Web's Largest listing of crowdsourcing and crowdfunding events

Events

Advertise

Web's Largest Directory of Sites

2,412 crowdsourcing and crowdfunding sites

Working the Crowd: Employment and Labor Law in the Crowdsourcing Industry

document Distributed Knowledge, Open Innovation

Alek Felstiner – Working the Crowd DRAFT-5.16.10

WORKING THE CROWD: EMPLOYMENT AND LABOR LAW IN THE CROWDSOURCING INDUSTRY
Alek Felstiner*

Labor markets, like almost every aspect of our economy and culture, have begun an inexorable migration into cyberspace. As budget-conscious employers embrace Internet technology to access larger labor pools, the traditional concept of a fixed workforce comprised of individually selected employees has begun to disintegrate. Stable workforces are being replaced by networked “crowds.” Wired reporter Jeff Howe introduced the term crowdsourcing1 to describe this relatively new phenomenon, made possible by sophisticated software advances, fast and cheap bandwidth penetration, and increased access to personal computing technology.2 The term “crowdsourcing” has since expanded to encompass a variety of practices, but Howe defines it as “the act of taking a job traditionally performed by a designated agent (usually an employee) and outsourcing it to an undefined, generally large group of people in the form of an open call.”3 Through crowdsourcing, new platforms for online work allow firms to connect with enormous numbers of prospective laborers and to distribute burdensome tasks to an amorphous collection of individuals, all sitting in front of computer screens. Though crowdsourcing has been called “the biggest paradigm shift in innovation since the Industrial Revolution,”4 the already-maturing market for crowd labor remains almost entirely unregulated. Or, to be more accurate, courts and regulatory agencies have yet to apply existing employment and labor laws, and legislatures have taken no action to extend those laws or to otherwise regulate crowdsourcing. Such hesitation

J.D. 2011, (UC Berkeley). Jeff Howe, The Rise of Crowdsourcing, WIRED, June 2006, at 176, 179. Though the term is new, the practice of crowdsourcing is hundreds of years old. For example, the first edition of the Oxford English Dictionary was compiled through an open call for volunteers to read books and submit “quotation slips.” See SIMON WINCHESTER, THE MEANING OF EVERYTHING: THE STORY OF THE OXFORD ENGLISH DICTIONARY 53-57 (2003). 2 See Jonathan Zittrain, Ubiquitous Human Computing 1-2 (Univ. of Oxford Legal Research Paper Series, Paper No. 32, 2008), available at http://ssrn.com/abstract=1140445 (“Networks connect people as well as devices, and when they are cheap and easy to use it means that those intellectual tasks more efficiently performed elsewhere by other people can be broken out and distributed…Cheap networks mean that nearly any mental task can become unbundled, no matter how minor”). Jeff Howe traces the roots of crowdsourcing somewhat differently, citing “a renaissance of amateurism, the emergence of the open source software movement, the increasing availability of the tools of production, and…the rise of vibrant online communities organized according to people’s interests.” JEFF HOWE, CROWDSOURCING 17-18 (2d prtg. 2009). 3 Jeff Howe, Crowdsourcing: A Definition (June 2, 2006), http://www.crowdsourcing.com/cs/2006/06/crowdsourcing_a.html. 4 Wendy Kaufman, Crowdsourcing Turns Business On Its Head, (NPR radio broadcast Aug. 20, 2008), available at http://www.npr.org/templates/story/story.php?storyId=93495217.
* 1

1

Electronic copy available at: http://ssrn.com/abstract=1593853

Alek Felstiner – Working the Crowd DRAFT-5.16.10

should not surprise us, given the law’s generally slow reaction time and the likelihood that regulators have a limited awareness of the crowdsourcing industry. But reluctance to regulate may also stem from the unique and daunting legal problems created by the crowdsourcing labor model. Crowd labor has no physical job site. It is performed and compensated entirely in cyberspace, often anonymously, and governed—to the extent that it is governed at all—by compulsory “clickwrap”5 participation agreements. Crowd work is characterized by many-to-many connections between workers and employers, with some connections lasting as little as a minute or two. What constitutes an employment relationship in such an environment? Can a worker genuinely operate as an independent contractor? What responsibilities, if any, attach to the companies that develop, market, and run crowdsourcing platforms? This Article confronts some of the thorny questions that arise in applying traditional employment and labor law precepts to crowdsourcing, and offers some provisional solutions. To that end, the Article makes two related arguments for regulatory intervention. First, the nature of the certain crowdsourcing models makes workers uniquely vulnerable and prevents them from achieving decent wages and working conditions. Second, current employment and labor laws already fail to adequately address the modern physical workplace, and without immediate intervention, those laws will become completely irrelevant and anachronistic as more and more work shifts into cyberspace. Part I describes how crowdsourcing works, and explains why particular subsections of the paid crowdsourcing industry expose employees to substandard working conditions without much recourse to the law. It also examines crowdsourcing’s unique advantages and inherent risks, and explores the impact on existing industries. Part II offers a case study of the Mechanical Turk platform developed by Amazon.com, detailing the demographics and motivations of the workers, the compensation they receive, and the terms that purport to govern the employment relationship. Part III investigates the legal status of the “crowd,” analyzing the threshold question of employment status as well as the complications involved in attempting to apply existing work laws to online labor markets. It also evaluates the possibility of gathering in Amazon and other
“Clickwrap” refers to terms and conditions imposed on users of a software platform, usually through a pop-up window or check box on their computer screen. Users are obligated to agree to the terms and conditions before they can utilize the software.
5

2

Electronic copy available at: http://ssrn.com/abstract=1593853

Alek Felstiner – Working the Crowd DRAFT-5.16.10

crowdsourcing vendors as “joint employers.” Part IV makes the case for regulatory intervention, presenting recommendations for legislatures seeking to expand legal protections for crowdsourced employees, suggestions for how courts and administrative agencies can pursue the same objective within our existing legal framework, voluntary “best practices” for firms and venues involved in crowdsourcing, and examples of how crowd workers might begin to effectively organize and advocate on their own behalf.

PART I. CROWDSOURCING Almost every internet user has probably participated in a crowdsourcing project of some kind. Due to the diversity of existing crowdsourcing operations, and the extent to which they permeate the online environment, many of us as internet users engage with some form of crowdsourcing every day. This Article focuses on crowdsourcing models in which the parties explicitly arrange for compensated labor, in the form of piece rate or wages. In doing so it ignores other models, such as disguised crowdsouring, 6 contest-based crowdsourcing,7 and expert networks.8 It also does not address the tricky and largely theoretical question of

This may be the most common form of crowdsourcing, taking place behind a veil of software, usually unbeknownst to the user. In the process of accessing websites, playing online games, or participating in e-commerce, internet users perform small tasks inserted into the flow of activity. Online security protocols provide an excellent example. When users wish to log on to a secure system, they are asked to read a distorted word or series of words and type them into the response field. Ostensibly, this process is designed to ensure that only real, live users—as opposed to automated software “robots”—can access the system. Most users likely do not know that they are actually using their unique human powers of pattern recognition to read words that a computerized optical scanner cannot recognize. They are, in fact, seeing individual words from old, scanned issues of the New York Times, or other physical documents in the process of being digitized. reCAPTCHA, the company that developed this program, offers web security in exchange for the eyes and brains of the system’s users. See reCaptcha.com, What is reCAPTCHA?, http://recaptcha.net/learnmore.html (last visited May 16, 2010). Instead of “disguised,” Jonathan Zittrain employs the alternate term “epiphenomenal,” in reference to tasks “gleaned as a by-product of people’s activities rather than because they aim to perform them.” Zittrain, supra note 2, at 5-6. 7 Contest-based crowdsourcing (or competitive crowdsourcing) functions as an open competition, with firms broadcasting a problem or complex task to the crowd in the understanding that many crowd members may accept the challenge and perform the work but only one (or a small group) will receive the reward. Innocentive, founded in 2001, offers a worldwide “Open Innovation Marketplace,” where “seekers” can submit “challenges” to a pool of over 180,000 “problem solvers.” Seekers then select their preferred response and pay out the reward. See Innocentive.com, FAQ, http://www.innocentive.com/crowd-sourcing-news/faq (last visited May 16, 2010). Another popular example of competitive crowdsourcing is Threadless.com, an online T-Shirt company that solicits T-shirt designs and selects a new batch to print each week, rewarding the designer with a cash prize and bonuses for each reprint. See Mark Weingarten, “Project Runway” for the T-Shirt Crowd, BUSINESS 2.0 MAGAZINE, June 18, 2007, http://money.cnn.com/magazines/business2/business2_archive/2007/06/01/100050978/index.htm. Similar to competitive crowdsourcing, an increasing number of companies operate what Jeff Howe calls “idea jams,” in which the firm asks the crowd to submit ideas for new products and services, or changes to existing products and services. Howe, supra note 2, at 134. Dell’s Ideastorm project (www.ideastorm.com) offers a typical example, in which users submit and vote on suggestions, and Dell implements them at its discretion, without compensation. Also, in the last few years, a handful of companies have abandoned traditional advertising campaigns and turned to competitive crowdsourcing for promotional materials. Companies typically solicit the crowd to produce a short advertisement on a particular topic, with some general guidelines, and then reward the chosen submitter with cash or other benefits, including the use of the ad in the company’s campaign. See Howe, supra note 2, at xxi-xxiii. 8 Instead of hiring in-house expertise or making long-term consulting contracts, firms can access an online network of experts in almost any field, and farm out complex technological or business questions. See About Us – Gerson Lehrman Group,
6

3

Alek Felstiner – Working the Crowd DRAFT-5.16.10

whether certain types of voluntary crowd labor, such as gaming and social networking, might actually constitute compensable work.9 Instead, it concentrates on the segment of the crowdsourcing industry in which workers perform what can best be described as “cognitive piecework”—sets of tasks, performed and compensated within the online platform.10 On some platforms, the tasks require low to moderate skill and can be performed in a comparatively short period of time. Others require more qualifications and expertise. Cognitive piecework models tend to follow a common tripartite structure, consisting of vendors, firms (also referred to herein as “companies” or “employers” depending on context), and workers. Crowdsourcing vendors develop a “platform” upon which firms can broadcast their tasks, and workers can accept, perform and/or submit the work.11 This platform may take the form of a task list, or may be more actively mediated or automated by the vendor. As a condition of access to the platform, workers and firms must assent to some kind of participation agreement, invariably written by the vendor. These agreements often bind participants to other terms of use separate from those governing the platform, including privacy policies and conduct requirements. Firms then post their tasks to the platform for acceptance by the crowd of workers, or have their tasks automatically funneled to workers by the vendor.

http://www.glgroup.com/about.html (last visited May 16, 2010); See also Gurustorms, How Does Gurustorms Work?, http://www.gurustorms.com/brainstorm/how_it_works_home_2 (last visited May 16, 2010). 9 In many online environments, the distinction between work and passive (or leisure) activity becomes difficult to draw. For a discussion of the intricacies involved in separating work from leisure in virtual environments, see Miriam Cherry, Working for (Virtually) Minimum Wage: Applying the Fair Labor Standards Act in Cyberspace, 60 ALA. L. REV. 1077, 1097-1105 (2009). This Article does not discuss crowdsourcing endeavors undertaken by government, public interest groups, and charitable organizations. However, these projects do demonstrate the phenomenal power of harnessing a networked pool of committed participants. For example, NASA’s Clickworkers project used volunteers to search through massive sets of Mars photographs and identify topographical formations, a task that otherwise would have taken months to complete. See Michael Szpir, Clickworkers on Mars, AMERICAN SCIENTIST, May-June 2002, at 1. AMT providers donated their time to search satellite imagery of the Nevada desert for evidence of missing aviator Steve Fossett. See The Search for Steve Fossett: Turk and Rescue, THE ECONOMIST, Sep. 22, 2007, at 61. Members of Moveon.org combed through and vetted 9,000 advertising photo submissions in 30 minutes. The Extraordinaries Blog, Moveon.Org and Crowdsourcing: A Chat with Peter Koechely (Sep. 29, 2009), http://www.theextraordinaries.org/2009/09/moveonorg-and-crowdsourcing-a-chat-with-peter-koechley.html. The Katrina PeopleFinder Project attempted to tackle the overwhelming task of compiling all data about displaced Hurricane Katrina victims into a searchable, single-format database. Emily Gertz, KatrinaWiki, Katrina PeopleFinder: Distributed Computing Responses to Disaster (Sep. 5, 2005), http://www.worldchanging.com/archives/003437.html. More recent crises have spawned a new generation of crowdsourcing efforts dedicated to efficiently compiling and sifting information during humanitarian crises. See Crisis Commons, http://wiki.crisiscommons.org/wiki/Main_Page (last visited May 16, 2010); Ushahidi, http://www.ushahidi.com/about (last visited May 16, 2010). Perhaps the most striking example of non-commercial crowdsourcing is BlueServo, which in 2008 partnered with the Texas Border Sherriff’s Coalition to set up a network of cameras and sensors along the Texas-Mexico border, creating a “virtual fence.” Users monitor real-time streaming footage of border areas, looking for “suspicious criminal activity.” See BlueServo.net, About Us, http://www.blueservo.net/about.php (last visited May 16, 2010). 10 The term “cognitive piecework” is borrowed from social informatics researcher Lilly Irani. See Lilly Irani, Tweaking Technocapitalism: Turkopticon (Jan. 30, 2009), http://www.differenceengines.com/?p=146. 11 This Article uses the terms “crowdsourcing platform” and “crowdsourcing venue” to refer to the actual cyberspace location (such as www.mturk.com) where firms and crowd workers connect, and where work is often performed, tracked, and compensated. Loosely, the “platform” refers more to the actual cyberspace location, whereas “venue” indicates the legal or conceptual location.

4

Alek Felstiner – Working the Crowd DRAFT-5.16.10

Amazon’s Mechanical Turk platform (“AMT”) exemplifies the cognitive piecework model of crowdsourcing. Firms register on AMT to access an immense pool of workers (called “Providers”), estimated at 200,000 in total.12 The firms (called “Requesters”) post “Human Intelligence Tasks” (or “HITs”), which typically involve basic computing and language skills – such as tagging photos according to their contents, rewriting sections of prose, transcribing audio, choosing representative screenshots from a short video clip, responding to survey questions, or performing internet research. Anywhere from 20,000 to 100,000 HITs are available at one time, and Requesters post 20,000 to 40,000 new HITs every day.13 AMT, which is described in Part II, has more or less cornered the market on the most brief and “unskilled” tasks. But other crowdsourcing companies have adopted a similar pattern with slightly larger units of work.14 For example, Rent-A-Coder provides a ready labor pool of software coders to firms seeking discrete and fairly straightforward coding projects.15 LiveOps uses a networked crowd of communication workers to create virtual call centers for tech support and direct marketing.16 The oDesk and Elance crowdsourcing platforms offer a wide array of professional services, including administrative support, design, engineering, writing, and web development.17 OnForce maintains a network of information technology professionals who can be dispatched to needy companies, on demand.18 LiveWork runs the gamut, from customer support to medical transcriptions, by combining individual providers and business product outsourcers.19 The most important thing to understand about crowdsourcing vendors is that they are more than just glorified job listing services. In various ways, they play an active and fundamental role in establishing the

BRENT FREI, SMARTSHEET.COM, PAID CROWDSOURCING: CURRENT STATE AND PROGRESS TOWARD MAINSTREAM BUSINESS USE 6 (2009), http://www.smartsheet.com/paid-crowdsourcing-current-state-and-progress. 13 Panos Ipeirotis, Mechanical Turk Monitor, http://hyperion.stern.nyu.edu/mturk/; http://hyperion.stern.nyu.edu/mturk/arrivals.php (last visited May 16, 2010). 14 As the tasks grow in skill and duration, they may or may not continue to be compensated on a strictly “piecework” basis. This article continues to use the term “cognitive piecework” to refer to the discrete and sequential nature of the work performed – not specifically to the method of compensation. 15 See RentACoder.com, Common Questions, http://www.rentacoder.com/RentACoder/DotNet/misc/About/CommonQuestions.aspx (last visited May 16, 2010). 16 See LiveOps.com, Call Center Outsourcing with At-Home Agents from LiveOps, http://www.liveops.com/on-demandoutsourcing/call-types.html (last visited May 16, 2010). 17 See oDesk.com, FAQs, http://www.odesk.com/w/faqs (last visited May 16, 2010); Elance.com, Who Uses Elance?, http://www.elance.com/p/who_uses_elance/index.html (last visited May 16, 2010). 18 See Onforce.com, FAQs, http://www.onforce.com/page/faqs (last visited May 16, 2010). 19 See LiveWork.com, Hiring an On-Demand Workforce, http://pages.livework.com/client.html (last visited May 16, 2010).
12

5

Alek Felstiner – Working the Crowd DRAFT-5.16.10

market conditions for crowd labor. Though vendors serve different purposes and occupy different positions in their respective market segments, they share the central objective of allowing employers to forego traditional routes to procuring labor supply—namely, maintaining a labor force or contracting out to someone who does. Like other similarly transformative reconfiguration of labor, such as subcontracting, independent contracting, and temporary staffing, employers have flocked to crowdsourcing and built it into a formidable industry in its own right.

A. The Crowdsourcing Industry As one might imagine in an industry built by and dependent upon the internet, crowdsourcing has grown primarily in online-friendly or online-exclusive sectors of the economy. These include web content, advertising, software development, audio/video transcription, database building, communication design, and market research. Information on what sizes and types of firms use crowdsourcing is sparse and incomplete, but the general consensus is that crowdsourcing attracts two kinds of employers: 1) small firms that rely entirely on crowdsourcing for their existence—such as SpunWrite, an “article spinning”20 provider; and 2) medium-sized firms looking to source specific projects or operational segments more cheaply or quickly.21 Paid crowdsourcing has experienced remarkable growth in the last ten years. According to Smartsheet, which provides online work management services to companies that use crowdsourcing, the paid crowdsourcing labor pool contains over one million workers. Those workers have earned $1-2 billion in the last decade. Crowdsourcing vendors, who typically charge a fee or commission, make over $500 million annually.22 Smartsheet places paid crowd labor into four categories, distinguished by volume of tasks, level of compensation, and degree of automation. “Micro tasks” generally appear in high volume, with

See SpunWrite.com homepage, http://www.spunwrite.com/ (last visited May 16, 2010). Article spinning is described in more detail in Part III. 21 See Frei, supra note 12, at 7; Paylancers.com, Cottage In, Cottage Out (Oct. 5, 2006), http://paylancers.blogspot.com/2006/10/cottage-in-cottage-out.html (describing why larger firms have yet to embrace Amazon’s Mechanical Turk crowdsourcing platform). 22 Frei, supra note 12, at 7. The figures presented in this section refer to the entire paid crowdsourcing industry, including competitive crowdsourcing and expert networks. Because crowdsourcing categories are porous and undefined, no studies have yet broken down the crowdsourcing industry by “model.”
20

6

Alek Felstiner – Working the Crowd DRAFT-5.16.10

correspondingly low compensation levels and near-complete automation. These include locating and copying information from websites, tagging online photos with relevant information, and categorizing online products. “Macro tasks” take slightly more time, because they are less automated and call for more discretion on the worker’s part. But they still appear in high volume and pay very little. Examples of macro tasks would include providing survey feedback, writing a short review of a website, product, or piece of entertainment, or compiling a list from multiple sources. “Simple projects” are not automated, and tend to pay more and demand more worker investment than tasks. Designing a simple website, building a database, or writing a basic piece of code would constitute a simple project. Complex projects are the most rare, because their performers often command higher rates of pay and require more active supervision. Complex projects are usually one-offs, and may overlap with work customarily performed in-house or by an established contractor. These include building the back-end of a complicated interactive website, designing a patentable product, or preparing a substantial business report.23 Crowdsourcing vendors tend to tailor their operation to one or more of these categories. For example, the anonymity and rigidity of the AMT platform makes it irrational to request a project of any complexity. The vast majority of the HITs on AMT are micro tasks, with some macro tasks mixed in. oDesk, on the other hand, has built a platform where firms and workers can interact and negotiate prior to entering into a transaction. Workers can set their own hourly rates and control other terms, while employers gain access to a more qualified and stable workforce—making possible the performance of a wider variety of work.24 It appears that crowdsourcing vendors decide what kinds of firms and workers they wish to attract, then construct an online platform and a set of policies to support those transactions. All that said, we know very little about crowd workers themselves, at least in abstract statistical terms. A few crowdsourcing vendors gather information about who actually performs this work, but they tend not to publicize it, perhaps because anonymity and fungibility of labor represent two of crowdsourcing’s prominent advantages. Some researchers and crowd workers themselves have begun to collect information on AMT, discussed in Part II, but for the most part, the demographics of the crowd remain a mystery.
23 24

Id. at 3. See oDesk.com, FAQs, http://www.odesk.com/w/faqs (last visited May 16, 2010).

7

Alek Felstiner – Working the Crowd DRAFT-5.16.10

There is, however, some information about compensation levels, and some indication of the ways in which crowdsourcing vendors choose to structure (or refrain from structuring) the employment relationship. At the low end of the spectrum, Providers on AMT earn an average rate of $1.25 per hour for their cognitive piecework.25 CrowdFlower sets a default rate of $2.00 per hour, based on estimated completion times.26 oDesk, which offers a panoply of services across the skill spectrum, reports an average hourly rate of $9.32.27 The oDesk rates do vary widely among job categories. Software developers earn $15.00 per hour on average, whereas designers and consultants earn closer to $10.00, technical writers earn $8.00, and data entry providers earn around $3.00.28 With hundreds of firms using any given platform, and potentially tens of thousands of workers, most crowdsourcing vendors make some effort to impose a default structure on the employment relationships. They tend to explicitly specify that providers of crowd labor will serve as independent contractors.29 Vendors also usually set all the ground rules regarding qualifications for work, supervision, payment, dispute resolution, and access to the platform. As the next Section illustrates, many of the benefits and risks of crowdsourcing flow directly from this somewhat unusual arrangement.

B. Why Crowdsourcing? And Why Not? Crowdsourcing is still relatively new. Many of its grand promises and dire predictions have yet to unfold. Nevertheless, firms and employees have already surged into the crowdsourcing market, seeking the unique advantages of the model and accepting—consciously or not—the associated risks.30 This Article does not discuss the advantages and drawbacks of crowdsourcing for the purpose of evaluating whether
25 JOEL ROSS, ET AL.,

WHO ARE THE TURKERS? WORKER DEMOGRAPHICS IN AMAZON MECHANICAL TURK 3 (2009), http://www.ics.uci.edu/~jwross/pubs/SocialCode-2009-01.pdf. See Part II for more details on the AMT workforce. 26 See CrowdFlower, http://crowdflower.com (last visited May 16, 2010). The default rate appears in a job-building screen only accessible after signing up for an account. 27 oDesk.com, oConomy – Global Statistics by Country and Skill, http://www.odesk.com/community/oconomy (last visited May 16, 2010). 28 oDesk.com, Hourly Rates by Job Category, http://www.odesk.com/community/oconomy/rate_statistics (last visited May 16, 2010). 29 See, e.g., Amazon.com, Amazon Mechanical Turk Participation Agreement § 3a-b, https://www.mturk.com/mturk/conditionsofuse (hereinafter AMT Participation Agmt.) (updated Apr. 29, 2009) (last visited May 16, 2010); oDesk.com, oDesk Marketplace User Agreement § 2.7, http://www.odesk.com/help/help/policies/user_agreement (last visited May 16, 2010); OnForce.com, Policies, http://www.onforce.com/policies (last visited May 16, 2010). 30 The motivations discussed here are limited to economic motivations. Social benefits and risks of crowdsourcing, though undoubtedly important, lie outside the scope of this Article.

8

Alek Felstiner – Working the Crowd DRAFT-5.16.10

crowdsourcing is ultimately “worth it,” or whether the industry will survive in the long run. The objective here is to explain what distinguishes crowdsourcing from past models of organizing work, primarily as a way of understanding what forces brought us to this point. i. What Firms Get Out of Crowdsourcing The employer advantages of crowdsourcing are difficult to overstate. When leveraged strategically, employers can reap rewards in a variety of ways, some of which resemble more traditional outsourcing models, and some of which are entirely unique to crowdsourcing. Two primary, much-touted advantages are the twin grails of scalability and on-demand labor. Given a sufficiently large network (not difficult to assemble on the internet), the crowd can accomplish tasks of practically any size. The workforce can also grow and shrink over time, according to the firm’s needs. Employers do not have to hire superfluous in-house staff, or locate and pay a premium for outside contractors capable of scaling the workforce up or down on demand. Employers can enter and exit crowdsourcing venues at their whim, without any significant transaction costs or logistical hurdles. They can also use the constant availability of a global labor pool to avoid the delays commonly associated with identifying and vetting outside contractors. Some firms using AMT can even obtain time-sensitive results in some approximation of “real time,” without having anyone oncall. For example, “human-augmented search” companies accept trivia questions by phone and use AMT to answer them in a matter of minutes.31 Better still, this flexibility comes at a relatively low cost. Depending on a firm’s quality standards, crowdsourcing can be astoundingly cheap. Crowd workers receive low wages, no benefits, no job security, and have not much prospect at present of organizing to change these conditions. Employers do not need to provide facilities and support for a workforce, nor do they need to pay overhead fees to an outside contractor. Because the employment relationship tends to be fleeting and largely anonymous, at least in some

See Katherine Mieszkowski, “I make $1.45 a week and I love it”, SALON.COM, July 24, 2006, http://www.salon.com/technology/feature/2006/07/24/turks/.
31

9

Alek Felstiner – Working the Crowd DRAFT-5.16.10

platforms, crowdsourcing adds little or no personnel cost. An employer does not need to hire managers to supervise the crowd, and can avoid turnover and recruitment expenses.32 Companies able to configure or retrofit their business to fit existing crowdsourcing platforms will realize the greatest gains in efficiency. Their best bet is to structure their operation in such a way that it relies on the completion of a high volume of discrete tasks (disaggregation), or on the unbundling of a task too large to attack in full (disintegration). Previously, companies seeking to avoid the inevitable bottlenecks and prohibitive delays of assigning such tasks to employees or subcontractors would simply turn to computers. But computers cannot necessarily perform these tasks as efficiently or reliably as the human brain. Now, those companies can get the benefit of human cognition without the bottlenecks or delays. Photo tagging, audio transcription, and web research represent only the tips of the iceberg. Crowdsourcing has the potential to transform the entire structure of the information economy. Going forward, companies will begin to gather their productive labor and necessary services in small and medium-sized chunks from an amorphous and ever-shifting cloud of workers. Crowdsourcing gives employers access to a broader range of skills and experience than they could ever achieve on their own, even through a combination of staff and outside contractors. The genius of this model is that employers do not have to locate workers who possess valuable attributes—in crowdsourcing, the people you need find you. ii. Inherent Risks The many advantages available to employers through crowdsourcing do not come without some related risks. Employers in most crowdsourcing platforms will almost certainly lose some control over the work and the manner in which it is performed. The ordinary employment relationship gives employers de facto control over the process and outcomes of work. Contractual relationships for specific services usually provide for similar control, if not waived by the parties. But crowdsourcing venues vary widely, and while some still

There remains some dispute as to whether the cheapness of crowdsourcing will persist. Jeff Howe argues that because “what unites all successful crowdsourcing efforts is a deep commitment to the community,” any employer who treats the crowd as a cheap labor source is “doomed to fail.” Howe, supra note 2, at 15. We don’t know what exactly constitutes a “deep commitment to the community,” but the success of crowdsourcing platforms such as AMT—which does very little to protect workers or give them a voice—may belie Howe’s contention. Treating individual workers as sources of cheap labor has hardly proven to be a recipe for failure in the past. If some are willing to work for substandard wages and benefits without legal protection, by necessity or choice, there is no reason to believe that their willingness will evaporate in cyberspace.
32

10

Alek Felstiner – Working the Crowd DRAFT-5.16.10

offer ways to exert considerable control, others necessarily carve out areas of worker discretion upon which the remote employer cannot impinge. With loss of control generally comes a loss of the certainty and accountability that might normally characterize a formal employment or contractual relationship. Even the most committed crowd worker will have less at stake than a formal employee, especially when building a positive reputation as a crowd worker only offers limited value because that reputation remains non-portable. Thus some crowdworkers may prove less concerned with meeting specifications and adhering to compliance policies. Firms care about control and accountability because distributing tasks to an anonymous pool can lead to real difficulties in ensuring the quality of the product. Smartsheet has asserted that low quality and unexpected results are “the single biggest factor[s] in companies choosing to abandon paid crowdsourcing.”33 Satisfaction clauses,34 or the right to reject sub-par work, can partially alleviate this problem. Of course, any added effort spent reviewing and rejecting work cuts into the savings that motivate firms to try crowdsourcing in the first place. Some vendors, such as CrowdFlower, have built entire applications to track quality and filter results.35 Employers might also address this problem by putting higher qualification restrictions in place for the task, where possible, or by using multiple workers on a single task to check or confirm work. These methods may limit risk, but the core problem of quality control will remain unless a firm is willing to make a significant investment in the sorts of online quality-assurance mechanisms that approximate real-world supervision and control. In addition to concerns around control and quality, firms may also encounter serious intellectual property risks in distributing tasks to a large pool of anonymous workers. Employer can and likely will attempt to design their requests so as to protect any proprietary material, but who knows what a crowd worker may be able to glean by completing even a small contribution to a valuable piece of intellectual

Frei, supra note 12, at 8. See, e.g., AMT Participation Agmt., supra note 29 at § 3a-b; oDesk.com, Billing and Payments, http://www.odesk.com/help/help/policies/billing_payments_policy#quality (last visited May 16, 2010) (imposing a satisfaction clause on fixed-price work only). It is worth noting that unlike other crowdsourcing venues, oDesk actually guarantees payment for hourly work. See oDesk.com, oDesk Guarantee, http://www.odesk.com/help/help/payments/financial_activity/guarantee (last visited Apr. 20, 2010) (“oDesk is the first and only service to guarantee that an hour billed is an hour worked and that an hour worked is an hour paid”). 35 Crowdflower.com, Solutions-Overview, http://crowdflower.com/solutions/index (last visited May 16, 2010).
33 34

11

Alek Felstiner – Working the Crowd DRAFT-5.16.10

property? As mentioned above, crowdsourcing vendors often impose privacy or nondisclosure policies as part of their participation agreement, or alongside it,36 but the those same vendors also tend to disclaim any responsibility for injuries suffered as a result of participation.37 The absence of any oversight may render those policies ineffectual, leaving proprietary information somewhat unprotected. The extent to which these drawbacks will become prohibitive depends on the size and structure of the firm, as well as the nature of the work performed. Over time, crowdsourcing models will likely develop that minimize these risks and make crowdsourcing “safe” for employers of all types.38 Despite the possibility of some fluctuation in benefits and risks, it seems a safe bet that the overarching rewards of crowdsourcing will endure as the industry grows. iii. What Attracts Workers to Crowdsourcing? Depending on the employee, crowd labor may confer unique value and opportunity. Compare the experience of joining a networked labor pool, from the comfort of your home or coffee shop, with the effort and drudgery of travelling to a workplace and occupying a prescribed space for a set period of time and doing tasks assigned by a supervisor, with little independence or flexibility. The primary advantage of being a crowd worker is the freedom you have to choose when and where you will work, how long you spend, and what work you perform. All you need to get started is a computer and a reasonably fast internet connection. So, as with crowdsourcing employers, the barriers to entry for crowd workers prove quite low, as do the costs and risk associated with exit. Such flexibility would have been unprecedented in the job market of the twentieth century, and remains quite rare today.39

See, e.g., Amazon.com, Amazon Mechanical Turk Privacy Notice, https://www.mturk.com/mturk/privacynotice (updated Aug. 5, 2009) (last visited May 16, 2010); oDesk.com, Privacy Policy, http://www.odesk.com/help/help/policies/privacy_policy (last visited May 16, 2010). 37 See, e.g., AMT Participation Agmt., supra note 29, § 8. 38 Of course, the advantages of early adoption may also diminish as more firms enter the arena. 39 These and other attributes of crowdsourcing make it a theoretically potent force for economic development in rural and wardamaged areas of the developing world. It is a low risk endeavor requiring little capital investment or employee training, ideal for NGOs, local governments, and entrepreneurs. People in developing countries can work directly for firms around the globe, without the sometimes costly and exploitative interventions of an outsourcing contractor. See Leila Chirayath Janah, Kenya Dispatch #4: Refugees and Remote Work (June 22, 2009), http://www.socialedge.org/blogs/samasourcing/archive/2009/06/22/kenya-dispatch4-refugees-and-remote-work (describing Samasource, a non-profit that trained Somali refugees at a UNHCR camp in Dadaab, Kenya to perform basic internet tasks of the type commonly posted on AMT).
36

12

Alek Felstiner – Working the Crowd DRAFT-5.16.10

Additionally, the choice of tasks built into crowdsourcing models means that employees can affirmatively select tasks to fit their unexplored interests or their existing knowledge base. Though monopoly of certain crowdsourcing platforms by particular sets of employers sometimes diminishes the available choices, crowd workers can still generally self-select to achieve a variety of assignments, to pursue learning in a specific area, or to entertain themselves on someone else’s dime. Taking full advantage of crowdsourcing opportunities may also help workers realize substantial gains in personal productivity. Crowdsourcing promises to convert our “spare cycles”40—essentially, periods when the brain is operating but not producing anything of value—into productive time. Instead of playing onscreen Solitaire or surfing the web, AMT and LiveOps put you to work tagging photos or taking restaurant orders over the phone. Without such platforms, how would a person go about monetizing the stray ten-minute increments that crop up throughout the day? A single employer would not likely hire an hourly employee to work during those scraps of time. But with crowdsourcing, every waiting room and bus stop becomes a temporary workspace.41 iv. The Drawbacks of Performing Crowd Labor First and foremost, crowd workers tend to receive extremely low pay for the cognitive piecework they perform. They usually earn no benefits and enjoy no job security, and in fact, the platform may seek to prevent them from doing so.42 There are no true legal protections for crowd laborers, and the cyberspace they occupy remains essentially unregulated for employment and labor law purposes. In addition to these fundamental drawbacks, crowd workers also encounter problems with information asymmetry, deception, and privacy. On AMT and similarly lopsided platforms, workers have very

The term “spare cycles” is borrowed from computer technology. A “cycle” refers to the process of a computer goes through to retrieve information from memory and execute an action. Distributed computing programs use the “spare cycles,” or downtime, of every computer in a large network to tackle computing tasks too big to perform in one location. Crowdsourcing adapts this concept to the human brain’s untapped cognitive powers, and attempts to put our brains’ spare cycles to productive use. See Clive Thompson, The Human Advantage, WIRED, July 2007, at 166. 41 In a broader perspective on productivity, Jeff Howe suggests that crowdsourcing may supply an answer to the “long-standing human conundrum” that “the amount of knowledge and talent dispersed among the numerous members of our species has always vastly outstripped our capacity to harness those invaluable quantities.” Howe, supra note 2, at 19. Howe argues that crowdsourcing reflects the “fundamentally egalitarian principle” that “every individual possesses some knowledge or talent that some other individual will find valuable.” Id. at 134. Theoretically, crowdsourcing can connect those who possess particular talents and knowledge with those in need of them, without the customary associated costs and barriers. As a result, those valuable assets will not be wasted through neglect, but put to productive use. 42 AMT Participation Agmt., supra note 29, at § 3b.
40

13

Alek Felstiner – Working the Crowd DRAFT-5.16.10

little information about their prospective employers and only limited information about the tasks to be performed. Essentially, they see only what the vendor and the employer want them to see. Employers, on the other hand, can usually see workers’ employment history (on AMT, this takes the form of rejection and acceptance rates). Many crowdsourcing vendors give employers the right to reject unsatisfactory work product, without necessarily relinquishing the right to use it.43 Information asymmetries, especially when combined with satisfaction clauses and the absence of a reliable dispute resolution system, will inevitably permit some fairly bald forms of deception. Employers can order work, receive it, and then reject it as unsatisfactory without much justification. To some degree, employers can also disguise the nature and quantity of the work in order to secure consent. On AMT, a worker may expect, based on the employer’s posting, to be paid a certain piece rate for each photo he or she categorizes. The worker may not discover until after accepting the job that the rate applies to batches of photos.44 At that point, the worker can either “return” the job, which negatively affects the worker’s reported completion rate, or finish the job under unforeseen and unsatisfactory conditions. These disclosure deficiencies may also present crowd workers with unusual moral and ethical challenges. Jonathan Zittrain has observed that because workers don’t know for whom they are working, and for what their work will actually be used, crowd labor can “deprive people of the chance to make judgments about the moral valence of their work.”45 Some companies use crowdsourcing to produce more authenticfeeling internet “spam” or fake product reviews. While irritating, and perhaps unethical, such abuses remain fairly innocuous. Their main danger is that they will clog the platforms so completely that firms and workers searching for more “legitimate” work will not be able to find each other. But we can easily see more disconcerting scenarios on the horizon. A pacifist software developer on Rent-A-Coder could end up writing a piece of software for a weapons manufacturer that becomes part of a sophisticated and deadly missile targeting program. Or, as Zittrain hypothesized, the government of Iran could cheaply identify protest

See, e.g., AMT Participation Agmt., supra note 29, at § 3a-b; oDesk.com, oDesk Marketplace User Agreement § 8, http://www.odesk.com/help/help/policies/user_agreement (last visited May 16, 2010). 44 AMT does show Providers a sample HIT from the set they will be doing, which in many cases will prevent such misunderstandings. But there is no real check on the accuracy or consistency of these HIT previews. And, in some cases, the amount of work required for each HIT simply does not become clear until the Provider begins to perform it. 45 Zittrain, supra note 2, at 5.
43

14

Alek Felstiner – Working the Crowd DRAFT-5.16.10

participants by using AMT to cross-reference photographs of the nation’s population against pictures taken during the protests.46 Finally, some crowdsourcing platforms involve at least a possibility of privacy violation. A great number of firms and researchers use AMT to conduct surveys, collect anecdotes or testimonials, and perform market research. In the process, workers often disclose personal information without a clear guarantee of confidentiality or responsible use by the Requester.47 Unfortunately, the privacy policies imposed on both parties, if even enforceable at all, may not cover such disclosures.

C. Crowdsourcing and Existing Industries Of course, the impact of crowdsourcing is not, and will not be confined to the firms and workers who actively participate. In fact the first coverage of crowdsourcing focused less on its potential to create new labor markets and more on its propensity to destroy existing ones. We must recognize that negative externalities, whether real, substantially probable, or utterly paranoid, have shaped crowdsourcing discourse from the beginning and will certainly shape emerging political and legal responses. Jeff Howe’s 2006 Wired article began by chronicling the plight of stock photographers, whose industry has partially collapsed following the emergence of iStockphoto and other “microstock” suppliers.48 “Microstock” refers to stock photography agencies that source images primarily from a large pool of amateur photographers and charge only a small fraction of the conventional stock photo price, without royalties. Where professional stock photographers charge $100 and up for a single photograph, the almost entirely amateur49 photographers on microstock websites sell for $1 to $5 a photo, royalty free.50 As microstock

Jonathan Zittrain, Work the New Digital Sweatshop, NEWSWEEK, Dec. 9, 2009, at 41. Professor Zittrain estimates that such an identification project would cost Iran’s government $17,000 per protestor on AMT. 47 See Part II for a summary of Turkers’ concerns with AMT, including privacy concerns. 48 See Howe, supra note 2, at 178. 49 For ninety-six per cent of iStock contributors, photography is not a primary occupation. Id. at xxi. 50 Id. at 178.
46

15

Alek Felstiner – Working the Crowd DRAFT-5.16.10

steadily grows,51 it threatens to swallow traditional stock suppliers and drive down the price of stock photographs. Crowdsourcing forces professional photographers to compete in a market that largely ignores their experience, high-end equipment, and reputations. Many creative professionals and providers of skilled services (such as software development) fear that their industries will go the way of stock photography. A particularly heated debate has arisen in the field of graphic design. Some established professional designers have loudly declaimed crowdsourcing platforms offering “spec” design, such as crowdSPRING and 99designs.52 crowdSPRING allows firms in need of creative labor—coding, graphic design, etc.—to post a project and receive completed work from as many responders as choose to perform the task. In the traditional creative marketplace, designers or coders would submit only a bid or proposal for a project, not the completed project itself. Firms using crowdSPRING pay in advance, receive actual work product from each responder, then simply choose what they like. 53 The chosen responder gets paid for its labor, and the rest do not. Some designers fear that design work submitted by amateurs, with no guarantee of compensation, will degrade the overall quality of design and lead to misperceptions about its importance. They also worry that spec design competitions will eliminate the role of designers as counselors and researchers, remove the collaborative aspect of the designer-client relationship, and lower the perceived value of the services that experienced graphic designers can offer. 54 Microstock and design contests do not exactly fit the cognitive piecework model exemplified by AMT. But it is not difficult to imagine crowdsourcing vendors like AMT replacing existing industries in data entry, audio transcription, tech support, even legal services.55 The same principle applies: replace a full-time or

iStockphoto is the third-largest purveyor of stock photos. The company expects to clear $200 million in revenue for 2009. Daryl Lang, iStock: We’ll Clear $200M This Year, PDNONLINE.COM, June 24, 2009, http://www.pdnonline.com/pdn/content_display/photo-news/stock-and-syndication/e3i772f176924f862d4e48e594b4c9c7e39. 52 See Jeff Howe, Is Crowdsourcing Evil?: The Design Community Weighs In, WIRED.COM, March 10, 2009, http://www.wired.com/epicenter/2009/03/is-crowdsourcin/; AIGA, AIGA Position on Spec Work, http://www.aiga.org/content.cfm/position-spec-work (last visited May 16, 2010); Neil Tortorella, Ten Reasons, http://www.nospec.com/articles/ten-reasons/ (last visited May 16, 2010). 53 See crowdSPRING.com, How it Works, http://www.crowdspring.com/how-it-works/ (last visited May 16, 2010). 54 Supra note 52. 55 3 Geeks and a Law Blog, Crowdsourcing – The Live Experiment (May 7, 2009), http://www.geeklawblog.com/2009/05/crowdsourcing-live-experiment.html.
51

16

Alek Felstiner – Working the Crowd DRAFT-5.16.10

subcontracted employee with workers from the pool. As cognitive piecework models expand, and platforms become more sophisticated, other information technology industries might also find themselves up for grabs. Of course, one cannot consider these negative externalities without also acknowledging the positive externalities created by crowdsourcing. Stock photographs has become much cheaper, as has graphic design. Individuals and small businesses can obtain products and services previously unavailable due to the prohibitively high cost of paying for professionalism.56 It is also worth noting that the very concept of crowd labor replacing existing industries has been challenged on the grounds that the products and services generated by the crowd reach different sectors of the consumer market, or are simply too different to overlap with those generated by existing industries. After all, a high-end design firms isn’t necessarily losing small business clients if those clients would never have engaged the firm in the first place. The cognitive piecework model may well intersect with existing service industries, but its capacity to harness economies of scale also gives it the potential to create entirely kinds of services and open up heretofore unimagined industries.57

PART II. AMAZON’S MECHANICAL TURK “[Mechanical Turk] gives us a snapshot of a depressing future in which legions of clickslaves toil away at identifying duplicate Web pages for less than minimum wage. Amazon says it hit on the idea for Mechanical Turk when it realized that there were some tasks that even the smartest computers couldn’t perform. I’ve got an alternate theory. Maybe the computers just didn’t want to.”58 Apparently, Amazon did not create AMT with the intention of marketing a crowdsourcing service. The company, which sells or facilitates the sale of a seemingly unlimited number of retail products, had over time built up millions of web pages describing those products. Some were bound to be duplicates, but it turns

Another benefit of deskilling and deprofessionalization, though not exactly an externality, is that more people can enter previously exclusive professions. It’s a fairly obvious point, but in times of high unemployment, the greatest advantage of crowd labor may be that it’s available. 57 For example, ReTel Technologies uses crowd labor to analyze snippets of retail store security video in order to create demographically-keyed heatmaps of a store interior. Crowd workers track the age and sex of the customers, identify which product areas the customers spend the most time browsing, etc., in order to give the owners a precise picture of how their customers respond to product groups and sales-floor configuration. Video-analyzing software would have great difficulty making these kinds of judgments, based on pattern recognition. See ReTel Technologies, ReView Analysis Suite, http://www.reteltechnologies.com/home/review-retail-analysis-suite/ (last visited May 16, 2010). 58 Jeff Howe, Taking Measure of Mechanical Turk (Nov. 3, 2006), http://www.crowdsourcing.com/cs/2006/11/taking_measure_.html.
56

17

Alek Felstiner – Working the Crowd DRAFT-5.16.10

out that computer programs aren’t particularly efficient or effective at recognizing duplicate web pages. The human mind, on the other hand, can perform that task in a matter of seconds. So Amazon hit on the idea of paying users a few cents for every duplicate page they could find.59 In addition to the quote about “legions of click-slaves,” Howe has also described AMT as “the lowest-common denominator variety of crowdsourcing.”60 Perhaps this pessimism stems from AMT’s inauspicious beginning. Amazon used its valued customer base to perform routine, menial tasks it would otherwise have to assign to employees or contractors. This smacks of race-to-the-bottom outsourcing, and displays very little of the collaboration, innovation, and creativity that inspires Howe and crowdsourcing’s other evangelists. Yet AMT is almost certainly the largest crowdsourcing platform on the web, and has become the first stop for many individuals and firms seeking cheap, on-demand crowd labor. As such, it provides an excellent case study for evaluating the legal ramifications of crowdsourcing. This Part explains how AMT works, lays out the rules imposed by Amazon on Requesters and Providers, and describes the people who participate and their reasons for doing so.

A. How AMT Works Amazon provides an online platform for firms or individuals to solicit and accept “Human Intelligence Tasks” (HITs). “Requesters”—those soliciting the HIT—and “Providers”—those accepting— must sign up for an Amazon.com account, which subscribes them to an online payment service and creates an AMT identity. In the process, they provide Amazon with some personal and tax information. Users must also consent to AMT’s User Participation Agreement. Requesters post their task(s) to the website, including the compensation (called a “reward”) and the duration of the HIT. Providers (also colloquially known as “turkers”) browse through the HITs, look at example tasks, and accept whichever HITs they choose. Once the Provider accepts an HIT, he or she must
Jason Pontin, Artificial Intelligence, With a Little Help From the Humans, N.Y. TIMES, Mar. 25, 2007, at 35. Jeff Howe, Mechanical Turk Targets Small Business (Aug. 1, 2008), http://www.crowdsourcing.com/cs/2008/08/mechanicalturk.html. It may be worth noting that Howe made the comments quoted here in 2006 and 2008, respectively. A prefatory note in the second printing of his book on crowdsourcing exhibits a more ambivalent tone. See Howe, supra note 2, at xiv.
59 60

18

Alek Felstiner – Working the Crowd DRAFT-5.16.10

complete it within the designated time, and submit the work through the website. The Requester can then accept or reject the work and authorize payment. There is no set schedule for the acceptance or rejection of an HIT, or for the payment of any rewards earned. The Provider can choose whether to have the money transferred to a bank account, or to an Amazon.com gift certificate account. If a Requester accepts the HIT and pays the Provider, the Requester must also pay a 10% service fee to Amazon, on top of the reward amount. Amazon also requires that Requesters place the full amount of the reward, plus the service fee, in a payment account before posting the HIT. In some cases, Requesters will also pay “bonuses” to a Provider, at the Requester’s discretion.61

B. Amazon’s Terms of Use Requesters and Providers have no real knowledge of each other on AMT. Because of the way Amazon structured the platform, a Requester’s posting of an HIT operates as a unilateral contract offer, which the Provider essentially accepts through performance. AMT contains no opportunity or method for negotiation. As a result, Amazon’s Terms of Use, to which both parties have agreed, functions as the transaction’s only governing document. But Amazon does not wish to involve itself at all in the RequesterProvider relationship, as this disclaimer from the Participation Agreement clearly articulates: 2. Amazon Mechanical Turk's Role. Amazon Mechanical Turk provides a venue for third-party Requesters and third-party Providers to enter into and complete transactions. Amazon Mechanical Turk and its Affiliates are not involved in the transactions between Requesters and Providers. As a result, we have no control over the quality, safety or legality of the Services, the ability of Providers to provide the Services to Requesters' satisfaction, or the ability of Requesters to pay for Services. We are not responsible for the actions of any Requester or Provider. We do not conduct any screening or other verification with respect to Requesters or Providers, nor do we provide any recommendations. As a Requester or a Provider, you use the Site at your own risk.62 Yet despite this disclaimer, the Participation Agreement sets fairly strict guidelines for the use of the site and does attempt to govern some aspects of the Provider-Requester relationship. In addition to the prepayment obligation, the Participation Agreement essentially builds in a mandatory satisfaction clause, which
Amazon.com, Requester Website FAQs, https://www.mturk.com/mturk/help?helpPage=requester (last visited May 16, 2010); Amazon.com, Worker Website FAQs, https://www.mturk.com/mturk/help?helpPage=worker (last visited May 16, 2010). 62 AMT Participation Agmt., supra note 29, at § 2.
61

19

Alek Felstiner – Working the Crowd DRAFT-5.16.10

authorizes the Requester to “reject” any submitted HITs, without paying the Provider, giving a justification, or forfeiting the work. The decision to accept or reject a submitted HIT remains entirely within the Requester’s discretion.63 The Participation Agreement also mandates that Providers will only submit work, and Requesters will only accept it, through the AMT website. This prevents the parties from contracting independently, and ensures that Amazon will receive its service fee.64 Amazon also has an interest in the maintenance, or rather the neglect, of Prepaid HIT accounts. First, Prepaid HITs are non-redeemable and non-transferable. If an account lies dormant for thirty months, Amazon automatically converts the funds to Amazon gift certificates. Second, if Amazon removes a Requester for violating the Participation Agreement, Amazon gets to collect the balance after paying out any unpaid HITs.65 Providers, unsurprisingly, get the short end of the stick. Along with the satisfaction clause, the Participation mandates that the work product will be “work made for hire,” which means that all ownership rights (including intellectual property) vest with the Requester upon performance, whether or not the Requester chooses to actually pay for the work.66 The Agreement also asserts that Providers will perform services as independent contractors, and not as employees. In that capacity, Providers must acknowledge and agree, among other things, a) “not to use robots, scripts, or other automated methods to complete the Services,” b) to furnish the Requester with “any information reasonably requested,” and c) to agree that they (Providers) will not be entitled to any employee benefits, and will not be eligible to recover worker’s compensation if injured.67 Amazon can also cancel a Provider account at any time for violation of the various interconnected terms of use imposed by Amazon upon registration. When this happens, the Provider may forfeit any earnings left in his or her Amazon account. The Agreement does appear to contain some privacy safeguards for Providers, in that it permits Requesters to use information “solely to the extent necessary for you to use the Site and for no other

Id. § 3a-b. Id. 65 Amazon.com, Mechanical Turk Prepaid HITs Terms and Conditions §2-3, https://www.mturk.com/mturk/prepaidterms (last visited May 16, 2010). 66 AMT Participation Agmt., supra note 29, at § 3b. 67 Id.
63 64

20

Alek Felstiner – Working the Crowd DRAFT-5.16.10

purpose, including…solicitation, advertising, marketing, unsolicited e-mail or spamming, harassment, invasion of privacy, or otherwise objectionable conduct.”68 Requesters often obtain private information from Providers in response to a paid survey. But the Agreement leaves wide latitude for Requesters to use Provider information within the AMT platform, in ways that might compromise privacy without constituting an invasion or “objectionable conduct.” Perhaps in reaction to the inevitable problems arising under these terms, Amazon also clearly strives to extricate itself from any disputes that may emerge during the course of dealing. Section 3f of the Participation Agreement provides that “[b]ecause Amazon Mechanical Turk is not involved in the actual transaction between Providers and Requesters, Amazon Mechanical Turk will not be involved in resolving any disputes between participants related to or arising out of the Services or any transaction.”69 This would seem perfectly clear and conclusive, but becomes more complex due to Amazon’s existing guarantees for account users. The “Amazon A-to-z Guarantee” covers all payments made through the Amazon Payments system, but exempts “payments for services”—which category presumably includes AMT.70 However, Amazon Payments’ Buyer Dispute Program specifically applies to “transactions that are not covered by Amazon A-to-z Guarantee,” and states that users “may still seek our assistance in resolving dispute for these items by submitting a dispute.”71 These contradictory terms, which AMT users must accept in order to participate, leave a slightly confusing picture of to what degree Amazon will take responsibility for disputes. Finally, the Participation Agreement contains the expected (and broad) General Release. But rather than just disclaiming all liability without explanation, the release clause is premised specifically on the fact that “Amazon Mechanical Turk is not in

Summary This Article confronts some of the thorny questions that arise in applying traditional employment and labor law precepts to crowdsourcing, and offers some provisional solutions. To that end, the Article makes two related arguments for regulatory intervention. First, the nature of the certain crowdsourcing models makes workers uniquely vulnerable and prevents them from achieving decent wages and working conditions. Second, current employment and labor laws already fail to adequately address the modern physical workplace, and without immediate intervention, those laws will become completely irrelevant and anachronistic as more and more work shifts into cyberspace.

Description Crowdsourcing relies on the crowd for its very existence. Legal intervention can only buttress and protect the organized efforts of crowd workers; it cannot replace those efforts. If crowd workers follow the crowdsourcing model itself, and harness their shared potential to be greater than the sum of their parts, they could transform virtual work and the alter the future of the information economy.

Tags: crowd labor, Crowdsourcing, crowdsourcing industry, employment, labor laws, Virtual Work

Posted by Rebecca Gutierrez

Apr 23, 2011 03:17 am GMT Viewed 3790 times

Author: Alek Felstiner Source

Contact us if you are owner of this content and want to make changes.

Download Flag This

Events

Working the Crowd: Employment and Labor Law in the Crowdsourcing Industry

Comments

Crowdsourcing.org

Categories

Content

Crowdsourcing.org, an initiative by massolution®

Events

Related Content

Working the Crowd: Employment and Labor Law in the Crowdsourcing Industry

Comments

Crowdsourcing.org

Categories

Content

Crowdsourcing.org, an initiative by massolution®