Large-Scale Data Processing Technique - Research Paper Example

? HYPOTHETICAL DESIGNS Envisioned Research Problem A popular, large-scale data processing technique that has been extensively utilized in recent times for tasks that need direct human input is crowdsourcing. Crowdsourcing, according to Howe and Robinson (2006), refers to a novel online business model and problem-solving technique that utilizes the creative capabilities of a distributed pool of individuals through an open call (Brabham, 2008). Popular examples include Inn-Centive, Threadless, the Golcorp Challenge, Netflix, user-generated advertising competitions, Amazon Mechanical Turk, and iStockphoto among others (Brabham, 2008). The fact that crowdsourcing is dependent on large, distributed; global network of individuals, raises a set of new challenges. These challenges, which include dishonesty and plausible misjudgments, threaten the quality of results obtained through this process. Certain measures have, however, been put in place, to ensure high quality, and error free results. There is little or no attention given to the efficiency and throughput of the crowdsourcing process or the integrity of the results obtained. It is argued that the numbers of task workers and tasks are always small, thereby, resulting in crowdsourcing techniques that are not conscious of the number of tasks, potential worker behavior and efficiency of the process. This research aims at proposing a crowdsourcing, result-improvement technique that is independent of task complexity and sizes and ensures result quality, integrity, as well as efficiency and throughput of the process. The hypothesis being studied in this case is that crowdsourcing result-improvement techniques that are task size and complexity independent ensures result integrity, quality, efficiency as well as throughput. Hypothetical Designs Quantitative design This design will utilize experimental research method - methods that aim at maximizing replicability, generalizability, and objectivity of results; mostly concerned with prediction (Creswell, 2009). The focus will be to test several existing crowdsourcing techniques including r-Redundancy, v-Voting, and Vote Boosting techniques on a large number of tasks that will be handled by a large number of users. These techniques will be considered as experiment participants. The independent variable in this case would be the crowdsourcing techniques, including r-Redundancy, v-Voting, Vote Boosting techniques and the technique that this research will propose. The independent variables will be studied in two level; low task, less complex task level, and high number of tasks, and complex tasks level. Consequently, the dependent variables will be integrity, quality, efficiency as well as throughput. The experiment will be set in such a way that, the tasks set for testing, will have two definitive parameters including the accuracy of the tasks initial states and the number of options available per decision. 8 sets of 100,000 tasks with 3, 4, or 5 options and 75%, 85%, and 95% as the accuracy for the initial set tasks. There are about 4 to 10 decisions distributed normally. The user network or population tested also has two parameters including mean probability for committing errors and for dishonesty. Values of 3%, 6% and 20% are used for both dishonesty and making of errors. These probabilities were distributed exponentially over [0, 1] around their mean values. Simulations for about 40 input-aggregation functions with each one receiving one input are run repeatedly. For this experiment, the proposed quantitative hypothetical design is deemed to be extremely expensive even in the event that only few points in a parameter space are covered. Qualitative Design In this case, this research method will aim at understanding and discovering the perspectives, thoughts and experiences of previous researchers and participants in the same field in order to understand reality, purpose and meaning (Trochim & Donnelly, 2008). The focus will be to review and evaluate literature on previous and current research work on several existing crowdsourcing techniques including r-Redundancy, v-Voting, and Vote Boosting techniques on a large number of tasks that will be handled by a large number of users. In this case, the independent variable would be the crowdsourcing techniques, including r-Redundancy, v-Voting, Vote Boosting techniques and the technique that this research will propose. The dependent variables will be integrity, quality, efficiency as well as throughput. Thorough evaluation will be conducted while proper consideration will be given to dishonesty, errors, efficiency or results and on throughput. This will help understand purpose, reality and meaning and make an informed decision. This design will not be appropriate for this study because, it will be difficult to test the proposed crowdsourcing technique against the existing ones since there is no literature or case studies to this effect. Additionally, it is difficult for the researcher to ignore their own perceptions, biases and experiences and pretend to objective while reviewing and evaluating previous case studies and literature on existing techniques. Mixed Methods Design This design combines both quantitative and qualitative research methods with the aim of ostensibly bridge their differences with the aim of meeting the objectives of the research (Research methods: Design of investigations, 2004). The independent variable in this case would be the crowdsourcing techniques, including r-Redundancy, v-Voting, Vote Boosting techniques and the technique that this research will propose. The dependent variables will be integrity, quality, efficiency as well as throughput. This design will focus on proving whether or not, crowdsourcing result-improvement techniques that are task size and complexity independent ensures result integrity, quality, efficiency as well as throughput. This will be achieved through experimental design in, which simulations will repeatedly be done on the existing techniques, to ascertain their level of efficiency, throughput, and results integrity. This will be complemented with a review and evaluation of literature on the existing crowdsourcing techniques including r-Redundancy, v-Voting, and Vote Boosting techniques on a large number of tasks that will be handled by a large number of users. The results obtained will then be compared to the simulation results of the technique that this research will propose. This design hypothesis would be ideal and appropriate for this study, since, data will be collected and evaluated using varying methods and strategies that would result in completing of overlapping and non-overlapping strengths and weaknesses. It will complement the weakness of quantitative, experimental design method that is; it’s extremely expensive nature. Additionally, this design is complementary, pluralistic, and inclusive (Eysenck, 2004). References Brabham, D. C. (2008). Crowdsourcing as a Model for Problem Solving. The International Journal of Research into New Media Technologies, 14(1), 75–90. Creswell, J. W. (2009). Research design: Qualitative, quantitative, and mixed methods approaches (3rd ed.). Thousand Oaks, CA: Sage Publications. Eysenck, M. W. (2004). Psychology: An International Perspective. New York: Taylor & Francis. Research methods: Design of investigations. (2004). Psychology Press Ltd. Trochim, W., & Donnelly, J. (2008). The research methods knowledge base (3rd ed.). Mason, OH: Cengage. Read More