Walker, Jon DMC
(2016)
Wisdom of the Crowd Mechanisms.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
As Web 2.0 facilitates the collection of a vast amount of interactions, a phenomena, known as the wisdom of the crowd, is increasingly enlisted to justify using those interactions as surrogates for expert opinions. This dissertation explores this phenomena through an analysis of the micro elements of two wisdom of the crowd simulations: (1) Hong and Page’s (2004) Diversity trumps ability model and (2) Luan et al.’s (2012) Fast and Frugal simulation. The focus of this study is on the micro elements that contribute to those simulations’ results. This focus leads to the identification of a search mechanism that favors exploitation as a first step followed by exploration as defined by March’s (1991) Exploration/Exploitation simulation.
Three new methods for creating a group of experts were developed and were shown to be not only superior to the Top 10 agents but also superior to the more diverse random group of ten agents which consistently outperformed the Top 10 agents in the Hong-Page model. It was also shown that these expert groups were more efficient in incorporating the entire range of heuristics possessed by the universe of agents. The problem spaces were manipulated in various manners and the effect of such manipulations demonstrated. Additionally, group process losses were demonstrated through the simulation of a Hidden Profile scenario in which skills possessed by only one agent were ignored by the group. The effect of the dichotomization rate in the Fast and Frugal paradigm was highlighted and the effect of an alternative dichotomization rate demonstrated along with increasing the number of cues and manipulating the degree of correlation among them. Additionally, a set of perfect cue weights was developed for the Fast and Frugal paradigm and a simulation showed how a single agent executing the paradigm to choose the correct alternative saw its ability deteriorate as the cue weights progressed from the perfect order to all cues being equally weighted while groups of agents experienced increasing accuracy over the same progression.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
2 June 2016 |
| Date Type: |
Publication |
| Defense Date: |
4 March 2016 |
| Approval Date: |
2 June 2016 |
| Submission Date: |
2 June 2016 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
220 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Information Sciences > Information Science |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
crowdsourcing |
| Date Deposited: |
02 Jun 2016 14:03 |
| Last Modified: |
15 Nov 2016 14:33 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/28113 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[img]](https://d-scholarship.pitt.edu/style/images/fileicons/application_pdf.png)
![[feed]](/images/feed-icon-32x32.png){kind=icon}