Rules of Crowdsourcing: Models, Issues, and Systems of Control

Abstract

In this article, the authors first provide a practical yet rigorous definition of crowdsourcing that incorporates “crowds,” outsourcing, and social web technologies. They then analyze 103 well-known crowdsourcing web sites using content analysis methods and the hermeneutic reading principle. Based on their analysis, they develop a “taxonomic theory” of crowdsourcing by organizing the empirical variants in nine distinct forms of crowdsourcing models. They also discuss key issues and directions, concentrating on the notion of managerial control systems.

Keywords:

• crowdsourcing
• outsourcing
• e-business
• social web
• social media
• advanced internet technologies
• Web 2.0

ACKNOWLEDGMENTS

The authors would like to thank the editor and the two annonymous reviewers of this journal for their valuable comments and suggestions. The reviewers and editors of the Fourth Global Sourcing Workshop (March 22–25, 2010, Zermatt, Switzerland) also provided helpful comments on an earlier version of this article.

Notes

1. We use the terms social web, social media, advanced internet technologies, advanced web technologies, Web 2.0, and just “Web” interchangeably in this paper, as they are all contemporary terms but simultaneously realizing that no one term will endure forever due to the rapidly evolving nature of these technologies.

2. In fact, empirical research on 166 R&D challenges has shown that online communities can be better at solving research questions than a limited pool of internal researchers (Lakhani et al., 2007).

3. The crowdsourcing model is therefore distinct from pure “Web 2.0” or social media applications, which primarily stress the “social” aspect of community.

4. In short, though firmly built on the Web 2.0 platform, the crowdsourcing model transforms the Web into a virtual production platform through the implementation of sophisticated sourcing strategies.

5. We thank an anonymous reviewer for pointing this out.

6. As described on the website, “In our experiment in ‘crowdsourcing,' where we employ you, the listener, in an act of journalism, we asked you to go outside and count the number of SUVs on your block, as well as the number of regular cars” (http://www.wnyc.org/shows/bl/suv_map_07.html).

7. For example, the Cambrian House website mentions a spin-off crowdsourcing venture called VenCorps (www.cambrianhouse.com/vencorps) as well as an additional candidate organization, OGGTours (http://oggtours.com).

8. In line with the convenience sample approach, we also considered as candidates any crowdsourcing organizations we came across in our everyday readings, etc.

9. Krippendorf (2004, p. 3) describes content analysis as “systematic reading of a body of texts, images and symbolic material, not necessarily from the author's perspective.”

10. In the context of the interpretive IS research method, Klein and Myers (1999)paraphrase Gadamer's principle of the hermeneutical understanding process as follows: “The process of interpretation moves from a pre-cursory understanding of the parts to the whole and from a global understanding of the whole context back to an improved understanding of each part, i.e., the meanings of the words” (p. 71).

11. According to Klein and Myers (1999) decontextualized reading is a violation of the fundamental principle of the hermeneutic circle.

12. See the Appendix for a complete list of all 103 organizations categorized by model type.

13. Amazon Mechanical Turk, http://www.mturk.com/mturk/welcome Accessed March 27th, 2008.

14. http://www.nielsenbuzzmetrics.com/pr/releases/141 (Accessed September 25, 2009).

15. http://www.epinions.com (Accessed January 1, 2009).

16. For example, Citizendium.com permits authoring only with real name under the supervision of editors who demonstrated the proven expertise in a specific domain. Google's Wiki service (Knol.Google.com) also allows only real name authoring, and the author potentially receives monetary compensation through advertisement links. Note that, although Wikipedia is frequently cited as one of the premier crowdsourcing models (e.g., Bruns, 2007;Howe, 2008), we exclude it and similar wikis from our definition of crowdsourcing. The major reason is that the wiki platform itself is now distributed as a piece of software that embeds “a series of workflow process” (Stvilia et al., 2008, p. 984) to improve and maintain the quality of user-generated contents. Some readers may argue that the embedded “series of workflow process” mechanism can be considered a “managerial control system” to improve information quality. Or others may insist that a Wiki is not just a technological platform, but a socio-technical collaborative platform where online users collectively generate corpora of human wisdom. Despite such reasonable contentions, the Wikipedia model has been copied into various version of software, including Joomla, MediaWiki, DokuWiki, Wiki2Go, WikyBlog, RhizomeWiki, jspWiki, JbossWiki or Soks, to name a few. The diffusion of the Wikipedia model into various types of software implies that Wiki itself has become a free-of-charge commodity that can be easily copied and distributed. However, we do include Google's Knol, a variant of Wikipedia, as a crowdsourcing business model. Although adapted from Wikipedia, Google turns it into a business model by employing concrete managerial control schemes. For example, unlike traditional wikis, Knol supports authorship as a copyright holder, gives the option to display Google's advertisement links as revenue sources for both authors and Google, and allows authors to open, moderate, or close collaborative writing at their will. These unique features make it hard for competitors to copy, and this idiosyncrasy brings about a strategic advantage that cannot be easily copied by others.

17. CAPTCHAs (Completely Automated Public Turing Test to Tell Computers and Humans Apart) are “wavy” or distorted text images that appear next to or below web registration forms or text input boxes. This text is used to identify if a real human inputs information or if a spam bot is trying to fool the web application.