1:
2: \documentclass[12pt]{article}
3: \usepackage{fullpage,isolatin1,times,verbatim}
4: \usepackage{psfig}
5:
6: \title{Towards Solving the Interdisciplinary \\ Language Barrier Problem}
7: \author{Sébastien Paquet}
8:
9:
10: \renewcommand{\baselinestretch}{1}
11: \thispagestyle{empty}
12:
13:
14: \begin{document}
15:
16:
17: \bibliographystyle{plain}
18:
19: \maketitle
20:
21: \vspace*{3.0cm}
22:
23:
24: \emph{There are two great secrets to success
25: in life. The first is to not tell
26: everything you know.}
27:
28: Anonymous
29: \vspace*{1.0cm}
30:
31: \emph{Everybody laughs in the same language. }
32:
33: Anonymous
34:
35: \subsection{Abstract}
36:
37: This work aims to make it easier for a specialist in one field to find
38: and explore ideas from another field which may be useful in solving a
39: new problem arising in his practice. It presents a methodology which
40: serves to represent the relationships that exist between concepts,
41: problems, and solution patterns from different fields of human
42: activity in the form of a graph. Our approach is based upon
43: generalization and specialization relationships and problem
44: solving. It is simple enough to be understood quite easily, and
45: general enough to enable coherent integration of concepts and problems
46: from virtually any field. We have built an implementation which uses
47: the World Wide Web as a support to allow navigation between graph
48: nodes and collaborative development of the graph.
49:
50:
51: \section{Introduction}
52:
53: \subsection{The Ills of Specialization}
54:
55: The problem we are concerned with in this work is extremely common in
56: today's world. Anyone who has consulted medical doctors without having
57: himself or herself a good grasp of medicine has encountered a language
58: barrier first-hand : whereas the doctor (usually) knows what
59: he\footnote{This text uses the masculine to include the feminine.} is
60: doing and will be able to discuss it with other experts in the same
61: field, his patient will understand very little of what is going on in
62: such a dialogue, because it will be replete with terms that mean
63: nothing to him. Moreover, frequently the doctor will not be able to
64: fully convey his understanding to the patient. This can be a
65: frustrating predicament, especially as the issues under discussion may
66: greatly matter to patient.
67:
68: To \emph{specialize} is to concentrate on a particular, necessarily
69: restricted activity or field of study. Observing the contemporary
70: world, it is not difficult to discern a massive trend towards the
71: specialization of humans, which translates into an explosive growth in
72: the number of fields of human activity. Specialization is a way of
73: increasing overall efficiency; one of the better known examples is
74: Henry Ford's introduction of the assembly line for manufacturing
75: cars. By assigning to each member of the production team a single task
76: to be performed on the cars as they passed by slowly down the assembly
77: line, Ford realized a tremendous gain in efficiency compared to
78: traditional assembly methods.
79:
80: While specialization is desirable from that particular point of view,
81: it also has a downside: the more a person becomes specialized, the
82: less he can meaningfully discuss problems that matter to him with
83: other people. In effect, the specialist is very often restricted to
84: collaboration with other specialists in the same area. As one gets
85: more and more specialized, pools of colleagues grow ever smaller. This
86: is unfortunate because discussions with specialists in another area
87: often prove to be fertile ground, as ideas and strategies which were
88: first developed in one field often turn out to be adaptable to a
89: problem in a different field. Indeed, breakthroughs often result from
90: interdisciplinary collaboration: it is not unusual that effective
91: tools for tackling a long-standing problem in an area are found in
92: another area. For instance, the recent scientific successes of genome
93: sequencing owe much to the collaboration between biologists and
94: computer scientists. Another adverse consequence of the isolation
95: arising from specialization is that many people reinvent the wheel for
96: themselves because they are unaware of similar work that has been done
97: elsewhere.
98:
99: Interdisciplinary communication is thus desirable from the point of
100: view of progress; that is, it is helpful in solving problems,
101: especially the more important and challenging ones. Consequently,
102: finding efficient ways of communicating with outsiders is becoming an
103: increasingly pressing problem for people who are not content with
104: speaking only with an inner circle of colleagues. A way to alleviate
105: the problem is to learn another specialty, which will provide
106: opportunities for discussion with a larger circle of people. However,
107: learning a specialty usually involves a considerable time investment;
108: moreover, there are so many different specialties\footnote{Crane and
109: Small reported that there were 8530 definable knowledge fields in
110: 1987. It is reasonable to presume that the figure has grown
111: significantly since then.} that even selecting a promising one can be
112: a difficult problem in itself.
113:
114: \subsection{A Language Issue}
115:
116: Human communication is an activity whose goal is to convey meaning
117: from one person to another. Language is a set of signals (such as
118: words, images or gestures) which serves to conduct it. In order for
119: two people to communicate, it is necessary that they agree on signals
120: and their meanings. A language barrier exists whenever the signals one
121: person uses are not recognized or have different meaning for the other
122: person.
123:
124: Since different specialties are concerned with different things,
125: specialists develop different languages that enable them to
126: communicate efficiently amongst themselves. The advantage in
127: developing a specialized language is that of conciseness. Expressing a
128: complex, multi-faceted construct in a single word is preferable to
129: always using everyday (commonly used) language. This is because the
130: unequivocal description of the construct in everyday language is
131: usually much longer. As a case in point, consider the fact that
132: mathematical equations used to be expressed in words. Following the
133: development of a symbol system providing shorthand representations for
134: recurring expressions (i.e. \emph{x} being substituted for the phrase
135: ``the unknown''), it became easier to think about complicated
136: mathematical problems. However, at the same time an additional
137: requirement was imposed unto whoever wanted to understand what
138: mathematicians were doing: it became necessary to learn their
139: specialized language.
140:
141: In the time of Leonardo da Vinci, it was possible for a dedicated
142: individual to become reasonably versed in most existing disciplines,
143: providing ample occasion for cross-fertilization between
144: fields. Sadly, nowadays this is no longer possible. The
145: interdisciplinary language barrier problem is universal, because one
146: can be trained in no more than a few specialties, and consequently
147: cannot easily understand the languages that are used in the other
148: specialties.
149:
150:
151: \subsection{Connections Between Languages}
152:
153: How does one convey the meaning of words\footnote{in what follows, we
154: use ``word'' to mean any kind of signal.} belonging to a language
155: which they understand to someone who has never heard of them before?
156: (This is a central problem in teaching.) A well-known strategy
157: consists in using words known to both teacher and student to explain
158: what a new words means. Those will either be words taken from everyday
159: language, or specialized words which are known to be understood
160: because an area of overlap exists between the teacher's and the
161: student's backgrounds.
162:
163: We wish to stress the point that this overlap, the common ground
164: between teacher and student, can be a major determinant of the
165: efficiency of this process. For instance, if both master the language
166: of mathematics at a sufficiently advanced level, the teaching of
167: elementary physics should be greatly facilitated. This is because
168: numerous and deep connections have been established between those two
169: fields. Thus a learner acquainted with the concept of derivative will
170: most likely quickly understand that of velocity, seeing it as a
171: special case of a derivative of a spatial variable with respect to a
172: temporal variable. This \emph{specialization relationship} between
173: concepts can be seen as a
174: \emph{shortcut} from one field to the other, since understanding one of
175: the concepts enables one to understand the other quickly, \emph{without
176: building it from the ground up}. A similar argument could be made
177: concerning the situation of someone familiar with the concept of
178: velocity who wishes to understand that of derivative. If we abstract
179: out the labels of space and time from the concept of velocity we get
180: that of derivative. In that direction, the shortcut is rather a
181: \emph{generalization relationship}. Specialization and generalizations
182: are two sides of the same coin: the former consists in putting
183: particular things in as-yet-empty boxes, while the latter consists in
184: removing them.
185:
186: The last example referred to recognizing something new as a particular
187: form of something that is already understood, and vice
188: versa. Sometimes connections between concepts are not as obvious. For instance, a
189: physicist who is introduced to the concept of chemical reaction speed
190: might perceive a similarity between it and the concept of velocity,
191: which might help him grasp the former concept. What happens here is
192: that both concepts may be seen as special cases of a common, more
193: general concept: that of a derivative with respect to time. In this
194: case the shortcut is an indirect, two-link path from the particular to
195: the general and back to the particular. Note that even though the
196: learner may perceive the similarity between two concepts, he may not
197: readily express the common abstraction in words.
198:
199: Shortcuts of the three types mentioned above (specialization,
200: generalization, and similarity) have much value because they enable
201: learners to quickly learn about a new topic, capitalizing on their
202: mastery of a specialty. Having access to a large, organized body
203: of such shortcuts, would provide specialists with a quick and rather
204: painless way of learning about new topics. It would consequently
205: enable them to discuss meaningfully with specialists from other fields,
206: thereby giving them a chance to overcome the interdisciplinary
207: language barrier with much less effort than is currently required.
208:
209: The natural question which arises from the previous discussion is the
210: following one: are there really that many shortcuts between the
211: concepts belonging to different fields of human activity? If so, are
212: there ways of identifying them, organizing them, and sharing them?
213: The goal of the present work is to examine those questions and provide
214: elements of answers to them.
215:
216:
217: \subsection{Mathematics as ``Shortcut Science''?}
218:
219: The previous examples, as well as simple observation of the methods
220: used in several sciences, point to the discipline of mathematics as a
221: kind of substrate underlying many specialized concepts. Indeed,
222: mathematics does provide a kind of integrative unity to a very large
223: body of knowledge comprising most of physics and chemistry, parts of
224: engineering, computer science and economics, etc. One could think of
225: it as a well-organized, useful network of passageways between those
226: fields. A mastery of mathematics will definitely help someone learn
227: any of those fields.
228:
229: However, the network is far from completely connecting every concept in
230: every field and subfield. If such were the case, able mathematicians
231: would likely rule the world. What is missing? What kinds of knowledge
232: have not been abstracted out and organized by mathematicians?
233:
234: Insofar as it can be considered as knowledge, \emph{know-how} or
235: \emph{problem solving knowledge}, which relates to the actions
236: that enable one to accomplish a particular task or to reach a
237: particular goal, is an obvious and glaring omission. Among other
238: disciplines, computer science has made a few forays into this area,
239: especially as regards the \emph{specification} of know-how, but a
240: proper and general \emph{organization} of know-how is nowhere to be
241: found at present. Many problem solving strategies have been
242: identified, a few relationships have been identified between them, but
243: for most problems that appear for the first time there is currently
244: nowhere to look for potentially appropriate strategies.
245:
246: Problem solving is a central human activity. Most occupations are
247: defined by the set of problems they are concerned with. Problem
248: solving knowledge that is specific to an occupation is often referred
249: to as the ``tricks of the trade'', and (sadly) is seldom well
250: documented. Acquiring it quite often involves a sometimes costly
251: process of trial-and-error which is called ``experience''. More often
252: than not, success in an occupation depends on achieving mastery of
253: those tricks. The usefulness of properly documenting problem solving
254: processes lies, among other things, in the potential it has for
255: reducing the amount of trial-and-error that is necessary to learn a
256: trade.
257:
258: \subsection{Scope of this work}
259:
260: \begin{comment}
261: \emph{I am sending you a longer letter than usual, because I did not
262: have sufficient time to make it shorter.} - Blaise Pascal
263: \end{comment}
264:
265: This work makes the (admittedly audacious) suggestion that a simple
266: approach laid upon the foundation of specialization and generalization
267: relationships and problem solving strategies may enable concepts,
268: problems and solution strategies from all fields of human endeavour
269: which are describable by language to be structured together in a
270: coherent, usable manner. This structure may be seen as a large graph
271: linking these entities together. Moreover, it proposes a practical
272: means of building and providing access to such an organized body of
273: knowledge which is based on currently available technology.
274:
275: Of course, our conjecture cannot be proved true without exhaustively
276: organizing all (symbolic) human knowledge, a task which is beyond the
277: reach of any individual. However, demonstrating it to be false would
278: require identifying something which demonstrably does not relate to
279: anything. The mere act of defining such a thing would put it in
280: relationship with something, which would create a contradiction.
281:
282: Having put aside questions of absolute truth or falsity regarding our
283: claim, one might still legitimately attempt to evaluate the
284: practicality of the methodology presented here. In order to address
285: this question, we have applied the methodology to produce a
286: small-scale prototypical structure comprising concepts from
287: diverse fields of activity, to test the practicality and the general
288: applicability of our approach.
289:
290: \begin{comment} We have also documented the process by
291: which the prototype was created, so that the reader interested in
292: verifying for himself whether the approach is usable may do so easily.
293: \end{comment}
294:
295:
296: \subsection{Citation Warning}
297:
298: Texts such as this one are usually replete with references to related
299: work by other researchers. As the topic under discussion here is quite
300: broad, every reader will undoubtedly find relationships between this
301: work and other work, perhaps even his own. Without the shadow of a
302: doubt, the present work reinvents several wheels which have already
303: been invented in other fields of study.
304:
305: However, the reader will find a remarkable paucity of citations in
306: this document. There are two reasons for this state of affairs, one
307: benign, the other more serious. The first is that the author will
308: never have time to read and understand everything in every discipline
309: which could be related to this work. The second is that, because the
310: problem considered here is intrinsically interdisciplinary, the
311: author doesn't have at his disposition proper \emph{means} of finding
312: everything which could be related to this work. This unfortunate
313: predicament actually constitutes the original motivation for this work.
314:
315:
316:
317: \subsection{Structure of the Document}
318:
319: The remainder of this document further explains and provides support
320: for what has been suggested in this introduction.
321:
322: Section \ref{approach} explains our approach to the problem. It is
323: divided in two. Subsection \ref{graphs_nodes_links} explains the
324: basics of graphs, which we subsequently use to explain our
325: approach. Other ways of describing the same approach are possible, but
326: the idea of a graph is intuitive and allows for easy-to-grasp visual
327: representations.
328:
329: Subsection \ref{concepts} argues that any concept
330: that is expressible in a language is defined by way of its
331: relationships to other concepts. It describes the basic structure of
332: a type of graph node which may serve to represent a concept, and types
333: of links which may originate from such a node.
334:
335: Three particular kinds of concepts are central to our work and are
336: explained in Subsection \ref{special_concepts}. They are: problems,
337: solution patterns, and strategies. Another useful concept is that of
338: domain, which is explained last.
339:
340: Section \ref{implementation} describes our prototypical implementation of
341: a concrete example of a product of our approach. We conclude in
342: section \ref{conclusion} by summarizing this work and providing
343: directions for future research.
344:
345:
346: \section{A General Approach to Organizing Concepts and Problems}
347: \label{approach}
348:
349: \subsection{Graphs, Nodes and Links}
350: \label{graphs_nodes_links}
351:
352: A \emph{graph} is a set of abstract entities called \emph{nodes} that
353: are connected by abstract entities called \emph{edges}. The word
354: ``edge'' has a geometrical connotation which may be confusing in the
355: context of this work; we use the word ``link'' instead. Links
356: represent relationships that exist between nodal entities. A link
357: which represents an asymmetrical relationship (for instance, the
358: relationship between parent and child) has a direction, that is, one
359: of its endpoints is labelled as the origin node and the other is
360: labelled as the destination node. Sample visual representations of
361: graphs are found in figure
362: \ref{graph}.
363:
364: \begin{figure}[htbp]
365: \centerline{\hbox{\psfig{figure=graph.eps,height=3.0cm}}}
366: \caption{Examples of graphs. Nodes are represented as circles and
367: links as line segments (indicating symmetrical relationships) or
368: arrows (indicating asymmetrical relationships).}
369: \label{graph}
370: \end{figure}
371:
372: The concept of a graph is useful in a multitude of contexts, for
373: instance for representing transportation or communication networks. It
374: has also been extensively used to represent relationships between
375: concepts. Our approach uses nodes to represent concepts, and links to
376: represent specific kinds of relationships that exist between them.
377:
378:
379: \subsection{Concepts}
380: \label{concepts}
381:
382: Dictionaries define the meaning of words using other words. Teachers
383: do the same with students who are learning new ideas. This suggests
384: the general statement that a concept that is expressible in a language
385: is defined by way of its relationships with other concepts. This is
386: not something which can be proved. The only way to go further in this
387: work is to temporarily accept that it may be true and see where this
388: leads.
389:
390: We have said that we want to define a type of graph node which may
391: serve to represent any concept. We have already explained what we mean
392: by definition, generalization and specialization. Our proposition is
393: the following. A concept node features:
394:
395: \begin{enumerate}
396: \item a \emph{name}, which is a signal which may be used to refer to
397: the concept among people who understand it \footnote{We remark that
398: words often have multiple meanings. This means that one particular
399: expression may appear as a name in a number of distinct concept
400: nodes. Obviously, the rest of the node contents disambiguates the name.};
401: \item a \emph{definition} which contains links to the concepts which
402: are involved in it;
403: \item a set of links to concepts that are \emph{generalizations} of
404: the concept;
405: \item a set of links to concepts that are \emph{specializations} of
406: the concept;
407: \item a set of links to other \emph{concepts} whose definition
408: directly involves the concept;
409: \item a set of links to \emph{problems}\footnote{We explain what we mean by a
410: problem in section \ref{problems}.} in which the concept is directly involved; and
411: \item a link to a \emph{domain}\footnote{We explain what we mean by a
412: domain in section \ref{domains}.} which this concept primarily relates to.
413: \end{enumerate}
414:
415: We give an example of a concept node which could describe
416: the concept of an ordinary quadratic function.
417: \begin{enumerate}
418: \item The name is the expression ``real-valued quadratic function of a real
419: variable''.
420:
421: \item The definition is the phrase
422: ``A \emph{real-valued polynomial function of a real variable} involving
423: \emph{terms} of the \emph{second degree} at most.'' Italicized words indicate that
424: links to the corresponding concept nodes exist.
425:
426: \item Generalizations could include, but are not limited to:
427: ``polynomial function solvable by radicals'', ``function having a
428: global extremum'', and ``high school level mathematical
429: concept''. Note that the first generalization classifies the concept
430: from the standpoint of algebra, the second one highlights a property
431: that could be useful in an optimization context, and the third
432: indicates a property that could be useful in an educational
433: context. The three are quite different; their relevance depends on
434: what one is looking for.
435:
436: \item Specializations could include:
437: ``real-valued quadratic function of a real variable having no real
438: root'', ``real-valued quadratic function of a real variable having
439: one real root'', ``real-valued quadratic function of a real variable
440: having two real roots'',
441: ``real-valued quadratic function of time'',
442: ``real-valued quadratic function of a spatial variable''.
443: Here the first three specializations have a mathematical flavour
444: while the latter two have a physical flavour.
445: \end{enumerate}
446:
447: Let us recall at this point that the overall goal is to obtain a graph
448: that is easy for humans to use. This implies that someone who is
449: creating a node must strive to ensure that none of the above two lists
450: grow to become unmanageably large. In order to achieve this, a guiding
451: principle needs to be applied. Our prescription is the following:
452: \emph{systematically hide complexity by introducing new, meaningful nodes
453: wherever possible}. What this means for instance is that instead of
454: listing ``real-valued quadratic function of straight-line distance'',
455: ``real-valued quadratic function of geodesic distance'', etc. as
456: specializations, one should instead bundle them together into
457: ``real-valued quadratic function of a spatial variable''. The same
458: principle should be applied for generalizations. A similar principle
459: is often applied successfully in the context of writing, where
460: explicit structure serves to help the reader deal with the overall
461: complexity of the text.
462:
463: \begin{enumerate} \setcounter{enumi}{4}
464: \item An example of a concept whose definition directly involves this
465: concept is that of a paraboloid.
466:
467: \item Problems could include links to the following problem
468: nodes: ``Obtain the roots of a real-valued quadratic function of a
469: real variable'', ``Find the extremum of a real-valued quadratic
470: function of a real variable'', ``Teach the concept of quadratic
471: function''.
472:
473: \item Finally, the domain link could connect the concept node to the domain of
474: mathematics.
475:
476: \end{enumerate}
477:
478:
479: \subsection{Special Concepts}
480: \label{special_concepts}
481:
482: In the last subsection, we defined the structure of nodes which
483: represent concepts. Here we define slightly different structures for
484: nodes which describe problems, solution patterns, strategies, and domains.
485:
486: Subsection \ref{problems} explains what we mean by a problem. It
487: describes the basic structure of a type of graph node which may serve
488: to represent a problem, and types of links which may originate from
489: such a node.
490:
491: Subsection \ref{solution_patterns} explains what we mean by a solution
492: pattern to a problem. It describes the basic structure of a type of
493: graph node which may serve to represent a solution pattern, and types
494: of links which may originate from such a node.
495:
496: Subsection \ref{strategies} explains what we mean by a strategy. It
497: describes the basic structure of a type of graph node which may serve
498: to represent a strategy, and types of links which may originate from
499: such a node.
500:
501: Subsection \ref{domains} explains what we mean by a specialized domain
502: of activity and argues that a domain can be approximately defined by
503: the set of concepts, problems and solution patterns which it is
504: concerned with. It describes the basic structure of a type of graph
505: node which may serve to represent a domain, and types of links which
506: may originate from such a node.
507:
508: At this point in our exposition, we would suggest that the reader take
509: a look around our prototype implementation, which is described in
510: section \ref{implementation}. This should make it easier to understand
511: what follows.
512:
513:
514: \subsubsection{Problems}
515: \label{problems}
516:
517: Up until now we have talked about problems without defining more
518: precisely what we meant by that term, hoping that the reader has
519: encountered enough problems in his life to have an intuitive grasp on
520: that concept.
521:
522: Our conception of a \emph{problem} is similar to that proposed by
523: Polya\cite{Polya}: \emph{a goal in a context which we can express
524: clearly, but that is not immediately accessible}. This is a very broad
525: definition. Examples of common problems are: to find something to eat
526: in a cluttered refrigerator, to obtain recognition from others, and to
527: understand something.
528:
529: What constitutes a problem for a person is not necessarily a problem
530: for another. For instance, for most people getting out of bed is an
531: immediately accessible goal, while this is far from being the case for
532: a baby. Thus a possible line of action for someone who does not know
533: how to solve a problem is to collaborate with someone who does. (This
534: may be a problem in itself.)
535:
536: A problem is fully specified when both goal and context are
537: unequivocally specified in terms of concepts. We propose to represent
538: problems in nodes which feature:
539:
540: \begin{comment}
541: - A link to solution patterns that this problem is involved in.
542: \end{comment}
543:
544: \begin{enumerate}
545: \item a (possibly empty) set of \emph{names} used for referring to the
546: problem;
547: \item a \emph{description} of the problem, comprising goal and context;
548: \item a set of links to \emph{generalizations} of the problem; that is,
549: instances of the problem with a generalized form of the goal, a
550: generalized form of the context, or both;
551: \item a set of links to \emph{specializations} of the problem; that is,
552: instances of the problem with a specialized form of the goal, a
553: specialized form of the context, or both;
554: \item a set of links to \emph{solution patterns} to more complex
555: problems which require, as a substep, that this problem be solved,
556: providing motivation for solving the problem;
557: \item a set of links to \emph{solution patterns} that apply to the problem;
558: \item a link to a \emph{domain} which this problem primarily relates to.
559: \end{enumerate}
560:
561: We give an example of a problem node which could describe the problem
562: of solving an ordinary quadratic equation, given that we know the
563: quadratic formula $\frac{-b\pm\sqrt{b^2-4ac}}{2a}$. We ought to warn
564: readers who are used to solving this problem without even thinking
565: about it that seeing the problem from the perspective of someone who
566: is tackling it for the first time will likely require some effort.
567:
568: \begin{enumerate}
569: \item A name could be ``Solve an ordinary quadratic equation''. Note
570: that it is easier for a specialist to refer to the problem by its
571: name, while it is easier for a non-expert to refer to the problem by
572: its longer but more explicit description.
573: \item The description of the problem could be the following:
574: ``Obtain the \emph{roots} of a \emph{real-valued quadratic function
575: of a real variable}, using the \emph{quadratic formula}''. Links to
576: the relevant concepts, i.e. that of a root, that of a real-valued
577: quadratic function of a real variable, and that of the quadratic
578: formula are provided.
579: \item An example of a generalization of the problem could be:
580: ``Obtain the roots of a polynomial function solvable by radicals,
581: using an appropriate formula.''
582: \item A (perhaps a little contrived) example of a specialization of
583: the problem could be: ``Obtain the roots of a real-valued quadratic
584: function of a real variable of the form $ax^2+c$, using the quadratic
585: formula.''
586: \item An example of a solution pattern which involves solving this
587: problem could be that consisting in finding the intersection of a line
588: with a parabola by (1) finding the corresponding quadratic equation; and
589: (2) solving it.
590: \item A solution pattern that is applicable to the problem
591: could be that consisting in computing the solution by feeding the
592: polynomial coefficients into the quadratic formula.
593: \item Finally, the domain link could connect the problem node to the domain of
594: mathematics.
595: \end{enumerate}
596:
597:
598: \subsubsection{Solution Patterns}
599: \label{solution_patterns}
600:
601: Solving a problem amounts to finding a precise line of action which
602: will enable one to attain the goal. The line of action is made up of a
603: series of subproblems to be solved. This is what we mean by the term
604: \emph{solution pattern} \footnote{The equivalent term in computer
605: science is \emph{algorithm}. \emph{Procedure} is another word which
606: refers to the same concept}. Finding (and remembering) a solution
607: pattern is interesting because it is reusable once found: if the same
608: problem arises again, it suffices to apply the same pattern again to
609: solve it. For instance, once the standard solution pattern to ordinary
610: quadratic equations has been remembered, any such equation can be
611: solved without hesitation. The more general the solution pattern, the
612: wider its applicability. Of course, a very general solution pattern
613: can hide much of the actual complexity because the subproblems it
614: features can be complex problems in themselves. By contrast, a very
615: specialized solution pattern is quite straightforward to put into
616: application, but has a correspondingly restricted applicability.
617:
618: A solution pattern is fully specified when all the subproblems which
619: compose it are identified. The solution pattern is guaranteed to be usable
620: if every one of its subproblems has a known solution pattern. If such
621: is not the case, the solution pattern may be usable, but this is
622: conditional upon finding solution patterns to every subproblem.
623:
624: We propose to represent solution patterns in nodes which feature:
625:
626: \begin{enumerate}
627: \item a link to the associated \emph{problem};
628: \item a link to a \emph{strategy} which generalizes the solution pattern (strategies are defined in the next section);
629: \item a \emph{specification} of the solution pattern, containing links to the
630: \emph{subproblems} that it entails;
631: \item a link to a \emph{domain} which this solution pattern primarily relates to.
632: \end{enumerate}
633:
634: We give an example of a solution pattern node which could describe a
635: way of solving an ordinary quadratic function.
636: \begin{enumerate}
637: \item A link to the problem ``Obtain the roots of a real-valued
638: quadratic function of a real variable, using the quadratic formula'' could be provided.
639: \item A link to the strategy ``Obtain a result from an appropriate input and a procedure'' (explained next) could be provided.
640: \item The solution pattern consists of a single subproblem: ``Evaluate the quadratic formula on the polynomial coefficients''.
641: \item Finally, the domain link could connect the problem node to the domain of
642: mathematics.
643: \end{enumerate}
644:
645: \subsubsection{Strategies}
646: \label{strategies}
647:
648: \begin{comment}
649: \emph{I do not think at all that I am able to present here any procedure of
650: investigation that was not perceived long ago by all men of talent;
651: and I do not promise at all that you can find here anything quite new
652: of this kind. But I shall take pains to state in clear words the rules
653: and ways of investigation which are followed by all able men, who in
654: most cases are not even conscious of following them. Although I am
655: free from the illusion that I shall fully succeed even in doing this,
656: I still hope that the little that is presented here may please some
657: people and have some application afterwards.}
658:
659: Bernard Bolzano (1781-1848), quoted by Polya (How to Solve It, 1957)
660: \end{comment}
661:
662: This work makes the (perhaps controversial) assumption that every
663: problem solving process can be modelled as a sequence of two steps:
664: (1) putting the problem in relationship with a very general, abstract
665: form of problem for which the (abstract) solution is known; (2)
666: translating the abstract solution into terms that are specific to the
667: problem. We call \emph{strategy} an abstract problem form accompanied
668: by its associated solution which has wide enough applicability to be
669: used in almost any domain. A strategy captures the essential aspects
670: of a solution pattern. Strategies are interesting because they suggest
671: reusable forms of solution for wide classes of problems.
672:
673: We illustrate the process of going from solution pattern to
674: strategy by considering the following simple problem. Suppose a friend has
675: secretly chosen a number between 1 and 10 and asks you to repeatedly
676: guess what it is. Each time you make a guess he will tell you whether
677: your guess is indeed the secret number. The problem is to find the
678: secret number. You choose a simple solution pattern: you try every
679: number in order, from 1 to 10, until your friend confirms that you
680: have found the secret number. This is an appropriate solution pattern,
681: in the sense that you are guaranteed to find that number.
682:
683: What would be the strategy in such a case? It is often hard to tell
684: what is going on in one's head; we can only make conjectures as to
685: what really happened. Our point of view is that strategies can be
686: identified, which may or may not correspond to the actual mental
687: process. Although knowing more about that process would certainly be
688: interesting, the only goal we are concerned with here is finding an
689: abstract form of the problem for which a valid solution is known. The
690: important thing to recognize is that once we have found a potentially
691: appropriate strategy, it is easy to check whether is indeed qualifies
692: as a generalization of the solution pattern at hand.
693:
694: What can be abstracted out of our problem without essentially changing
695: it? First, note that the concept of number is not essential, for we
696: could have replaced numbers with letters of the alphabet or
697: billiard balls. The important thing is that we had a
698: finite set of possible objects to choose from. Second, what is
699: essential in your interaction with your friend? Surely your friend
700: could be replaced by a machine which tells you whether you have found
701: the right object. The essential thing here is that you have a way of
702: checking whether a given object satisfies the property of being the
703: secret object. Finally, what was the use of counting to ten? It was to
704: provide a way of selecting an untested object at each guess.
705:
706: Thus the abstract version of our problem is the following: ``Find an
707: object satisfying a particular property in a finite set of objects,
708: given a way to check an object for the property and a way to find an
709: untested object''. The strategy solution pattern is the following:
710: ``As long as an object satisfying the property is not found, find an
711: untested object and test it''. People seem to use such a strategy
712: everyday; finding something to eat in a cluttered refrigerator,
713: shopping for clothes or selecting an appropriate screwdriver from a
714: toolbox are examples of problems where this strategy applies.
715:
716: In a sense, strategies are even more useful than domain-specific
717: solution patterns in the sense that they are relatively easy to grasp
718: and apply to large sets of problems. They are also very useful from an
719: interdisciplinary standpoint because they provide clear connections
720: between methods that are used in different fields.
721:
722: We propose to represent strategies in nodes which have a structure
723: similar to that of problems:
724: \begin{enumerate}
725: \item a (possibly empty) set of \emph{names} used for referring to the
726: problem;
727: \item a \emph{description} of the problem, comprising goal and context;
728: \item a set of links to \emph{strategic generalizations} of the
729: problem; that is, strategies that apply to problems involving a
730: generalized form of the goal, a generalized form of the context, or
731: both;
732: \item a set of links to \emph{strategic specializations} of the
733: problem; that is, strategies that apply to problems involving a
734: specialized form of the goal, a specialized form of the context, or
735: both;
736: \item a \emph{specification of the solution pattern} to the problem,
737: containing links to the \emph{strategic subproblems} that it entails;
738: \item a set of links to \emph{domain-specific specializations} of the
739: strategy; that is, domain-specific solution patterns which are
740: specializations of the strategy;
741: \item a link to a \emph{domain} named ``strategies''.
742: \end{enumerate}
743:
744: In the example of the applying the quadratic solution formula, the
745: associated strategy node could take the following form:
746: \begin{enumerate}
747: \item Possible names could include ``Apply a recipe'', ``Do it by the book''.
748: \item The description of the problem is ``Obtain a result from an
749: appropriate input and a procedure''.
750: \item A possible generalization could be the strategy ``Obtain a
751: result from a procedure and an input'', the precise nature of the input (for
752: instance, as being appropriate versus almost appropriate\footnote{ We
753: characterize an input as almost appropriate with respect to a
754: procedure if we have a way of converting it to an input
755: that is appropriate for that procedure.})
756: being abstracted out.
757: \item A specialization could be the problem ``Effortlessly obtain a
758: result from an appropriate input, a procedure, and a helper''\footnote{Calling upon a
759: computer or a graduate student to perform the work should come to the
760: minds of experienced researchers.}.
761: \item There would be a single strategic subproblem in this case:
762: ``Apply the procedure to the input''.
763: \item A domain-specific solution pattern could be the one described in
764: the previous subsection;
765: \item Finally, the domain link would connect to the domain of strategies.
766: \end{enumerate}
767:
768:
769:
770: \subsubsection{Domains}
771: \label{domains}
772:
773: The concept of domain provides a convenient way of classifying
774: specialized concepts, problems and solution patterns which are somehow
775: related together. Almost everyone today thinks of himself as a
776: specialist in a particular domain. Thus we have quantum cosmologists
777: who are physicists who are specialized in quantum cosmology;
778: neurosurgeons who are medical doctors who are chiefly concerned with
779: brain surgery; hitmen who are criminals that specialize in murder; and
780: so on. Being able to label people is useful. Upon learning that
781: someone has expertise in a particular domain, we know that this person
782: masters the concepts and problems of the domain. If we need to solve a
783: problem which we know to be related to that domain, we know that this
784: person may be able to help.
785:
786: We propose to represent domains in nodes which have the following
787: structure:
788: \begin{enumerate}
789: \item A \emph{name} which people who are specialized
790: in the domain use to recognize each other;
791: \item A set of \emph{generalizations} which are disciplines which may
792: be considered to be the roots of the domain; usually this is the
793: domain in which specialists of the domain are first trained.
794: \item A set of \emph{specializations} which are disciplines which may
795: be considered to be the branches of the domain;
796: \item A set of \emph{prominent concepts} which are the central
797: concepts used in the domain;
798: \item Finally, a set of \emph{prominent problems} which are the central
799: problems that a specialist in the domain knows how to solve.
800: \end{enumerate}
801:
802: In the case of mathematics, the domain node could read as follows:
803: \begin{enumerate}
804: \item The name is ``mathematics''.
805: \item A possible generalization could be ``science''.
806: \item Possible specializations could include: ``fundamental
807: mathematics'', ``algebra'', ``number theory'', ``geometry'',
808: ``topology'', and ``mathematical analysis''.
809: \item Possible prominent concepts could include: ``set'', ``number'',
810: ``function'', ``axiom'', ``proof'', and ``definition''.
811: \item Possible prominent problems could include: ``prove a
812: proposition'' and ``find an unknown from data''.
813: \end{enumerate}
814:
815: Domain boundaries are quite fuzzy and constantly evolving. New
816: domains pop up ever more quickly as the trend towards specialization
817: accentuates. On the other hand, some domains become obsolete. Thus we
818: see less and less blacksmiths, alchemists and carriage makers as time
819: goes by. An unfortunate aspect of this kind of natural selection among
820: specialties is that interesting solution patterns may die along with
821: their trade unless those processes are documented before their
822: complete extinction. We remark that a graph such as the one described
823: in this work could obviously represent the forgotten specialties
824: alongside the current ones.
825:
826:
827: \section{An Implementation}
828: \label{implementation}
829:
830: The previous section described a basic architecture for organizing
831: concepts and methods. Here we describe a simplistic but concrete form
832: which this architecture may take. We have built a prototypical knowledge
833: base which uses this basic design. We are certain that many
834: improvements could be made to further facilitate navigation among
835: concepts.
836:
837: The World Wide Web (WWW) consists of documents that are available on
838: the Internet. These documents (``web pages'') are written in
839: hypertext, which means that they may contain links to other
840: documents. There is no restriction on what documents a given document
841: may link to, as long as the target documents are themselves available
842: somewhere on the Internet. Users of the Web use WWW browsing software
843: to view documents and can move over to linked documents with a simple mouse
844: click.
845:
846: Our implementation uses this technology in a very straightforward
847: manner. Each concept, problem, solution pattern or strategy is explained
848: on a single web page, and the relevant links are provided in that
849: page. Every page follows the template corresponding to the kind of
850: concept it describes. This homogeneity enables one to explore the
851: knowledge base confidently once the nature of the basic kinds of links
852: have been understood.
853:
854: Thanks to the way the WWW is designed, the pages which constitute the
855: knowledge base need not all be in the same physical location. This
856: allows different parts to be developed by people working in different
857: locations. Of course, communication is necessary in order to establish
858: links between parts that are developed separately. The prototype is
859: accessible on the Web at the address
860: \texttt{http://www.iro.umontreal.ca/$\sim$paquetse/knoweb/000\_INTRODUCTION.html}.
861:
862: This design has the following strengths:
863: \begin{itemize}
864: \item The architecture is simple and easy to understand.
865: \item Consulting the knowledge base is easy. All one needs is a WWW
866: browser and access to the Internet. Large numbers of people
867: nowadays have such access.
868: \item Contributing to the knowledge base is easy. All one needs
869: is the ability to write Web pages and means of making them
870: available. Large numbers of people nowadays have such means.
871: \item The architecture allows contributors to work together to
872: strengthen their respective contributions by linking them.
873: \item While restrictions exist on the format that is used to represent
874: ideas, there is complete freedom in regard to the contents itself.
875: \end{itemize}
876:
877: It shares the following shortcomings with the basic WWW architecture:
878: \begin{itemize}
879: \item Changing or refining existing parts of the knowledge base may
880: result in large numbers of links coming from other parts of the base
881: being invalidated. Work is then needed to fix those links.
882: \item In order to maximize their usefulness, new contributions must be properly linked
883: to the existing base. This may require a lot of communication between
884: contributors.
885: \item There is no way of ascertaining the quality of contents that
886: has been contributed.
887: \end{itemize}
888:
889: We believe that despite the limitations of this initial design, such a
890: concrete way of organizing ideas has the potential to win the favour
891: of dedicated individuals who not only like to learn but wish to
892: share their understanding with others in an efficient manner.
893:
894: \begin{comment}
895:
896: \emph{"Believe you can, believe you can't; either way, you're right."} -Henry Ford
897:
898: First way to go is textbooks, especially those which have an
899: interdisciplinary flavour. There we find concept definitions, problem
900: definitions, solution patterns which can be put in the form of a graph
901: such as the one described here.
902:
903: Interdisciplinary people are the ones who can help the most to provide
904: shortcuts between fields.
905:
906: Contributors have complete freedom. Total control over their
907: contribution? Copyleft would allow progressive refinement: one could
908: copy what another has done and improve upon it, as happens in science
909: and art. Should enable dynamic, collaborative evolution while avoiding
910: broken links.
911:
912: A single name can have several meanings. So we need a distinct node for
913: each meaning.
914:
915: Need for representations other than text: visual, audio, movies, ...,
916: to facilitate incorporation of certain disciplines and improve
917: accessibility of the material.
918: \end{comment}
919:
920:
921: \section{Conclusion}
922: \label{conclusion}
923:
924:
925: We have proposed a simple methodology which aims at integrating
926: knowledge and know-how across disciplines in a coherent, usable
927: manner. This methodology is centered around
928: generalization/specialization relationships and a problem solving
929: perspective. We have provided a sample implementation of that
930: methodology which builds on the strengths of World Wide Web technology.
931:
932: We hope that the reader feels the importance of the interdisciplinary
933: language barrier problem and sees this work as an interesting first
934: step towards solving it. He is currently actively seeking feedback and
935: help to improve this work. He can be reached by email at the address
936: \texttt{paquetse@iro.umontreal.ca}.
937:
938:
939:
940: \begin{comment}
941: \emph{United we stand; divided we fall.}
942: Abraham Lincoln
943:
944: There probably are mistakes or names which do not reflect common usage
945: in the prototype, because the author was not aware of these
946: names. Likewise, some concepts have not been cleanly dissected because
947: the author did not see the use for the distinction.
948:
949: \end{comment}
950:
951:
952:
953: \bibliography{doc}
954:
955: \end{document}
956: