Superadditivity

Superadditivity

Superadditivity refers to a phenomenon where the total value or outcome of a system is greater than the sum of its individual parts. This occurs when interactions between the parts are encouraged, resulting in a synergistic effect. In decision-making, superadditivity reflects complementarity among decision criteria, indicating that they work together to enhance the overall outcome.From: Patient Safety [2019], DC optimization for constructing discrete Sugeno integrals and learning nonadditive measures [2020]

Multimodal Analytics for Automated Assessment

View Chapter

Purchase Book

Published in Duanli Yan, André A. Rupp, Peter W. Foltz, Handbook of Automated Scoring, 2020

Sidney K. D’Mello

In contrast, signals A and B are said to be nonredundant when they communicate different information (A → X; B → Y). If the signals are independent then the multimodal combination can result in both X and Y. Alternatively, one signal could dominate the other (X or Y) or could modulate the other by enhancing (XX or YY) or diminishing it (x or y). The most interesting case occurs when a multimodal combination results in an entirely new signal (Z), which cannot be explained by the responses of the unimodal signals (X, Y). This property – referred to as emergence or superadditivity – is closely aligned with the Gestalt principle of totality discussed above in that it maintains that the whole is greater than the sum of the parts.

DC optimization for constructing discrete Sugeno integrals and learning nonadditive measures

View Article

Journal Information

Published in Optimization, 2020

G. Beliakov, M. Gagolewski, S. James

A key feature of capacities is that they are not necessarily additive (for brevity, we call them nonadditive). Hence, they extend the traditional probability measures, enabling one to efficiently represent the variety of cases of interaction among multiple decision criteria [2,6,11,12]. It is generally accepted that an additive capacity reflects independence among the decision criteria, whereas superadditivity (more generally, supermodularity) reflects complementarity and subadditivity (submodularity) reflects redundancy. The Shapley interaction index (see [1,13]), as well as some alternative indices [12,14–16], measures the degree of interaction within a group of criteria.

Explore chapters and articles related to this topic

Multimodal Analytics for Automated Assessment

DC optimization for constructing discrete Sugeno integrals and learning nonadditive measures