Unravelling Expressive Delegations: Complexity and Normative Analysis

Authors: Giannis Tyrovolas, Andrei Constantinescu, Edith Elkind

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We consider binary group decision-making under a rich model of liquid democracy: agents submit ranked delegation options, where each option may be a function of multiple agents votes; e.g., I vote yes if a majority of my friends vote yes. Such ballots are unravelled into a profile of direct votes by selecting one entry from each ballot so as not to introduce cyclic dependencies. We study delegation via monotonic Boolean functions, and two unravelling procedures: MINSUM, which minimises the sum of the ranks of the chosen entries, and its egalitarian counterpart, MINMAX. We provide complete computational dichotomies: MINSUM is hard to compute (and approximate) as soon as any nontrivial functions are permitted, and polynomial otherwise; for MINMAX the easiness results extend to arbitrary-arity logical ORs and ANDs taken in isolation, but not beyond.
Researcher Affiliation Academia Giannis Tyrovolas1, Andrei Constantinescu2, Edith Elkind3 1Independent 2ETH Zurich 3University of Oxford
Pseudocode No Fulkerson s algorithm is given in the full version.
Open Source Code No The paper does not include any statement or link indicating that open-source code for the described methodology is provided.
Open Datasets No This is a theoretical paper on computational complexity and algorithms; therefore, no datasets are used or referenced for public availability.
Dataset Splits No This is a theoretical paper and does not involve empirical experiments with data splits for training, validation, or testing.
Hardware Specification No The paper does not mention any specific hardware used for experiments, as it is a theoretical work.
Software Dependencies No The paper does not mention any specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and does not describe any experimental setup details such as hyperparameters or system-level training settings.