Multicriteria Classification of R&D Projects in the Public Sector: An Approach with ELECTRE TRI-C and Intuitionistic Fuzzy Logic

The selection and allocation of public funds for Research and Development (R&D) projects in public educational institutions require Multicriteria Decision Aiding methods able to overcome the structural flaws of the simple arithmetic mean, since assigning equal weight to heterogeneous items allows unrestricted additive compensation, in which high scores on formal aspects mask severe feasibility deficiencies, whereas the requirement of crisp scores ignores the hesitation inherent in ad hoc reviewers' judgement. To address this problem in the allocation of 18 undergraduate research scholarships among 28 proposals from a federal institute call, a decision model is proposed based on ELECTRE TRI-C integrated with Triangular Intuitionistic Fuzzy Logic, in which the 10 generic items were replaced by eight institutional strategic criteria whose weights, elicited through the Revised Simos Cards Method (Z = 4), highlighted technical feasibility with the highest valuation (20.94%). Evaluations and reference profiles (b1, b2, b3) were modelled as triangular intuitionistic fuzzy numbers, assuming t2 = t3 in order to capture simultaneously the plausible score scenarios, the degree of certainty (μ) and the hesitation (η), to which the λ-weighted difference index (λ = 0.5) was applied, weighting value and ambiguity by the factor [μ + (1 − η)]/2 and, for benefit criteria, cancelling the upper parameters in the subtraction, so that the lowering of scores stems from the lower margin and the declared degree of certainty. As the procedure contracts the original scale, the indifference (q = 0.5), preference (p = 1.5) and veto (v = 4.0) thresholds were set on the defuzzified scale, with the veto restricted to the social impact, student participation and technical feasibility criteria, which nullified the outranking credibility of P08, P12, P18 and P26, all rejected on feasibility grounds. At the cutting level λ = 0.70, the algorithm directly assigned seven projects to Category C3 — P01, P02, P10, P19, P20, P23 and P24 —, while the 11 remaining places were filled through a managerial choice of internal ordering within Category C2 anchored on the σ(a,b3) degree, covering P05, P06, P11, P13, P14, P15, P16, P21, P25, P27 and P28. Compared with the simple arithmetic mean, the model revealed only moderate agreement (Kendall's τ = 0.497), changing three positions in the funded set. It is acknowledged as a limitation that the index penalises declared hesitation, potentially downgrading a proposal because of the reviewer's low confidence rather than its performance, a caveat that does not offset the gains in rigour, transparency and protection of public funds.