Computer Science > Computer Science and Game Theory
[Submitted on 10 May 2022 (v1), last revised 16 Sep 2023 (this version, v5)]
Title:Optimal grading contests
View PDFAbstract:We study the design of grading contests between agents with private information about their abilities under the assumption that the value of a grade is determined by the information it reveals about the agent's productivity. Towards the goal of identifying the effort-maximizing grading contest, we study the effect of increasing prizes and increasing competition on effort and find that the effects depend qualitatively on the distribution of abilities in the population. Consequently, while the optimal grading contest always uniquely identifies the best performing agent, it may want to pool or separate the remaining agents depending upon the distribution. We identify sufficient conditions under which a rank-revealing grading contest, a leaderboard-with-cutoff type grading contest, and a coarse grading contest with at most three grades are optimal. In the process, we also identify distributions under which there is a monotonic relationship between the informativeness of a grading scheme and the effort induced by it.
Submission history
From: Sumit Goel [view email][v1] Tue, 10 May 2022 23:11:09 UTC (483 KB)
[v2] Tue, 5 Jul 2022 13:22:06 UTC (23 KB)
[v3] Tue, 3 Jan 2023 17:40:56 UTC (106 KB)
[v4] Sat, 20 May 2023 02:30:44 UTC (143 KB)
[v5] Sat, 16 Sep 2023 05:19:12 UTC (84 KB)
Current browse context:
cs.GT
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.