Date of Completion

2014

Document Type

Honors College Thesis

Department

Computer Science

First Advisor

Josh Bongard

Keywords

crowdsourcing, computer science, machine learning, linear regression

Abstract

Obtaining hypotheses for scientific experiments can be an exceptionally challenging and time-consuming task. Even with specialized knowledge, researchers frequently overlook important variables that may be strongly correlated with their outcome of interest. This experiment explores the use of crowdsourcing as a potentially more efficient way of gathering hypotheses. Testing the effectiveness of crowdsourcing involved the creation of a website that would enable users to build an online survey consisting of questions that may be predictive of a specified response variable. Contributions were obtained from members of online communities interested in subjects related to the area of research being studied. Allowing users to answer preexisting questions and add new questions to the survey resulted in a substantial amount of data on hundreds of potentially related variables. This data could then be used to determine which of the proposed factors are most correlated with the outcome of interest. Strongly correlated variables could later be studied in more detail in future experiments. Though this experiment specifically aims to identify factors that may be predictive of someone’s personal savings, this method of crowdsourcing can be replicated by researchers in any field of study. Using crowdsourcing to gather preliminary data has the potential to improve the efficiency of academic research and increase the rate of scientific discoveries.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Share

COinS