Высшая школа экономики94.4 тыс
Следующее
Опубликовано 22 апреля 2016, 9:29
Honorary lecture by HERVÉ MOULIN (Adam Smith Business School at the University of Glasgow, HSE - St.- Petersburg) "Fair Division in the Internet age"
Moderator: F. Aleskerov (HSE)
The Internet is just starting to provide free websites implementing sophisticated microeconomic answers to (a handful of) concrete allocation problems. The most popular site, SPLIDDIT (spliddit.org/), in the development of which I am involved, applies for instance the Shapley value to split a taxi fare with multiple destinations. To assign credit for a joint venture (e.g., a multi-authored article), it uses the "impartial" division method I invented, where one's ballot about the relative shares of other agents cannot influence one's own share.
I will focus on very recent theoretical advances in the classic problem of fairlydividing valuable objects without the help of a benchmark “market price”. Think of family heirlooms among siblings, shifts among interchangeable workers, seats in overbooked classes to students, or computing resources in peer-to-peer platforms. It is feasible to elicit additive utilities from the participants, but more complex messages are not realistic. It turns out that the allocation maximizing the Nash product is together compellingly fair, easily computed, and provide little incentives to strategic misreport of preferences. It is about to be offered for free on Spliddit.
Surprisingly, this compelling rule to divide goods has no counterpart when we divide bads, or when each agent must receive a fixed total quantity of commodities. More research is needed to provide sound advices in those important cases.
Moderator: F. Aleskerov (HSE)
The Internet is just starting to provide free websites implementing sophisticated microeconomic answers to (a handful of) concrete allocation problems. The most popular site, SPLIDDIT (spliddit.org/), in the development of which I am involved, applies for instance the Shapley value to split a taxi fare with multiple destinations. To assign credit for a joint venture (e.g., a multi-authored article), it uses the "impartial" division method I invented, where one's ballot about the relative shares of other agents cannot influence one's own share.
I will focus on very recent theoretical advances in the classic problem of fairlydividing valuable objects without the help of a benchmark “market price”. Think of family heirlooms among siblings, shifts among interchangeable workers, seats in overbooked classes to students, or computing resources in peer-to-peer platforms. It is feasible to elicit additive utilities from the participants, but more complex messages are not realistic. It turns out that the allocation maximizing the Nash product is together compellingly fair, easily computed, and provide little incentives to strategic misreport of preferences. It is about to be offered for free on Spliddit.
Surprisingly, this compelling rule to divide goods has no counterpart when we divide bads, or when each agent must receive a fixed total quantity of commodities. More research is needed to provide sound advices in those important cases.