a community of scholars who take developmental case studies seriously. Sheldon White has described Lawler's study of his daughter's cognitive development as "...the finest single study of children's learning we have, in care, in detail, in breadth, and sensitivity of perspective...The work of The Intimate Study stands as a model of the way a child's thinking should be examined." Barbel Inhelder has noted that it is also "... The first highly convincing synthesis of cognition science and genetic psychology. An innovative study which highlights the computational approach to new understandings of the growth of mind."
  3. This approach has been followed by the Center for Educational Research and Innovation of OECD, Paris, which commissioned development of an experimental catalog based on Piaget's work by Professors Pauli, Nathan, and Grize. The Principal Investigator has begun a joint project with OECD to bring their catalog of experiments on-line as a Macintosh hypercard stack.
  4. This focus is preferred by Piaget's close colleagues. H. Gruber, for example, who followed Inhelder in Piaget's chair at Geneva, focusses on Piaget's intellectual development in his book (with Vonèche), "The Essential Piaget", subtitled "An Interpretive Reference and Guide". The Principal Investigator has begun exploring with Gruber construction of a prototype hypertext version of The Essential Piaget.
  5. Published as a paper book by Simon and Schuster, New York, 1986.
Note (1997): The work projects mentioned in notes three and four have been carried on by other people, centered in the Archives Piaget and the University of Geneva. Clearly they are better positioned and in fact more deeply knowledgable about Piaget's work. I am pleased if my enthusiasm for this sort of project has urged others into completing tasks beyond my capability.

Learning and Computing | Education | Computing | Psychology | Artificial Intelligence |