Volunteering vs. Work for Pay

上传人:yanm****eng 文档编号:594882 上传时间:2017-04-09 格式:PDF 页数:9 大小:775.80KB
返回 下载 相关 举报
Volunteering vs. Work for Pay_第1页
第1页 / 共9页
Volunteering vs. Work for Pay_第2页
第2页 / 共9页
Volunteering vs. Work for Pay_第3页
第3页 / 共9页
Volunteering vs. Work for Pay_第4页
第4页 / 共9页
Volunteering vs. Work for Pay_第5页
第5页 / 共9页
点击查看更多>>
资源描述

《Volunteering vs. Work for Pay》由会员分享,可在线阅读,更多相关《Volunteering vs. Work for Pay(9页珍藏版)》请在金锄头文库上搜索。

1、Volunteering vs. Work for Pay: Incentives and Tradeoffs in CrowdsourcingAndrew MaoHarvard Universitymaoseas.harvard.eduEce KamarMicrosoft RYiling ChenHarvard Universityyilingeecs.harvard.eduEric HorvitzMicrosoft RMegan E. SchwambASIAAYale Center for Astronomy & Astrophysicsmschwambasiaa.sinica.edu.t

2、wChris J. LintottUniversity of OxfordAdler Planetariumcjlastro.ox.ac.ukArfon M. SmithAdler Planetariumarfonzooniverse.orgAbstractPaid and volunteer crowd work have emerged as a means forharnessing human intelligence for performing diverse tasks.However, little is known about the relative performance

3、 ofvolunteer versus paid crowd work, and how financial incen-tives influence the quality and efficiency of output. We studythe performance of volunteers as well as workers paid withdifferent monetary schemes on a difficult real-world crowd-sourcing task. We observe that performance by unpaid andpaid

4、 workers can be compared in carefully designed tasks,that financial incentives can be used to trade quality for speed,and that the compensation system on Amazon Mechani-cal Turk creates particular indirect incentives for workers.Our methodology and results have implications for the idealchoice of fi

5、nancial incentives and motivates further study onhow monetary incentives influence worker behavior in crowd-sourcing.Over the last decade, crowdsourcing has emerged as an ef-ficient way to harness human intelligence for solving a widerange of tasks. While some crowdsourcing is unstructuredand organi

6、c, such as efforts to coalesce knowledge on top-ics in Wikipedia and software applications created by opensource projects, several crowdsourcing systems provide astructured environment that connects participants or work-ers with microtasks that are well-defined and self-contained.These systems typic

7、ally do not require workers to be expertsor to have strong familiarity with a task before starting tocontribute. We shall use the terms crowdsourcing and crowdwork interchangeably to refer to work done in these types ofsystems.In paid crowd work, workers are compensated for com-pleting tasks created

8、 by requesters in a marketplace or otherassignment mechanism. Online marketplaces for specifyingtasks and recruiting crowd workers include Amazon Me-chanical Turk (MTurk), oDesk, and CrowdFlower. MTurkhosts a large variety of tasks, including data verification,language translation, and audio transcr

9、iption. Other tasksinclude studies of human computation techniques and be-havioral experiments (Ipeirotis 2010). Workers performingtasks through MTurk are often aware of their compensationand self-organize to find the best-paying and most interest-ing tasks (Chandler, Mueller, and Paolacci 2013).Cop

10、yright c 2013, Association for the Advancement of ArtificialIntelligence (www.aaai.org). All rights reserved.Although crowdsourcing in the absence of monetary in-centives has appeared in many forms, volunteer-based crowdwork has recently expanded to organized platforms. Per-haps one of the most well

11、 known of these is the Zooni-verse,1 an online citizen science platform that connects sci-entists seeking human eyes on large amounts of data withparticipants (workers) interested in contributing to science(often called citizen scientists), and has been successful inproducing valuable data for resea

12、rch. Examples of Zooni-verse projects include Galaxy Zoo (Lintott et al. 2008),where galaxies are classified according to their shapes, andPlanet Hunters2 (Fischer et al. 2012; Schwamb et al. 2012;Lintott et al. 2013), where participants identify potential sig-nals of planets orbiting distant stars.

13、 Citizen science systemsrely solely on voluntary contributions of amateur partici-pants without providing any monetary compensation, andvolunteers run the gamut from a core community with strongintrinsic motivation (e.g. interest in a scientific discipline) tocasual participants who visit the site o

14、nce and leave (Rad-dick et al. 2013). Volunteers in unpaid crowdsourcing sys-tems are driven by different motivations than workers of paidcrowdsourcing platforms; volunteer crowd workers seek dif-ferent objectives, and some may be more knowledgeableabout a specific task than most workers in paid sys

15、tems.The many differences in motivation and incentives be-tween paid and unpaid crowd work are not yet well under-stood, and a primary question is to characterize how differ-ent types of financial incentives influence the behavior ofpaid workers relative to volunteers. These differences are es-pecia

16、lly interesting with regard to the influence of incentiveson the performing of tasks that are ambiguous or difficult, asdifferent financial incentives may influence the amount oftime workers spend on tasks and the quality of work per-formed. If workers are motivated solely by monetary com-pensation on a platform with no quality control, economictheory predicts that they will shirk and produce work of min-imally acceptable quality. For

展开阅读全文
相关资源
相关搜索

当前位置:首页 > 学术论文 > 其它学术论文

电脑版 |金锄头文库版权所有
经营许可证:蜀ICP备13022795号 | 川公网安备 51140202000112号