Page 94 - Kỷ yếu hội thảo khoa học lần thứ 12 - Công nghệ thông tin và Ứng dụng trong các lĩnh vực (CITA 2023)
P. 94

78


                           Aggregation Methods in the Wisdom of Crowds:

                                               A Literature Review




                                          Thanh Trung Nguyen , Hai Bang Truong
                                                                                  2
                                                               1
                                1  Faculty of Information Technology, Ho Chi Minh City University of
                                     Foreign Languages and Information Technology, Vietnam
                                                trungnt2@huflit.edu.vn
                                   2  Saigon International University, Ho Chi Minh City, Vietnam
                                               truonghaibang@siu.edu.vn





                           Abstract.  The  phenomenon  of  the  Wisdom  of  Crowd  has  been  observed  in
                           several  problem  domains  where  the  collective  opinions  of  groups  tend  to  be
                           more  accurate  than  those  of  individuals.  In  this  context,  the  aggregation  of
                           individual opinions is a crucial factor that determines the success of unleashing
                           the  wisdom  of  the  crowd.  This  paper  aims  to  conduct  a  literature  review  on
                           various  aggregation  methods  employed  in  recent  researches.  We  found  that
                           despite  the  increasing  number  of  studies  that  deal  with  large  and  intricate
                           datasets,  conventional  aggregation  methods  such  as  arithmetic,  geometric,
                           weighted  aggregation,  and  mode  continue  to  be  widely  used  in  the  field  of
                           Wisdom of Crowds.


                           Keywords:  Wisdom  of  Crowd,  Aggregation  methods,  Arithmetic  average,
                           Geometric average, Weighted average.



                     1     Introduction


                     The  concept  of  Wisdom  of  Crowds  refers  to  the  approach  of  the  aggregation  of
                     multiple individual estimates to obtain a better collective one. Such an approach can
                     outperform  individuals,  even  domain  experts,  in  various  prediction  and  estimation
                     tasks [1].  Surowiecki claims  that  a  mathematical  or  statistical aggregation over  the
                     judgments of a group of individuals can be more accurate than those of the average
                     individuals because of the benefit of error cancellation [1]. In his work, he describes
                     four characteristics that make a crowd intelligent. First, the group should be diverse,
                     as  this  allows  for  various  individuals  to  complement  each  other  by  contributing
                     unique  pieces  of  information.  Second,  a  decentralized  structure  is  crucial  for  the
                     group,  without  any  centralized  authority  directing  or  influencing  the  answers  of
                     individuals.  Third,  it  is  also  essential  that  the  individuals  within  the  crowd  act
                     independently of one another. Fourth, when the information of many individuals is
                     pooled,  they  must  be  aggregated  into  a  collective  opinion,  with  numerical
                     contributions and statistical methods often serving as the basis for aggregation. While
                     diversity,  independence,  and decentralization  are  important  factors,  the  aggregating
                     method plays a key role in consolidating and synthesizing the individual opinions and



                     CITA 2023                                                   ISBN: 978-604-80-8083-9
   89   90   91   92   93   94   95   96   97   98   99