Distributed Decisions in Networks: Laboratory Study of Routing Splittable Flow

Abstract

We study network games in which users choose routes in computerized networks susceptible to congestion. In the “unsplittable” condition, route choices are completely unregulated, players are symmetric, each player controls a single unit of flow and chooses a single origin–destination (O–D) path. In the “splittable” condition, which is the main focus of this study, route choices are partly regulated, players are asymmetric, each player controls multiple units of flow and chooses multiple O–D paths to distribute her fleet. In each condition, users choose routes in two types of network: a basic network with three parallel routes and an augmented network with five routes sharing joint links. We construct and subsequently test equilibrium solutions for each combination of condition and network type, and then propose a Markov revision protocol to account for the dynamics of play. In both conditions, route choice behavior approaches equilibrium and the Braess Paradox is clearly manifested.

A network ridesharing experiment with sequential choice of transportation mode

“…Most of the research cited immediately above has documented scenarios where tacit coordination fails. In contrast, our present research, as well as previous experimental research on the market entry game (e.g., Erev and Rapoport, 1998;Seale and Rapoport, 2000) and the Braess paradox (e.g., Gisches and Rapoport, 2012;Morgan, Orzen, and Sefton, 2009;Rapoport et al, 2006Rapoport et al, , 2009Rapoport et al, , 2014) provides evidence for limited success of tacit coordination in small groups.…”

“…In previous experiments on strategic choice of route or transportation mode with either negative (congestion) or positive (cost-sharing) externalities, participants made their choices independently and anonymously under what we call the simultaneous protocol of play (e.g., Gisches and Rapoport, 2012;Liu et al, 2015;Rapoport, Mak and Zwick, 2006;Rapoport, Gisches, and Mak, 2014;Selten et al, 2007). Among these, Liu et al (2015) examined the same choice framework as in the present study.…”

Cost‐sharing in directed networks: Experimental study of equilibrium choice and system dynamics

“…The externalities resulting from the decisions of each user are negative i f individual benefits are a decreasing function of the number of other group members making similar choices. Examples include choice of routes in congestible networks (Cominetti et al., 2006, 2009; Correa and Stier‐Moses, 2011; Rapoport et al., 2009, 2014; Mak et al., 2015), and choice of time of departure in directed traffic networks with multiple bottlenecks (Daniel et al., 2009). In many other settings, the externalities are positive, as when the benefit of a choice of route is an increasing function of the number of other users making the same choice.…”

“…This could happen, for example, when users of the same mode of transportation share the travel cost, as could happen with carpool and shuttle taxi. Previous experimental research on interactive decision behavior in networks (e.g., Daniel et al., 2009; Gisches and Rapoport, 2012; Mak et al., 2015; Morgan et al., 2009; Rapoport et al., 2006, 2009, 2014; Selten et al., 2007) has focused on network games with negative externalities. In contrast, ours is the first experimental study of cost‐sharing in traffic network games with positive externalities.…”

Accessing Online Data: Web‐Crawling and Information‐Scraping Techniques to Automate the Assembly of Research Data

“…; Rapoport et al. ); my commentary is similarly structured to relate to future research in these and complementary domains. Due to the introductory nature of this article, I focus on applied concepts and functional illustrations.…”