1 %%%%%%%%%%%%%%%%%%%%% chapter.tex %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5 % Use this file as a template for your own input.
7 %%%%%%%%%%%%%%%%%%%%%%%% Springer-Verlag %%%%%%%%%%%%%%%%%%%%%%%%%%
8 %\motto{Use the template \emph{chapter.tex} to style the various elements of your chapter content.}
9 \chapter{From the founding situations of the SIA to its formalization}
10 \label{intro} % Always give a unique label
12 % to alter or adjust the chapter heading in the running head
17 Starting from mathematical didactic situations, the implicitative
18 statistical analysis method develops as problems are encountered and
20 Its main objective is to structure data crossing subjects and
21 variables, to extract inductive rules between variables and, based on
22 the contingency of these rules, to explain and therefore forecast in
23 various fields: psychology, sociology, biology, etc.
24 It is for this purpose that the concepts of intensity of implication,
25 class cohesion, implication-inclusion, significance of hierarchical
26 levels, contribution of additional variables, etc., are based.
27 Similarly, the processing of binary variables (e.g., descriptors) is
28 gradually being supplemented by the processing of modal, frequency
29 and, recently, interval and fuzzy variables.
34 Human operative knowledge is mainly composed of two components: that
35 of facts and that of rules between facts or between rules themselves.
36 It is his learning that, through his culture and his personal
37 experiences, allows him to gradually develop these forms of knowledge,
38 despite the regressions, the questioning, the ruptures that arise at
39 the turn of decisive information.
40 However, we know that these dialectically contribute to ensuring a
42 However, the rules are inductively formed in a relatively stable way
43 as soon as the number of successes, in terms of their explanatory or
44 anticipatory quality, reaches a certain level (of confidence) from
45 which they are likely to be implemented.
46 On the other hand, if this (subjective) level is not reached, the
47 individual's economy will make him resist, in the first instance, his
48 abandonment or criticism.
49 Indeed, it is costly to replace the initial rule with another rule
50 when a small number of infirmations appear, since it would have been
51 reinforced by a large number of confirmations.
52 An increase in this number of negative instances, depending on the
53 robustness of the level of confidence in the rule, may lead to its
54 readjustment or even abandonment.
55 Laurent Fleury~\cite{Fleury}, in his thesis, correctly cites the
56 example - which Régis repeats - of the highly admissible rule: "all
58 This very robust rule will not be abandoned when observing a single or
60 Especially since it would not fail to be quickly
63 Thus, contrary to what is legitimate in mathematics, where not all
64 rules (theorem) suffer from exception, where determinism is total,
65 rules in the human sciences, more generally in the so-called "soft"
66 sciences, are acceptable and therefore operative as long as the number
67 of counter-examples remains "bearable" in view of the frequency of
68 situations where they will be positive and effective.
69 The problem in data analysis is then to establish a relatively
70 consensual numerical criterion to define the notion of a level of
71 confidence that can be adjusted to the level of requirement of the
73 The fact that it is based on statistics is not surprising.
74 That it has a property of non-linear resistance to noise (weakness of
75 the first counter-example(s)) may also seem natural, in line with the
76 "economic" meaning mentioned above.
77 That it collapses if counter-examples are repeated also seems to have
78 to guide our choice in the modeling of the desired criterion.
79 This text presents the epistemological choice we have made.
80 As such it is therefore refutable, but the number of situations and
81 applications where it has proved relevant and fruitful leads us to
82 reproduce its genesis here.
84 \section{Introduction}
86 Different theoretical approaches have been adopted to model the
87 extraction and representation of imprecise (or partial) inference
88 rules between binary variables (or attributes or characters)
89 describing a population of individuals (or subjects or objects).
90 But the initial situations and the nature of the data do not change
92 It is a question of discovering non-symmetrical inductive rules to
93 model relationships of the type "if a then almost b".
94 This is, for example, the option of Bayesian networks~\cite{Amarger}
95 or Galois lattices~\cite{Simon}.
96 But more often than not, however, since the correlation and the
97 ${\chi}^2$ test are unsuitable because of their symmetric nature,
98 conditional probability~\cite{Loevinger, Agrawal,Grasn} remains the
99 driving force behind the definition of the association, even when the
100 index of this selected association is multivariate~\cite{Bernard}.
104 Moreover, to our knowledge, on the one hand, most often the different
105 and interesting developments focus on proposals for a partial
106 implication index for binary data~\cite{Lermana} or \cite{Lallich}, on
107 the other hand, this notion is not extended to other types of
108 variables, to extraction and representation according to a rule graph
109 or a hierarchy of meta-rules; structures aiming at access to the
110 meaning of a whole not reduced to the sum of its parts \footnote{ICI }, i.e. operating as a complex non-linear system. For example, it is well known, through usage, that the meaning of a sentence does not completely depend on the meaning of each of the words in it (see the previous chapter, point 4).