Chapters in part 3
Click a chapter to open its detailed mind map.
Object, scope and precautions of multi-criteria comparison
Mind map of chapter 18: transition from Part II's grammar to comparison criteria C_1…C_n, distinction between preferential question (MCDA) and descriptive (multidimensional data analysis), four-step method (criteria, matrix E(L,C_i), comparison without weights with Pareto, weighting + sensitivity) and precautions of use.
Building the analytical material: from the grammar of the common to criteria and reference matrices
Mind map of chapter 19: transformation of the book's grammar into usable material — battery of 10 criteria directly anchored to the variables (C_1=Dr, … C_10=J_p), two simulated plausible matrices M_d and M_p (20×10 each), construction via intervals, orientation/normalisation, and a frozen sandbox for the rest of the part.
The evaluation matrix: from the input matrix to the regime synthesis
Mind map of chapter 20: transition from matrices M_d and M_p to the synthesis matrix E(L, C_i), four layers of evidence (structural, observed, expert, perceived), reference rule via discrete upper median, context X (scenarios), multilevel aggregation and reading disagreement through box plots.
Analysing criterion by criterion: elementary preferences
Mind map of chapter 21: mono-criterion reading of the material — score gaps Δ_i = E(L_p, C_i) − E(L_d, C_i), sets 𝒢_p / 𝒢_d / 𝒞, dispersion and box plots on matrices M_d and M_p, smoothed profiles on four salient criteria, variants (indifference threshold, histograms, local weighting).
Comparing complete profiles: dominance, incomparability, weighting and sensitivity analysis
Mind map of chapter 22: from local to global — dominance/incomparability diagnosis (Pareto), check on reference matrix (incomparable: 9 criteria → L_p, 1 → L_d), introduction of weighted sum U_w(L), three illustrative profiles (A balanced, B C_1-priority, C relational), sensitivity analysis, non-compensatory admissibility thresholds, and five methods to produce weights (uniform, 100-point budget, swing, pairs, Borda).
Simulating scenarios: contexts X_1..X_4, score uncertainty and expert disagreement
Mind map of chapter 23: three complementary dimensions (context X dependence, score uncertainty, evaluator plurality), four scenario-types (X_1 strong segregation, X_2 high trust, X_3 media polarisation, X_4 repeated shocks), uncertainty representations (intervals, random variables), multi-experts E_k and multiple weights w^(k), Monte Carlo algorithm and complete example on the reference matrix.
Factor-analysis methods: re-reading regimes in the data space
Mind map of chapter 24: from the preferential side (ch. 20-23) to the descriptive side. Row/column duality scheme, PCA (axis 1 captures 84.87% of inertia, plane 1-2 = 96.36%, R/T/F↓/Tr/S/V/C cluster aligned, Dr opposed), MCA (relative profiles and χ² distance, barycentric principle), gains and limits of factor analysis.
Fisher's linear discriminant analysis: finding the axis that best separates the regimes
Mind map of chapter 25: shift to SUPERVISED analysis. Centres of gravity G_d and G_p, Fisher criterion J(v) = v'Bv / v'Wv, solution v* ∝ W^-1(G_p − G_d), results on the table (I_B/I_T ≈ 73%, LD1 coefs dominated by S, K↓, C, F↓, projected centres ±6.91), forward selection via Wilks' lambda and opening to CART trees.
Automatic clustering: searching for latent structures
Mind map of chapter 26: unsupervised clustering. Three typical geometries (absence, convex clusters, disc/ring), k-means on the 40 rows of M (perfect L_d/L_p agreement with overlap 𝒪 = 0 and inertia ratio I_B/I_T ≈ 73%), centroid reading, column clustering into two families (A: R, T, F↓, Tr, S, V, C cleaving; B: Dr, K↓, J_p transverse) and acknowledged limits.
Bayesian modelling
Mind map of chapter 27: cybernetic model L = (κ, ρ, η) + levers (J, Seg) + intermediate node ξ; Bayesian network on audit vector (R, T, C, J_p, Tr, F, D, K, Vs, S) with 5 simplifying hypotheses; three uses (prediction, diagnosis, updating); complete example in 7 steps (variables, structure, CPTs, Diamond/Plasma signatures, prospective simulation, diagnostic inference, regulation loop); indicators A_coh, A_ten, ΔH = A_coh − A_ten and gauge ΔH = 1 − H/log₂3.
From simulator to shared observatory
Mind map of chapter 28: presentation of the LaïciScope platform that embodies the book's method. Principles (transparency, traceability, non-decisional), GDPR anonymity with localStorage storage, four-domain workspace (Conversation, Causal Atlas, MCDA, Bayes), Lf/La testbed of 160 rows, 6-step / 8-panel tour, exportable objects (JSON project, calculation proof), epistemic hygiene rules, and six-project roadmap.
Conclusion of Part III: making trade-offs explicit, testing robustness
Mind map of chapter 29: Part III summary. Object (make heterogeneous evaluations comparable + equip trade-offs), six bricks (criteria grid C_i, matrix E(L,C_i|X), no-weight comparison, aggregation + sensitivity, multi-scenario simulation, LaïciScope observatory), methodological result (comparison makes disagreements more explicit, without removing them), three confusions to avoid, and opening on the shift from criteria C_i to documented indicators I_{i,j}.