Chapter 4
INFORMATION PROCESSING & METHOD

The Information Processing Paradigm

Nobody can reproach AI adversaries for being afraid to struggle against idées reçues. Thus, they explicitly question existence or possibility of information processing. All philosophical or phenomenological mechanisms are summoned, and first of all critique of assumptions, allowing to consider others one as dogmas, as we already have seen. Dreyfus thinks that to consider human being as an information processing belongs to this category (p. 164).

Here, the question is again raised of a characterization in term of fact, assumption, or dogma. It can already be said that information processing is as obvious as mental process in Searle – if the words processing and information mean something. Is it necessary to acknowledge them as paradigm, as a fact producing an image? That would have normally to satisfy phenomenologists, if they could link ideas. The psychological assumption of information processing can indeed be questioned, but it doesn’t seem so extraordinary. Rather than an assumption or an a priori, it is about a manner of speaking, a common sense way to say that some data are considered before acting or reasoning. It seems that it would be another assumption that would need a demonstration. Dreyfus himself agrees a usual form of the term information processing, but he counterbalances it by its semantic exclusion under referencialist motives: “Shannon […] carefully excludes as irrelevant the meaning of what is being transmitted.” (Dreyfus, p. 165).

Resistance to information processing possibly results from the anti-elementary obsession, not any more relevant here than elsewhere, since transmitted messages can be analogue. But the Shannon’s problem is only the signal/noise distinction. But this notion of noise could be generalized, from signal meaning (from signifier to signified, according to linguistic terminology). The fact that Shannon is or not aware of it has not any importance. Specifically, one can check gaps of spelling or phonetics, morphology, prosody, or syntax ones, even symbolism, semantics, or pragmatic ones.

This notion of noise represents too the motivation of information transmission, considered by the jargonizing philosophy under the ontological form in the: “emerges in recurrent patterns of breakdown and the potential for discourse about grounding” (Winograd & Flores, p. 69). More simply, the communication is grounded on a transmission of knowledge, between two persons not sharing the same data or the same opinions about world. The hypothesis according to whether knowledge acquisition (for AI) “is an unreachable goal” (idem, p. 98) appears as an empirical contradiction while computer knowledge is not absolute either. The sociologist Philippe Breton’s work, French specialist of the opposition at information theory, and consequently to AI, can be characterized in the same way. And what remains, indeed, when his personal standpoint is removed to his accumulation of arguments is an information on AI history. The vice homage to virtue.

Reasoning Automation

Searle unintentionally gives the solution to the information problem stated by Dreyfus. Once again, he produces a good illustration of the concrete and traditional origin of phenomenological or mentalist ideas:

“From the fact that money can have an indefinite range of physical forms […] it would once again be a miracle if they all produced exactly the same neurophysiological effect on the brain.” (Searle, p. 80).

Actually, the fact that the encoding material support is arbitrary allows, on the contrary, observing that conceptual value, information, is identical. Possibly, this neurological clarifying of the Sapir-Whorf’s hypothesis describes precisely a pathological state that cannot abstract phenomenal differences. Induced concepts (here, capital, income, currency, etc.), are themselves analyzable in abstracted terms, and they possess an intercultural validity [NOTE 60]. Elsewhere, Dreyfus sometimes denies himself this permanent recourse to an extremist reductionist argument himself:

“A computer is a physical object, but to describe its operations, one does not describe the vibrations of the electrons in its transistors, but rather the levels of organization of its on/off flip/flops.” (Dreyfus, p. 179).

The logic of an opposition to information processing can no longer be understood, because the small flips/flops in question are equivalent to hard-wired logical operations, which have some relations to meaning all right. Searle clarifies himself this standpoint when he considers (p. 59) that the cognitivism research program aims to study thought through computer programs (albeit considering programs analysis, rather than simulation, reveals a deductivist philosophical a priori). More generally, the information processing level is equivalent to the conscious processing level, or to reasoning in general as Dreyfus has stated himself (p. 163).

It can also be considered that the different reasoning levels are not directly linked to cognitive equipment (body/brain), but to the context of justification (values/rationalizations). But in both cases, we have what is called, in concerned fields (psychology and sociology, or logic), an information processing. The conscious processing paradigm in the reasoning course can be considered the police inquiry one, under its different variants. Dreyfus precisely admits sometimes this reality, thanks to these canonical fictional metaphors [NOTE 61]:

“This is not to deny that human beings sometimes take up isolated data and try to discover their significance by trying to fit them into a previously accumulated store of information. Sherlock Holmes and all detectives do this as a profession; everyone does it when he is in a very unfamiliar situation.” (Dreyfus, p. 265).

In this connection, we can also consider that decisions of justice are grounded on indices at best – and on the only mentalism at worst. Without going as far as not questioning a judicial procedure, its revision possibility isn’t likely to use the uncertainty principle! Aren’t documents some atoms layout, like drops of blood on the suspect’s jacket?

Thus fundamentally, all the Dreyfus’ intuitionist construction collapses! Because the problem is precisely the difference between problem solving, and an “orderly behavior without recourse to rules” (one of his chapters title), i.e. ritual and mechanical behavior. Finally, his thesis is limited in emphasizing that each time you are not bound to think, you have to act automatically! One of the computer interests, actually costing an extra calculation time, is not to get into the habit. This allows it not neglecting any solution, even if a higher appearance probability can be processed first.

For instance, in one of the Agatha Christie’s novels, the clue rests on the confusion caused by the name Evelyn, rarely male. In an equal criminal case, like in a medical diagnosis by expert system, the delinquent or rare sickness discovery is not useless. And this kind of exhaustive processing is not infinite however, contrary to what claims Dreyfus. By definition, rare cases never are very numerous to each step! And even if these cases were plethoric, a computer can process a large number of data with multivariate analysis, while a human being precisely cannot.

Reductionism Vs Information Processing

Of course, the negation of information processing is once again reductionist. Dreyfus shows up by his lousy empirical examples – when he risks giving some to specify his critiques (without playing on the only argument of past philosophical authority reference, or of current natural sciences):

“At this point, Neisser has recourse to an unjustified theory [according to which…]: ‘Physically, this page is an array of small mounds of ink, lying in certain positions on the more highly reflective surface of the paper.[62]’ But physically, what is there are atoms in motion, not paper and small mounds of ink. Paper and small mounds of inks are elements in the human world. […] What we normally perceive is a printed page. […] No man has ever seen such an eerie world.” (Dreyfus, p. 269).

It seems that this ultra-reductionist quibbling (contradictory with his previous standpoint that we just have seen) rests on the ‘physically’ term about what phenomenologists call ‘human world’. Actually, it should rather be equivalent to cultural world. The sociologist’s point of view could be entitled to claim a monopoly concerning descriptions of human being situated in a cultural context, of which different ‘world view’ have stricken Winograd and Flores (p. 8). Actually, this problem only raises with a philosophical approach as a starting point. The interest of the sociological point of view is precisely that it can observe that perceptions breakdown necessarily appears. When a subject is culturally moved, he tries to laboriously decipher the new culture like a new language. Incidentally, the contradiction can be noticed, for a phenomenologist, of lapsing in universalism.

In addition, concerning this same example, children and illiterates are precisely in front of what is described by Neisser. For printers, graphic arts, or page setting, it can even be said that a page is first seen as these elementary forms. Besides, dot, laser and ink jet printers deny these dubious affirmations. And any magazines reader is found in such a situation at reading of glazed paper magazines, because of reflection canceling contrast. These constraints have to be managed by professionals. More, but is it necessary to specify, it is too much certain that the maximalist pseudo-scientific reductionism of “atoms in motion” doesn’t make the phenomenological thesis truer. Elsewhere, no one has ever seen this ‘eerie’ world. At best, it concerns an attempt to put physicists on his side, by scorning kids, illiterates people, printers, and “the very special detached attitude which comes over a cognitive psychologist sitting down to write a book.” (idem).

The contradiction of the information processing refusal is elsewhere all the more obvious that (current) philosophy is essentially processing information from other fields – like in the successive example of the human brain model (see Progressive Modeling). The only continuity is its law-making pretension. When it is not characterized by showing again the same story, it thinks being indispensable, by imposing dogmatically to data conformity to external principles, while criticizing others’ assumptions.

It is this same import of information, by philosophy, from various fields, which allows it to claim to refute works of AI. Dreyfus speaks about developmental psychology which teaches us, against the information assumption, that “learning of specific details takes place on a background of shared practices […] not as fact and beliefs but as bodily skills for coping with the world” (Dreyfus, p. 47). But nothing prevent programming a background on which details take sense, while this learning seems only concerning procedural programming, instead of declarative one. A simple indexation program can first notice the general topic of a text, then to descend gradually to more precision; and a learned procedure can be used as model for a future adaptive behavior. Except the simple fact that if computer has no difficulties in storing data, why doing without. (There is enough real programming problems without inventing false ones, lying on a bad analyzes, or no analysis at all).

More generally, one may rightfully ask in what the reductionist idea is specifically phenomenological, because phenomena in question are established by sciences, and not by consciousness of subject in action, or by common sense. How philosophy can claim to autonomy by being engaged in experiment or observation pillaging of each specific area (physics, biology, psychology, sociology, etc.)? This a posteriori appropriation of scientific results would notice the despoiling, if it did not reveal a simple references management. The research of the last theory then proceeds of pedantry, contradictory when common sense is exalted. Unless it could only concern simple journalism (archaically arrogant or moralistic, like in under-developed or one-party countries)  – the ‘information in the philosophical meaning’ is deflated.

Comprehension and Information Processing

Finally, the interest of the information processing idea is precisely the generalization of the intelligence model regardless of its computer or human support. Newell and Simon, quoted by Searle (p. 29), are explicitly considering information processing this way, by clarifying its mechanism as “physical symbol manipulation.” From which the standpoint hostile to AI can be characterized as romantic archaism, because if body negation is another possible archaism, good to mention in AI partisans, it is not strictly included in the approach neutralizing supports rather than denying them. More, the question of information processing doesn’t consist in claiming that the purpose is to give a human comprehension to the machine. No, it is not a falling back! The problem is as follows: i) Either simulating human intelligence, working on observable behaviors, simulated by computer: answers, questions, knowledge already learned, etc. ii) Or generalizing human and computer comprehension in a common information processing.

But of course, AI adversaries refuse, on the taboo mode, that it can be spoken about comprehension for machines, or information processing for human beings:

“However, one can’t help being misled into feeling that if Bobrow uses ‘understands’ rather than ‘processes,’ it must be because his program has something to do with human understanding.” (Dreyfus, p. 135).

When Minsky does identify both terms, he actually generalizes the term ‘understanding’ to machine. If he analyzes this term as an operation consisting in processing information, this operative analysis worth better that a more or less strange morphological or etymological (philosophical or Lacanian) analysis, with puns idiomatic limitations. This generalization consists precisely in isolating what in the processing has to something to see with understanding. It can even be considered, despite the scandal, that this condition is already solved in a simple washing machine, or a thermostat, generalizing the term belief to environment representation:

“McCarthy says even ‘machines as simple as thermostats can be said to have beliefs’. And indeed, according to him, almost any machine capable of problem solving can be said to have beliefs. I admire McCarthy’s courage. I once asked him: ‘What beliefs does your thermostat have?’ And he said: ‘My thermostat has three beliefs it’s too hot in here, it’s too cold in here, and it’s just right in here’.” (Searle, p. 30).

Searle is opposed to the generalization of belief to machine by reproaching here the categorization simplicity. But what about simplistic human categorizations. One thinks immediately at racist categories of former South Africa (where, incidentally, I was told that Japanese were whites and Chinese blacks. Guess why?). In addition, before thermometer invention quantitatively scaling heat, temperature was measured in these qualitative terms. The problem being simply to represent mental states, the thermostat principle is therefore good enough, to what it is made for (technically speaking). What could be a mental state if it was not limited this way? According the principle of simplicity, Searle should have been satisfied by the answer (which hasn’t any comments, what amounts here – I suppose – to reprobation!). But the others’ arguments simplicity is presented as triviality, even indecency or presumptuousness. The process consists in inducing the answer by the devaluing describing like in the question: “Can an artefact think?” (Searle, p. 36), while any thought is an artifact. And the fact that thought is considered as human simply consists in terming a phenomenon, no in analyzing it. Actually, the very scandal rather consists in answering to a philosophical question, whose role is only to make people’s head swimming.

Language Updating

The philosophical standpoint can thus amount most often to the valorization of its discourse by the usage of archaisms. Conversely, the acceptance of human sciences possibility and of AI could amount to concepts updating, and to automation possibility of logic to handle them. Oddly enough, Searle notices himself this use of archaisms, which could be considered as a possibility of AI or human sciences justification, although he is challenging them:

“I am convinced that part of the difficulty is that we persist in talking about a twentieth‑century problem in an outmoded seventeenth‑century vocabulary.” (Searle, p. 14).

Nevertheless, when the generalization of the information processing notion is challenged, it in aid of the old notion of mind. Indeed, materialists, or behaviorist (cf. Searle, p. 14), can admit the phenomenological reaction against the mental negation. But these last ideas have been themselves notoriously designed as simple methodological precautions, in opposition of idealistic driftings. If these reservations are sometimes excessive, they nevertheless allow objectives observations about behaviors. These precautions avoid too the permanent tendency to confusion between real and conceptual, while phenomenology aims the archaic category reestablishment (generally for religious involvement), as Searle recognizes himself again: “But the use of the noun ‘mind’ is dangerously inhabited by the ghosts of old philosophical theories” (Searle, p. 12).

Traditionally, philosophy is situated in the framework of the opposition between “question of words and question of facts” (cf. Jean Largeault concerning Quine). It can be said that it is characterized by the lack of solution to this problem. Even more precisely, the philosophical mental state could be characterized as the pathology corresponding to this individual epistemological situation! Former words (consciousness, intentionality, subjectivity and mental causation), which do are verbal events, are describing subjective phenomena, but not as real as asserted by Searle (p. 17). Inept and labored constructions tending to ground them are showing the psychologization degree chasing its tail. We can attend here to a confusion between a problem statement and its solution, on the Searlian simplicity principle: “I want to go to the movies, and I do go to the movies, normally my desire will cause the very event that it represents, my going to the movies.” (Searle, p. 84) parodying Tarski.

All the work of negation of the updating possibility can, finally, to be read as a metaphor, in the sense that it operates on literary representations of the world. Or more exactly, it concerns a representation of the world with only literary means. These former words actually describe phenomena represent rather classic categories of personality, or novel characters’ psychology. This phenomenon can be explained at least by the simple fact that school training concerning social life doesn’t include sociology, but is grounded on literature (or a more or less mythological history).

This possibility of concept renewal is precisely refused, not only to data processing or AI, but also to cognitive sciences in general, or to theory of systems. The very excess of this refusal of any modification of the human sciences language is expressed then in a classic pathetic way:

“The corruption of language plays an important role. The language of the artificial intelligentsia, of the behavior modifiers, and of the systems engineers is mystifying. People, things, events are ‘programmed’ one speaks of ‘inputs’ and ‘outputs,’ of feedback loops, variables, parameters, processes, and so on, until eventually all contact with concrete situations is abstracted away.” (Weizenbaum, p. 253).

It could be put up against this author that corruption of words is not a mystical standpoint, but that purist and archaistic ideas is. It is necessary that words mean something, but not inevitably what they meant by the past. The modification of a word extension can be, or cannot be, a progress. It has to be judged on its formalization qualities. Beyond words, this author elsewhere refuses realities such a language is expressing. As we already have seen, the term of programming applied to human is obviously replacing the machine man topic. It describes ideas usually expressed by terms like conditioning, habit, routine, etc. The very fact that romantic thought could oppose to these phenomena also indicates this reality is existing for it.

This lack of assimilation and updating of concepts describing current situations can characterize an archaistic pathology. Talking about archaism can be considered as an evolutionist idea, so discredited today, albeit it could not be described itself as (positivistic) archaism without accepting the criterion in question. A reasonable evolutionism simply aims to explain cultural differences in term of competence distribution. Its ethnocentric version was actually itself merely an occident competence stage! But the corollary was that, in evolutionist time, it was obvious that vaccinations, technique, progress… had to be generalized. The concealed return of this universalist idea can elsewhere be noticed, with what have been called right of interference, for humanitarian reasons. The opposite situation can be read in selfishness, or racism, terms. In the current rejection of evolutionism, applied to the medical under-development states, let’s rather notice that insolvent populations are left dying of curable illnesses.

In fact, despite his computer competence and his original observations on many questions, the Weizenbaum’s solution is limited to a conservation of symbolic, religious or psychoanalytical language. Its characteristic is to sum up discourses of protagonists, and traditional symbols. In addition, it is obvious that assimilated cultural norm are human, and foreseeable sociologically, without being good a priori. An Aryan child in Germany of the thirties will be enthusiastically in Nazi youths. And only some restive anomic exceptions won’t follow the herd.

The problem of AI is the traditional philosophy one, currently embodied in social sciences. It consists in making void rationalizations operative – and not to add some news ones. Albeit as the human experience is continuing, myth production will continue. It is necessary to consider too that a few emotional aspects of these romanticist values is rather grounded on the fact they were not realized in the time of their elaboration. It is what can produce the vain wish or tragic boastfulness impression, from déclassé or mundane arrivistes, as those caricatured in Dostoievski’s Possessed – another Weizenbaum’s reference. All the pseudo-philosophy invoking these values while denying them is merely a pre-democratic stage reconstitution, which has been subsequently stabilized in literature or history. Once realized, these democratic values become so trite that some disenchantment can result, and lead to backward-looking attitude, those selected, during their studies, on romanticist references.

The former values or their representation invocation can represent a recourse (as admitted by situationnists, nevertheless radical leftists, in the sixties). But it precisely concerns strategies in an archaistic context, or against usurpers. Church or nationalism can be used to reverse communism in Poland, and in eastern country in general. These values do not represent modern organizer principles once the goal has been reached. And the usage archaic values form can have some consequences difficult to evacuate like Afghan fundamentalists after CIA assistance. Romantico-leftists attacks of adolescents or regionalists are the product of this system too. Regional/nationalists movements are grounded elsewhere on a romantic misinterpretation considering nation in racial or cultural terms. While its foundation is the power of the people against tyrants or oligarchies, and not natives against colonizers or foreigners.

Thus, hermeneutic approach does lie on a concern of syncretic conservation of former notions into news ones [NOTE 63]. This could be understood as the classically Hegelian’s notion of sublation (or Aufhebung: conservation and outdoing), but it actually amounts to the refusal of any updating. It even ends to a disenchantment of the world refusal, by the illusory conservation attempt of the (supposed) initial impression [NOTE 64]. As it is pretty obvious that trivialization of already known is unavoidable, or that the discovery inspiration flash cannot be continuous, the process is grounded on regression. A former idea presence in a new one is then the sign that the new one is not (entirely) established. This is a kind of counteracculturative strategy, i.e. a refusal of adaptation to a new reality or of its assimilation.

Neutrality

All things considered, by its updating, AI does amount to a neutralization of language by a rejection of anthropomorphism (except in its very name then!). That is unbearable for phenomenology (which prefers the contrary). Its neutrality and objectivity refusal reaches an obsessive characteristic, especially in Winograd and Flores. But the limitations of this idea remind us something too: that amounts any intellectual productions to ideology in the Marx’s meaning, possibly through the Sapir-Whorf’s hypothesis. This thesis summed up by heideggero-sartrian philosophy gives an outlet to the Marxist training making start singing this leitmotiv by these authors (Winograd & Flores, pp. 29, 71, 97, 156-157, etc., or in the Lucien Sfez’s preface of their French edition, pp. 6, 9). But, they hurry to neutralize this thesis since Marxism is no longer a current reference. Against science extolling objectivity, the negation of neutrality rather seems the new idée reçue, contrary to what Anthony G. Oettinger asserts (in the Preface of Dreyfus’ French edition, pp. XXI-XXII). The principle of this negation is the total romantico-Heideggerian commitment in the world and in history, forbidding any neutrality (Winograd Flores, pp. 32, 156-157). But commitment is not the human sciences distinctive feature, since the physicists, chemists, biologists are part of their object of research too.

More, the holist reference forgets there is differences between many knowledge level. Only ignorance can be complete. Even analysis of human realities can benefit from neutrality, first by the simple accounting of their existence. The familiarity with our environment, in which we are “thrown,” has itself been built. Why refusing to widen it to other extra-subjective or extra-cultural realities? These data and their teachings are elsewhere already available for a long time. It is true that they have disturbed the image of us as to produce such theories (Rousseauism, Sapir-Whorf’s hypothesis, racism, etc.). This lesson can be remembered to guard against problems that will be met in the modeling process since:

“In order for a program to be biased, it is not necessary for a programmer to consciously set out to insert prejudices [as he or she] must operate within his own background of prejudice.” (Winograd & Flores, p. 156).

But that doesn’t mean no one should program – the inhibition principle is recognized here. On the contrary, programming itself has already an objectifying value: either it requires clarifying rules [NOTE 65], or indirect consequences of these prejudices can be modeled. With the advantage that some unwished side-effects can be avoided, on principle, in both cases.

The axiological neutrality (values and commitment neutrality) is what allows, after a kind of terminological unification, to judge and to estimate innovation. The legitimacy of analogies, and their passage in a universal scientific corpus can be understood this way. Practical examples become scientific, and paradigmatic, when they are neutralized. And first when it is through with the alternative of valuing/devaluing usage, or with the literary pun about them. That restores the scientific value to the common discourse or even symbolist, since the stylistic aspect doesn’t become parasitic. Concerning the human behavior study, sociology is the formal system maintaining the real social dialogue at this level. Conversely, philosophy remains in the moralistic framework, because the adverse principles refusal happens by the rhetorical process of negative connotations. The good/evil opposition establishes a diabolizing rhetorical effect. This point represents the essence of religious explanation (Weizenbaum’s favorite).

Communication and Problem Solving

The information processing principle also leads to insist on the communication importance. Immediately, AI adversaries see here the opportunity of a new rejection. The problem seems to lie in the direct information transmission refusal, for the sake of commitment and interpretation (Winograd & Flores, p. 176, etc.). For Weizenbaum, it even seems there is not any possible communication except the mere reference to ineffable. Paradoxically, he grounds certainly his idea on a too much classic definition of science, which says that terms have to be defined in a regress to infinity. He then specifies that “any understanding of [words like trust, friendship, hope] must always be fundamentally metaphoric” (Weizenbaum, p. 212). That term metaphoric, valued when he is the one using this option, can be clumsy since that can mean the right to say no matter what, without any risk to be at fault ever. Weizenbaum adopts here the reference to ineffable, to uncertainty or to incompleteness, as being unique to human. But he neglects an essential principle of human knowledge, according to which ignorance won’t be left vacant. To manage this cognitive dissonance, the subject in his culture will start the production of a myth, generally archaistic, or metaphorical.

However, it is true that, since the time of the AI opponents’ books publication, this communication term was made fit every occasion. But the tendency using the word communication this ways is precisely the one generalizing the archaic standpoint of human beings manipulation for the ineffable sake. The alternative is classical: choosing between management of men and management of things (according to the Saint-Simonian opposition). It is not surprising that advertising people, whose professional ethos it is, and many former radicals, have this manipulating approach in mind when they exploit the philosophical opening, since it is the essence of their own strategy.

In this standpoint, Flores (most possibly more than Winograd, since that former Chilean minister has worked on cybernetic theories application to organizations problem) produces an quite odd theory grounded on the informal nature of managerial relationships. But his conversational idea seems excessively considering only oral (Winograd & Flores, p. 151), fitting better with his assumption. He seems to consider that an area like financial market is “purely linguistics” (Winograd & Flores, p. 174).

But the fact that oral is reigning at the highest level doesn’t mean that actions/share-stock are informal inside businesses or market places, in which there is also traces in writing. In businesses, the intermediate hierarchy translates the informal in formal operative. Elsewhere, managers complain sometimes of perturbations in passing on orders, equivalents to noise in information theory. This conversational theory pushed further to its limitation seems to rest on the ascendancy of the WINOFLORY method of jargon building, developed in a whole chapter, “Management and conversation” (pp. 143-162), on the topic “A business (like any organization) is constituted as a network of recurrent conversations” (Winograd & Flores, p. 254). This overcoding doesn’t modify the habit of managers, considering generally a business as a quantifiable totality. If actually there IS conversations, even sometimes deficits of negotiations, especially in conflicts resolutions as considered (pp. 168-171), these relationships are nevertheless formalizable. The authors elsewhere acknowledge some standardized practices (pp. 171-172). But formalization is not certainly established a priori in any conversation. This would be rather people prejudices (and not only DP men or computer scientists’ ones like above) which compared with those of others (before standardization). Human beings in real life are often confronted with suppleness or communication lacks. And the computerized modeling could provide a neutral ground allowing a dialogue between protagonists, especially due to ideological closure.

More specifically, AI adversaries consider communication analysis a human mechanizing solution, without seeing the interest of the systems or organization theory specialty for human concerns. This interest is admitted from individuals, but not for their specialty, like in the case of the trouble on campus in which the reality of conflicts is objected to better communication quest (Weizenbaum, p. 251-254). In fact, the Weizenbaum’s standpoint can lie on the over-valuing of the class struggle paradigm, whose verbal mannerisms can be acknowledged, albeit lessened by the American context, because if there are actually real conflicts, it is still necessary to define them, and to communicate them, in the better possible way. And if a good communication can solve some of them, why denying oneself the help of any available technique [NOTE 66]. Elsewhere, this author acknowledges the willingness of communication partisans.

However, where is the real conflict negation when cybernetics, or communication studies, aim to better represent problems to better solve them. From this point of view, democracy at works can be called problem solving – it replaces ‘the earth without evil myth,’ the society without class, or the utopian community of a pure culture or race. The new situation in which it is necessary to manage these disparate values is more prosaic, since words refer to facts and operations. The system analysis, communication, or sciences of organizations refusal, in Weizenbaum, is merely an avatar of this anti-sociological obscurantism.

The alternative to an information processing as personal competence would oblige to admit the existence of a good and evil opposition, as autonomous entities, informing, programming, and possessing individuals. This really is a regression to a theological thought, even to predestination, absolute in Protestants or relative in Jansenist. Witchcraft trials are the result already acknowledged in human history. The representation in term of problem resolution seems infinitely more positive. And why not to admit that problems are equivalent to Heideggerian breakdown, so beloved of Winograd & Flores, dogmatically opposed to this epistemological reconciliation solution (p. 12).

This refusal of formalisms of AI, or their representativeness in psychological term, leads to reject anything produced until today, and more especially the first steps of AI. In the rejection of the Simon and Newell’s GPS program (General Problem Solver) the WEIZY method strikes again: the GPS schema is called a nontheory; then a theory about problem solving, but trivial; then a problem solving pseudo-psychological theory, aiming Newell, Simon, Feigenbaum and Colby for this sacrilege (Weizenbaum, pp. 168-181).

What is observed when to the scandal origin is referred at: GPS works according to the means-end analysis principle, it thus belongs at a post-Cartesian framework of extension and precision of his Method, simply consisting in decomposing a problem into under-problems. The precision consists first in noticing differences between initial state and final state; then available operators are used to reduce differences. In order to avoid combinatorial explosion, heuristics (optimal use of rules about rules) are used, according to the expert systems principle.

Contrary to the trivialization in question, it is necessary to admit that the principle of GPS is a theory (true or false) of what is occurring during human reasoning. But for Weizenbaum, it is out of the question to admit human problem resolution, as materialized in a computer language. Nevertheless, human heuristics (true or false) exist for long, and their computerized operationalization, precisely embodies a sifting of their efficiency and value. This classic challenge of Cartesianism is even attacked in the leftist framework of the Vietnam War refusal. The ironic exaggeration considering any alien theory as ideological can be noticed [NOTE 67]:

“It is in fact entirely reasonable, if ‘reason’ means instrumental reason, to apply American military force, B-52’s, napalm, and all the rest to ‘communist-dominated’ Vietnam (clearly an ‘undesirable object’), as the ‘operator’ to transform it into a ‘desirable object,’ namely, a country serving American interest.” (Weizenbaum, p. 252).

On the contrary, if not blinded by an opposite ideology, anyone can abstract any peculiar circumstances. It is the own of any theory. Analysis in term of operator, problem, undesirable or desirable object, is not dependent of application targets or of reminded or used means (peace, discussion can be desirable objects or operators too). In Newell and Simon’s formalization, refused by Weizenbaum, action analysis is obviously retroactive. It is applicable to past wars, without computers or explicit problem resolution. For Vietnam above, the lack of computers would have most possibly been replaced by operations less accurate, but more massive, like during World War II. Elsewhere, Americans have lost the Vietnam War. It can be said too that it has been won by a better use of communists’ (communication) propaganda operators, and the unequal transparency, to the cost of Americans. But problem solving cannot be accused for controlling results, as GPS specifies too.

Communication and democracy

Political problems are problems of divergent interests structuring. Enforcing legitimacy by negotiation consists first in excluding resort to armed confrontation. Assemblies’ discussion lies in the fact of accepting this scandalous situation of dialogue, instead of erasing the adversary from the surface of the planet. In this framework, we can notice that claims simply amounts to expectation to a largest share of the cake (of the value added, or subsidies). But distribution is quantifiable. Even workers of a self-managed business cannot be paid more than those of capitalistic business. Self-management simply identifies shareholders and wage earners, what can specifically means they need a formal distinction between profit sharing and salary.

The best counterexample of this negotiated processing is the rumor problem, classical origin of pogroms, riots, revolutions, and possibly of a good fraction of students movements reminded by Weizenbaum. Demonstrations can be symptomatic of an irrational, or simply demagogical processing. They represent indeed a case of game with the limitations of confrontation, without assuming any consequences. This can, intrinsically, represent a regression to ritual conflicts, as the warlike term of struggle indicates. They do not solve much more problem too (like medical research), but lead to another grading (these having biggest pressure ability) which can be unfair too.

It is often difficult to solve social problems. But, here too, how can one wish having less knowledge? If it is not inevitably necessary to conclude or to choose automatically, a best problems formalization can be useful, especially if with a vast reference situations library. From this point of view, knowledge simply grows by accumulation, and by established solutions applying. For instance, the “‘system specification languages’ that deal with the domain of computational devices in a more general way” (Winograd & Flores, p. 177) allow to automate the business information system formalization. (Conversational) Decision Making System, currently specialized to the economic area, could be generalized to more sociological situations. This information theory is an attempt of human problem resolution. But then in the current human organizational framework, it can also be noticed that the problem processing work is certainly not flawless. As it can be illustrated by the literary following example:

“As an invalid, I’m entitled to get a flat in the first floor. I’ve addressed a request to that effect to the soviet of the district. And it’ll be soon three years the steps lasts for a pathetic problem that could be solved into three minutes. There was a letter of the combine about it. One has written to the newspaper. In vain. Where is the hitch? Here is a real question for thinkers. The Theoretician solve it simply: our society is a behemoths’ one, and itself, as a whole is a behemoth too. And our problems are behemoth-like. As their solutions are. But this explanation seems to simple to me.” (Alexandr Zinoviev, Living, A Robot Confession, p. 75).

As we can see, the Theoretician [NOTE 68] only solves the problem verbally, since the result doesn’t change (as it can be observed by a comparison between initial state and final state, on a GPS like mode). He has simply produced an adaptive rationalization, moralistic and cynical, like any philosopher – and any member of this socialist society (on the old fatalistic Russian model), institutionalizing irresponsibility. These very problems which are precisely those handled by information (or administration) theory. However, of course, the administrative solution would only be statistical: i.e. a largest number of this kind of problems could have been solved. A specific case could remain, because, it is also necessary to admit material limitations to action (e.g. it is necessary to wait for availability of an apartment, or an organ for transplants, etc., on pain of crime, privileges, or authoritarianism). Conversely, petitioner or humanitarian solutions, largely put on media events today, only solve the concerned cases, precisely leaving similar cases, possibly close ones, just as they are. It is what can be called appeasing one’s conscience lightly. Obviously, besides liberalism allowing choice, in the concerned case of a priority attribution, the simplest would be swapping apartments by mutual agreement with someone in the first floor. Especially as the odds are that apartments are the same. A possible administrative assistance with disabled persons could simply consist in paying transfer charges, or in giving compensation to displaced persons.

Another example – real this time – can be given, showing a conflict between values systems on different levels (judicial, company regulations, common sense): recently, in France, clerks of a State hospital admittance have refused to attend to a person fainting just in front of the hospital door:

i) Common sense speaking, the situation is absurd. It actually derives from an ambulances monopoly, in France, of transporting patient to the hospital. This kind of convention not being known by lay persons, the absurdity leads to a conflicts possibility, or a loss of time when the procedure is respected: the solution could have been the hospital reception had called an ambulance.

ii) Regulation speaking, a finicky application of procedures can be especially observed from the (supposedly human) hospital reception clerks. Their personal problem is of course avoiding problems of regulations transgression. It can also be noticed it seems to be a problem of (space or concepts) boundaries, like for a common computer.

iii) Legally speaking, this Kafkaesque situation is in contradiction with the law, for failure to assist a person in danger. In a hospital case, a public one what is more, the felony is accompanied by aggravating circumstances. The law generality is not abrogated, on principle, by an internal regulation or a private convention. Albeit this is often the case, especially in France, where jurisdictions of exception are all the time created.

In this episode, it can be observed the participation of a human agent doesn’t improve the general judicial formal structures, which happens to be good. On the other hand, it is actually possible that common sense couldn’t been rewarded if it had been applied. The moral fault, and the crime in question, can be considered as a symptom, signaling a personal conflict, which appears as a kind of zeal strike [NOTE 69]. Now, a processing in term of information system would not have had this problem since a priority order (law then regulation) could have been defined. A contradiction between rules can be assimilated to incoherence in the knowledge base. It could have therefore been automatically, and beforehand, pointed out by an artificial system [NOTE 70].

Intuition or method

All of this leads us to the fundamental method Vs intuition question. In epistemological works, it is natural to state (not always explicitly) the normative respect of a method as the exclusive scientific way. This tradition ascends to Francis Bacon (1561-1626), even to Roger Bacon (1214-1294), and go obviously through Descartes (1596-1650), to end with Herbert Simon, and computerization methods. It can be considered, in our standpoint, that this characteristic gives to the most innovative researches an algorithmic aspect showing respect to protocols. In the spiritualist lineage, the philosophical opposition to AI, and more generally to any method, is therefore done for the intuition sake, against rules (Dreyfus, p. 117) or, at best, admitting the association possibility of this intuition to rationality (Weizenbaum, pp. 254-256).

This intuitionist idea is however perfectly contradictory with the stemming idea of the classic philosophy (from Aristotelian’s information to Kantian’s subjectivism), nevertheless residual in phenomenologists. Most possibly it aims inheriting prestige of rigor that a stricter methodology can be claimed further: “all searching, unless directed by a preliminary structuring of the problem, is merely muddling through” (Dreyfus, p. 119). Besides this contradictory reminiscence, intuition is of course grounded here on the holistic assumption. Indeed, an overall intuition would make things easy against the laborious problem solving. But Dreyfus also doubts obviously that intuition could be introduced in computers by opposing the necessity to “understand the problem”[NOTE 71]. He uses for that an amazing reference:

“Polya, who is fully aware of the necessary role that insight plays in problem solving: […] ‘First, we must understand the problem. We have to see clearly what the data are, what conditions are imposed, and what the unknown thing is that we are searching for. Second, we must devise a plan that will guide the solution and connect the data to the unknown.’ [NOTE 72]” (Dreyfus, p. 116).

This apology of intuition is equivalent to nothing else than a misinterpretation that makes you wonder on interpretation capacities. Polya does propose a problem solving method, as the title of his book, How to solve it, indicates. Even without being a DP man, the plain mistake can be perceived, by making a concentration effort against rhetorical effects. Indeed, Polya does break down the different operations of specification (data, constraints, and unknowns) thus representing what Dreyfus calls intuition. The confusion maybe rests on the attention-getter (“to understand the problem”) satisfying the sermonizing holism .

This methodical approach can be precisely substituted to intuition which simply characterizes situations in which the method is not explicit (or has became an automatism). But even the simplest skills, as not explicitly formalized, are not ineffable for all that. And artists – philosophers’ reference intuitive – are more pragmatic. Pamela McCorduck very precisely points out this feature. She also notice that mathematicians as formalisms practitioners share this intuitionist belief:

“It took me a long time to understand that professional artists are not superstitious about creativity; they’re very concerned with it, and most of the artists I’ve talked to don’t think much of the theory of talent, and they admit they learn things by looking at other people’s work and thinking about it and asking them how you do things and so forth. Mathematicians don’t. Mathematicians never talk about how they think about mathematics, and they worship their creativity as a God-given gift. They are hypocritical about teaching students because on the whole they believe that you can’t teach students to be mathematicians – some of them have it or they don’t.” (McCorduck, Machines who think, p. 174).

The inversion goes only in appearance against nature, as it is in the philosophical neoplatonic tradition specific to some mathematicians. Anyway, observation contradicts on both of those aimed by phenomenological speculations.

Information Systems Design

More, the philosophical resistance to any method is deceptive, because problems arising from the lack of method have been observed in the computerization framework itself. The result of the more intuitive first approaches actually was catastrophic when programs have begun to be made of several millions lines of code. Elsewhere, this is one of the origins – real and technique – of the reproach addressed to data processing: “Decisions are made with the aid of, and sometimes entirely by, computers whose programs no one any longer knows explicitly or understands.” (Weizenbaum, p. 236). It became mythical only by confusion between machine and human responsibility. Weizenbaum is right to remind us difficulties generated by too complicated programs, especially in the first phase of data processing, when programming was poorly methodical.

But little by little, methods are developed, from the very difficulties, accumulated experience, and some pioneers’ work. Thus, software engineering shows the emergence of a method from the DP men’s or computer scientists’ practice, what represents a refutation of the Weizenbaum’s theoretical thesis. It happens to be an application of science of organizations to computer programs too. In both cases (data processing and administrative theory), Herbert Simon’s works and discredited specialty find a justification. Elsewhere, errors or experiments do not condemn beginning fields, since they permit to find new solutions. It can therefore be told that the hackers’ lack of method is largely solved today. This opposes to a psychologized and guilt provoking use of this critique in which the programmer in question is considered as compulsive (Weizenbaum, p. 111).

It could be considered too, more generously, that passion brings some satisfactions, as it would be the case in other professional practices. It is also possible to consider obsessive persons as victims (and not as guilty parties) of this lack of method [NOTE 73]. The very large program concerned new situations too, often without former equivalent. For those actually misled  by passion, it could be considered too as a compulsive answer that a conscientious professional applies when it is trapped in an insurmountable difficulty.

Communication, Representation, and Computer Analysis

One of the interests of software engineering methods is precisely to enable communication between the different specialties within a project, or with end users. Complex project development over long enough duration, with a large number of participants (possibly replaced on the way), supposes a common language and some terminological stability, to allow maintenance and ulterior modifications. In research, especially human sciences ones, problems from bad communication, and more generally work difficulties in large scale and long term projects, can be observed too.

In these methods framework, the formal representation of the business organization is grounded on system analysis, traditionally identifying a system and its environment; inputs and outputs describing relations between both; an operating system, an information system, and a monitoring system respectively to transform, analyze, or define a policy, inside the organization. The blueprints defined by the monitoring system lead the computerization project in the framework of a short, middle and long term management. It can be noticed too the current organization to be analyzed in term of means satisfying an aiming defined by the monitoring system. Not to confuse ends and goals, to distinguish strategy and tactics, can be equivalent to assumptions analysis. Pinpointing mistakes, critique, breakdown perception, can identify too the usual steps of basic development scheme, instead of using them controversially. An information system means what a business knows about itself. Dissatisfaction can concern any item of information system: actors’ unrest, insufficient results, new orientation, etc., characterizing the necessity of system adaptation to environment.

Specifically, experience in the computerization area has enabled to distinguish some usual steps. This time progress is called software life cycle. One interest of this approach is to consider a limitation in space and time for a project. Let’s notice incidentally that it can be an answer to the timeless and eternal pure research pretension. The lack of limitation on this matter, accompanied by unavoidable “always more credits,” possibly obviously meaning a secure income without any results checking.

More. Methods allow a dynamic representation of information system in distinguishing current from future model. This idea would profitably be distributed in philosophers’ epistemological thoughts, because current representations of science often aim to disqualify the current model under the pretext that there is a future model. This heresy with epistemological pretension is even spreading in research fields, as to become a bound, ritual, introductory or closing tirade. The purpose is certainly the justification of under way or planned programs. But skepticism deriving from and incarnating in anti-scientific trends can only have for consequence credibility decreasing of science, and therefore research funding shrinkage.

The epistemological interest of the systemic or methodic approach lies in the life cycle analyze: basic development scheme, preliminary review, detailed study, realization, implementation, validation. Planning methods (like PERT-method) have been developed too. They enable to anticipate the optimal organization, in time, finance, equipment and human resources. Such methods allowed saving time in the Apollo program, for instance. The life cycle can be specified taking into account its moments strong:

The feasibility study. This step allows determining the computerization or automation appropriateness. It can be generously considered that Dreyfus reinvents it when he is sarcastic about the spatial conquest of the prehistoric man (see Extrapolation and First Step). These methods of project planning precisely show the difference existing between prospective and gratuitous extrapolations. In their good days, philosophers assert themselves the necessity of this anticipation, in order not to throw oneself into empty space. But their disappointments possibly coming from their intuitive personal practices, have created some resistances, embellished by pseudo-methodological arguments.

Contrary to the Heidegger’s reference (Dreyfus, p. 276), the problem of goals does imply defining specifications. The opposition of human and computer or robotics frameworks can be solved by a progressive needs redefinition. This approach should be facilitated today by current programming methods (called objects oriented) favoring reusing (of subroutines) and properties inheritance (a class transmits its characteristics to its members). These techniques allow being free from strict programming tasks, precisely to focus on answering to needs

The appropriateness study aims decide to perform computerization or not. It can thus enable to give up this lunar project during some millennia – project elsewhere maintained by poets. Traditionally, as we have reminded, the decision rests on the specifications possibility (precisely defining inputs, outputs, and processing). In the beginning of data processing, it was about automating repetitious processes, or those necessitating difficult and long calculations. AI is equivalent to deal with hardly specifiable questions, since resting on blurred data or non-deterministic processes. Besides, it is necessary to consider not as identical problems met in scientific data processing, including a lot of processing and few data, and management, or human problems in general, including a lot of data, and less complex processing.

Gradually, human assisted or automatic specification means are implemented by software engineering tools. These programs assist in problem formalization (generally management ones). It is interesting to notice this automation does concern works previously judged as intellectual. Resistances from intellectuals are therefore simply equivalents to those about manual work automation. It’s then enough to pick in the corporatistic  arguments for artisan products. Indeed, the socio-professional structure imbalance and the unemployment risks, are not matters of indifference, and are recorded with the usual overcoding:

“’Computerization’ in its pejorative sense occurs with devices that were designed without appropriate consideration of the conversational structures they engender (and those that they consequently preclude).” (Winograd & Flores, p. 173).

It can be supposed that’s means computerization suppresses jobs, in leftist Heideggerian technocrats’ jargon! But, resistance to automation from those who automated other jobs belongs to a privileges carrying out, producers side, as much intellectual jobs are elite ones, because if overall effects are noticed (the least you can ask from holistic philosophers), industrial revolution especially had for consequence an increase of living standard, thanks to manufactured products, on the consumers side. It is therefore foreseeable that it could be the same for intellectual services, as it is already the case for commercial ones (travels, banking services, etc.).

Computerization methods do specify that, at short term, it is better to start from the existing facilities analysis. This is equivalent to empiricism, subject to encoding. The philosophers’ introspection can be admitted as starting point. But why to deprive oneself of observation and sociological inquiry on organizations. This general principle aims to distinguish a present state and a future state, to decide the transformation or the adaptation of a system judged defective or obsolete. The modification of an organization in a short-term outlook generally includes few risks. But of course, it is often judged poorly innovative, and can appear catastrophic in middle or long term. It can result some immobilism, like in the famous joke: “The new town hall will be built with stones of the former one, which will remain in charge as long as the new one isn’t finished.” But the creative alternative is also: “To wipe the slate clean! [NOTE 74]” on the utopian philosopher and some artists mode, whose bias and consequences are also well known.

In the long term, a larger theoretical liberty can be allowed – although Keynes has warned us that in the long term, we all would be dead. But risks of this option are often considerable in the short term. It is the entire problem of the so-called pure research. In the projects and methods framework, the long term outlooks very often cause huge wastes of resources, budget overruns, even simple swindling. The bet on new technologies is usually risked, as much they often evolve between the beginning and the end of the project. More often, an unrealistic anticipation is a frequent cause of failure: it is never so good to have to endure initial problems, and to be in advance on the market [NOTE 75]. Albeit some bluffs could succeed to discourage competition, as it was the case for Reagan’s administration stars war investments, having cause/induced the end of Soviet Union, and as it’s frequent in computer business. It is also necessary not to neglect the human factor. Notoriously, planning of new structures imposed by outside has always caused a big resistance from participants. The French sociologist Michel Crozier had studied it in his book La société bloquée [The blocked society]. And it seems this shuffling resistance to change is the preferential behavior in bureaucratic societies.

The mid-term is obviously a compromise to limit risks of both other scenarios (subject to avoid piling up both disadvantages). The insistence on the distinction between existent system and future one can thus be good policy, in AI or in these Methods [NOTE 76]. It is this choice that has been made by methods like MERISE (its name is inspired by the wild cherry, by reference to the transplant necessity in order to transform it into cherry tree), most used in France, largely inspiring the standard (AFNOR 67-101, August 1984) of the area.

The practical principle of preliminary or detailed analysis consists in distinguishing different levels: organizational (and informational possibly), conceptual, and physical:

i) The organizational, or logical, level consists in reasoning in term of event, operation, and result, allowing following the path of objects or information in the firm. The informational level is obviously about the materialization of messages at each step. An event can be external, intern, temporal (for instance an order, a bill, an expiry date), and releases a processing operation (data acquisition, calculation, etc.) inevitably producing a result. It is here that appeared rules of management, that define conditions to unite to realize an operation, a calculation (e.g. Incl. VAT = exc. VAT + VAT); or integrity constraints, also expressing knowledge of the world (e.g. Date of birth < Date of death). It is expressed too on graphs materializing the information circuit in the different workplaces. The interest of this organizational level is to represent exactly the activity process. At this level, the essential point is the operation location in time and space. As the organizational level insists on the situated action, phenomenology could have its fill of (or to be considered as doing organizational analysis without knowing it).

ii) The conceptual level is explicitly built by the suppression of locations and the time of the previous organizational diagram! One of the interests is here the automatable simplicity of the process (it suffices not to represent the different locations and to pack some tasks in more global ones). It would be possible to consider that traditional philosophy is precisely equivalent to conceptual analysis, removing time and practical locations of the model. We so could admit that an aspect of phenomenology is taking into account the protagonists,’ or common sense, categories (because of this process). But besides the lawmaker pretensions or reminiscence, the reintroduction of temporality and locality is equivalent to sociological and not philosophical inquiry.

It is interesting to notice that conceptual level is equivalent to an encoding that can be formulated closely to natural language. This standardization allows then to directly realizing database, thanks to entities-relationships model. For instance, with the analysis for the making of a bibliographical database, the fact: “an author writes a book” (or “a book is written by an author”) can be stated simply by defining a relationship “to write” and two entities “title(s)” and “author(s).” Entities then directly constitute (titles and authors) files, and relationships links corresponding to indexes (linking a title to an author). Relationships can be quantified by what is called cardinalities meaning that a book can have several authors, and that an author can write several books. The interest of data processing versus a traditional library manual file is obvious, while it allows cross-search. It especially avoids having to repeat the same information on several files (authors, topic, titles, etc.). Elsewhere, one of the rules of database implementation is that one information only has to be recorded once.

iii) The physical level finally allows characterizing the organization or computerization devices. Programs are realized only at this stage. Previous levels use a general metalanguage (a particular method jargon) or graphic diagrams, remaining close to natural language (and always able to get closer). The interest of current programming is thus to concentrate on problem analysis, in the user’s terms, if possible. In the analysis last step, translation into intermediate language and into machine language becomes automatic. Today, a database can be built (and of course be used) without knowing any programming.

These methods of real situation or professional activity analysis, however can seem heavy (technocratic or bureaucratic), compared to spontaneity or implicit experience. But it can be considered as recording and formalizing the optimal flow of any study, or as providing a corrective explanation of messy project. More simply, they constitute a guide to face a complex task, and be a security for a costly or harmful enterprise. They allow answering to the beginner’s loneliness – solitude also met in judicial inquiry, with consequences that can be guessed:

“The trainee inquirer is abundantly documented about outlines and limitations that law, worried of individual liberty protection, imposes to his or her action. But inside this framework, he (or she) is, paradoxically, left to himself in leading his research and grounding his demonstration. His instructors, superiors, magistrates themselves give him a trust, indeed flattering, but a few worrisome, as it sometimes come within an inch of disinterest.” (Claude Briançon’s Preface, in Paul Ravier & Jean Montreuil, L’enquête de police judiciaire [Police Investigation], p. 2, my translation).

The trainee inquirer knowing proscribing (negatives) rules reminded by Dreyfus could therefore benefit from prescribing (positive) rules of a method [NOTE 77]. The Ravier and Montreuil’s book allows us understanding in what the inquirer’s steps can be guided by a method, very similar to computerization ones! Anyway, who can pretend that any assistance can’t be useful? These computerization methods are precisely grounded on actors or public’s listening, which is a difference in technocratic habit, and are applicable to the designing of a general or specific problem.

Generalization

Any computerization of a business is equivalent to the automation of its information system or its divisions one. These methods precisely characterize the ability of formalization of the micro-world of a human activity, like an accountancy division, a library management, etc. This is the daily professional reality of the majority of individuals. And the fact this designing specialty can be isolated makes a professional skill of it too, already organized as such in methods departments. The best example, before the term was ever invented, of this expertise elicitation in expert systems matters, is the Taylor’s work. Hired as labor/foreman in a factory, he implemented an inquiry in what were then merely a bunch of artisans. His work especially consisted in measuring goods and men’s moves and times. He could thus define what he called Scientific Management, which has produced production line work. Negative effects of this solution have often concealed some positive effects like increasing productivity [NOTE 78].

Expert systems (ES) are one of the AI applications emphasizing this formalization possibility in quasi natural language. The specific expert system problematic is precisely characterized by the formalization of information system representing a human expertise. The same elements can be recognized under other designations:

• The first step is the expertise elicitation, by interviewing an expert. The knowledge engineer (ES specialist) use the maieutic approach of Socrates (the philosopher discredited by the SNCF) making clarify what is implicit in the daily professional practice.
• Knowledge representation takes place in own terms of the concerned expertise, what guarantees communication and monitoring.

• The expert system works itself in term of deduction and automatic resolution (like GPS).

• In order to counter the combinatory explosion of rules, meta-rules are used, allowing the mobilization of knowledge in a specific order, for instance.

• Expert systems implementation must allow an information coherence control. This means database structuring and keeping to integrity constraints for an easy update.

• Automatic learning, possibly by induction, or more simply by precision requests, allowing automatic or permanent maintenance.

• The ES possibly possess a (quasi) natural language front end, which allows a few ambiguities tolerance.

• One ES specificity can be added, its ability of inference explanation, breaking down the reasoning in order to justify to users the conclusions it has reached.

The experience elicitation practice concretizes, for administrators, DP men, computer scientists, or knowledge engineers for expert systems, what characterizes the sociologists or anthropologists specialty. The computerization of information systems represents therefore a de facto formalization of these social sciences fields. Elsewhere, some specialists in social sciences explicitly work on these topics. Since it is about expertise elicitation, it would be obviously absurd not to begin by using sciences whose specialty was precisely this collection. And that can explain too the human sciences interest for data processing, from its origin. Indeed then, it often concerned simple statistical data processing, what could appear restrictive or frustrating about overall phenomenon interpretations questions. But today, we precisely witness the possibility of extension of these data processing capacities to semantic values.

Traditionally in computing, the semantic networks or systemic models building characterize the necessary work to create databases. AI, from this point of view, is a generalization of this problematic, considering a natural language text (oral or written) as a database. As Umberto Eco has shown, in The Perfect Language Quest, the philosophers’ problem was always about concepts structuring. If phenomenology has given up this goal, it only disqualifies itself. The DP men or computer scientists have, very obviously, met these problems since they are universal. As Dreyfus’ M.I.T. students had noticed these questions were processed in their computer sciences courses [NOTE 79]. They have been solved in information systems implementation, which is always materialized, finally, by a database. Specifically, this information structuring has got the stages described by Eco: Hierarchical databases organize knowledge in tree structures, elements and classes. But it raises problems in case of reorganization or doubloons. Net databases allow numerous and multiple links between existing data. But their maintenance appears difficult too, especially in case of link removal. Relational databases constitute then a more versatile solution by organizing concepts on the entity-relationship model, in which a few characteristics define a notion (with the possibility to add some at any time).

Through the conceptual acknowledgement of these steps, it can be concluded these software-engineering methods constitute, literally, the (management) data processing contribution to human sciences, because they record the formalizable and partially automatable part of human experience. An application to management is only the beginning and obviously the natural human sciences outlet. Indeed, contrary to the anti-mercantile schema of the Christian tradition, merchant relationships constitute the conscious (phenomenological) ways of social quantification [NOTE 80]. Usually, what is blamed to these mercantile practices rather rests on protagonists’ lacks in accounting matter. And precisely, management data processing, thanks to microcomputers, allows sophisticated accounting methods to be generalized to the small economic units, even to personal accounts. The lack of formalization is therefore analyzable retroactively as a simple question of distribution or training costs.

Notes