Page 95 - Kỷ yếu hội thảo khoa học lần thứ 12 - Công nghệ thông tin và Ứng dụng trong các lĩnh vực (CITA 2023)

Thanh Trung Nguyen, Hai Bang Truong 79

knowledge within the crowd. It helps consolidate information, filter out noise,
identify consensus, handle conflicting information and enhance collective accuracy.
The reason why it is possible to aggregate multiple predictions and yield a superior
outcome compared to individual predictions can be attributed to the Law of Large
Numbers [2]. [3], predictions are modeled
as the truth plus a disturbance. The crowd wisdom is founded upon the aggregation of
individually independent guesses, which feature random or symmetrically distributed
errors. When numerous unbiased individuals make an estimate, the likelihood of
errors being made on both higher and lower sides of the correct answer becomes
balanced. By averaging the answers, the errors are mitigated due to the Law of the
Large Numbers.
Currently, prevailing researches on the Wisdom of Crowds phenomenon primarily
concentrate on the significance of diversity, decentralization, and independence.
Nevertheless, fewer studies have delved into the aggregation methods that aid in
consolidating a collective decision, although the process of consolidating a collective
decision from a group of individuals with varying opinions and perspectives can be
complex and challenging. Therefore, in this article, we will provide an overview of
the aggregation methods utilized in previous studies on the wisdom of crowds.
Understanding these methods is crucial in order to extract the collective wisdom from
a group of individuals.

2 Methodology

Studying the wisdom of crowds spans across various fields, resulting in a substantial
number of research papers. To address the problem statement, we utilize a traditional
review method explained in [4].
The first step involves identifying the topic of the literature review. As mentioned
above, our aim is to delve into the commonly employed aggregation methods in
recent studies on the wisdom of crowds, as well as the reasons behind their preferred
usage. In the second step, to find relevant papers, we employ Google Scholar as our
search engine, utilizing keywords like "wisdom of crowd," "collective intelligence,"
and "aggregation method." To streamline our search process, we limit our exploration
to approximately 10 result pages per keyword. We focus on retaining papers
published in 2015 or later. For papers published prior to 2015, we establish a filtering
criterion of 10 citations. Subsequently, in order to analyze and synthesize the
literature, we proceed with an initial review of the collected articles to gain an under-
standing of their content. This involves reading the abstracts, introductions, and
conclusions of the papers. After completing the initial overview, we revisit the articles
and conduct a more systematic and critical review of their content, employing a
structured approach proposed by Hendry and Farley [5].

ISBN: 978-604-80-8083-9 CITA 2023