Knowledge, Fairness, and Social Constraints
Authors: Haris Aziz, Sylvain Bouveret, Ioannis Caragiannis, Ira Giagkousi, Jérôme Lang
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | In this paper we define and study several such families of relaxations. We introduce the first one by an example. Ann, Bob and Chloe are PC members in the International Conference on Everything and are about to bid for papers to review. Four papers p1, . . . , p4 have been submitted. Ann and Bob both value each of these papers as 10, 6, 6 and 1, respectively. Chloe values them 1, 6, 6, and 10. All three have additive preferences. Each paper must be reviewed by exactly one reviewer. Ideally, one would like to assign papers to reviewers so that none of them would prefer the share of someone else to her own share; this is known as an envyfree allocation. It is easy to see that no such allocation exists here: since Ann and Bob have identical preferences, they are non-envious of each other only if they draw the same value from their assignments: the only ways to achieve this is to assign p2 to one of them and p3 to the other, or no paper to any of them; in any case, they will both envy Chloe. However, this line of reasoning implicitly assumes that each agent knows how the items are allocated. This is not necessarily the case in such a context. If each of the three agents ignores to whom the papers that they do not get are assigned, we can assign the papers in such a way that everyone considers it possible that they do not envy anyone else: assign p1 to Ann, p2 and p3 to Bob and p4 to Chloe. Ann, who does not know how p2, p3, and p4 are assigned, considers it possible that one of Bob and Chloe gets p2 and the other one gets both p3 and p4, in which case she does not envy anyone; such a line of reasoning also applies to Bob and to Chloe. Although this allocation is not envy-free, it is epistemically envy-free.1 |
| Researcher Affiliation | Academia | Haris Aziz Data61, CSIRO and UNSW Sydney, Australia haris.aziz@data61.csiro.com.au Sylvain Bouveret Univ. Grenoble-Alpes, France sylvain.bouveret@imag.fr Ioannis Caragiannis University of Patras, Greece caragian@ceid.upatras.gr Ira Giagkousi University of Patras, Greece giagkousi@ceid.upatras.gr J erˆome Lang CNRS, U. Paris-Dauphine, PSL, France lang@lamsade.dauphine.fr |
| Pseudocode | No | The paper includes formal definitions and theorems, but no explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statement or link regarding the availability of open-source code for the methodology it describes. |
| Open Datasets | No | The paper is theoretical and uses illustrative examples rather than empirical experiments with datasets. Therefore, it does not mention training datasets or their public availability. |
| Dataset Splits | No | The paper is theoretical and does not involve empirical experiments with datasets. Therefore, it does not provide training/test/validation dataset splits. |
| Hardware Specification | No | The paper is theoretical and does not involve empirical experiments. Therefore, it does not provide any details about hardware specifications used for running experiments. |
| Software Dependencies | No | The paper is theoretical and does not involve empirical experiments. Therefore, it does not provide specific software dependencies or version numbers. |
| Experiment Setup | No | The paper is theoretical and does not involve empirical experiments. Therefore, it does not provide details about an experimental setup, hyperparameters, or system-level training settings. |