A trend study on the impact of social media in decision making

Article history: Received: October 28, 2018 Received in revised format: January 20, 2019 Accepted: February 7, 2019 Available online: February 7, 2019 Social media has grown steadily during the last decade and it is now considered as a new opportunity to use for different purposes such as decision making. The primary objective of this paper is to review articles related to social media and decision making using manual and bibliometrics analysis methods, and to identify top themes in these articles. We have reviewed the papers published between 2008 and the first month of 2019 in Scopus where 1,159 articles were published in this period. These articles come from 733 sources and 3,459 authors. According to our survey, United States is the most productive country. Moreover, most collaborations occurred between two countries of United States and United Kingdom as well as between United States and China. The bibliometrics analysis examines global research in this field from the different point of views. © 2019 by the authors; licensee Growing Science, Canada.


Introduction
Social media (SM) includes divulgence websites that help relationship happening between users from different backgrounds, arising in a rich social structure.User-generated content persuades requirements and decision-making.According to the relevance of SM for different stakeholders, it has gained substantial attention from scholars of different areas such as information and decision sciences.To the best of authors' knowledge, there exists no comprehensive review that integrates and synthesizes the findings of the literature on social media (Kapoor et al., 2018).The subject matter of decision making is applied in a variety of fields, and the reviewed articles also show that the articles presented have a variety of backgrounds (Avudaiappan et al., 2018).A large amount of information on social media such as text, photos, and videos shared by users indicate there are two main categories on decision making.
 How social media manipulates users' decisions?
 How institutions, organizations, and governments use social media to make decisions?Prior to this, there were limited studies on the review of social media literature (Kapoor et al., 2018), and research was not focused on the decision-making using the social media.Many articles have been examined both in terms of the positive effect of this phenomenon (e.g.Gao et al., 2011;Munar and Jacobsen 2014b;Yates and Paquette 2011) and its negative impact (e.g.Dubé et al., 2014b;Kata 2012b;Reyna 2012).One of the major topics in these articles is the impact of social media on health and one of the most important articles is associated with the anti-vaccination problem.The use of social media by adherents of this theory and the lack of active participation by related organizations and specialists is causing the growth of this belief in public opinion.There are, of course, positive use cases such as checking the media to identify disasters or help the affected people in emergency situations (e.g.Widener et al., 2013).In the following, we will look at the literature search method and look for the selected articles in the Scopus database, most of the issues that have been addressed and explain the results of our findings.

Literature search method
To do our research, we acted in two ways:  Manual method: review abstracts of most cited articles,  software bibliometrics analysis: export articles information and import into analysis software.

Search based on keywords
First, in the Scopus database, the study searched all the articles that contained social media, and decision making in their abstracts, titles and keywords.But by reviewing the abstracts, we found that the number of irrelevant articles in this methodology is high.Also, there is a limitation on the Scopus database which is the maximum export size of BibTeX format (needs for bibliometrix) is 2,000 articles.So in both method, we searched articles using "social media" AND "decision making" in keywords.A total of 1,159 articles were found.After sorting them based on the most cited articles, we reviewed 150 first articles manually and exported all articles' information in BibTeX format in order to use in bibliometrix package.

Software analysis package
The bibliometrix R-package (http://www.bibliometrix.org)provides a set of tools for quantitative research in bibliometrics and scientometrics.It is written in the R language, which is an open-source environment and ecosystem.The existence of substantial, effective statistical algorithms, access to highquality numerical routines, and integrated data visualization tools are perhaps the strongest qualities that distinguishes R from other languages for scientific computation (Aria & Cuccurullo, 2017).Data were retrieved from Scopus web site using BibTeX format which is recognizable by Bibliometrix.Data were analyzed by using R studio v. 1.1.456, R v.3.5.1 (2018-07-02) and bibliometrix R-package (http://www.bibliometrix.org)(Aria & Cuccurullo, 2017).We have generated graphs and other information using biblioshiny.

Literature synthesis
In a short list of reviewed articles, various themes were identified based on their similarity.In Table 1 we have identified relevant themes of each article using the manual method.The first highest cited article reviews the effect of social media and collaboration in disasters and how it helps to collaborate and knowledge sharing.It is cited by 412 articles (Yates & Paquette, 2011).The second article is also related to disaster management.It describes the advantages and disadvantages of using social media for disaster relief coordination (Gao et al., 2011).Third place in rank belongs to the marketing field.It tries to tell us how can we use social media and extract information to make marketing decisions.This study has been cited by 271 articles (He et al., 2013).In Table 1 we can go through all of 150 reviewed articles.Table 1 shows 53 articles in the health field.From this point of view healthcare is the most important theme which is well developed.

Bibliometrics analysis results
In our survey, 1,159 articles retrieved from Scopus shows an average of 2.98 authors per article with collaboration index of 3.5.Average citation per documents is 7.194.These articles come from 733 sources and 926 articles were published by multiple authors.Fig. 1 shows that research and papers began in 2008 and have had an uptrend till 2018, reaching 241 articles in 2018, only declining slightly in 2017 and returning to the number of articles in 2015.

Fig. 1. Annual Scientific Production
Also, the journals that contained the most studied articles, containing "decision-making" and "socialmedia" keywords, were presented in Table 3, with the largest number of articles related to the LECTURE NOTES IN COMPUTER SCIENCE journal with 40 papers.

Table 3
Most productive sources On the other hand, Table 3 shows the core sources or focus of articles in journals, with only first 18 journals publishing nearly 20% of the total articles.According to Fig. 2, the maximum H-Index of the journals was 16, which is associated with the SCIENCE Journal, and 2 journals have H-Index 11, and other journals have H-Index 8 and lower.

Fig. 2. Sources' impact
A review in the first five journals, with the highest number of "decision-making" and "social media" subjects' articles, suggests that the number of related articles has been steadily rising and has peaked in 2015, 2016, and 2017 years, and is declining thereafter.

Fig. 5. Author scientific productivity
As shown in Fig. 6, the authors' H-index was maximum 4.  The authors of the studied articles are more affiliated with UNIVERSITY OF OTTAWA with 20 articles and then, respectively, in accordance with Table 4.
Fig. 7 shows the number of articles produced by the authors of different countries and the rate of cooperation of each country's authors with other countries' authors.For instance, authors of the United States have produced 286 articles, but the rate of American authorship co-authorship with other countries is about 10%.Subsequently, the authors of the UK ranked second with 62 papers, and the authorship rate for contributing articles to other authors with other countries is 27.4%.

Fig. 7. Corresponding author's country
Table 5 shows the total number of citations referenced to articles and the average citation of articles produced by the authors of each country.For example, 286 articles produced by American writers totaled 2691 citations and received an average of 9.4 citations per article.From the series of articles studied, Yated et al. (2011) with 410 citations ranked first.This article has an average of 51.2 citations per year.In total, it can be said that 11 articles have had more than 100 total citations, and other articles have received fewer than 100 citations.The statistical status of some of the articles with the most citations is presented in Table 5.As shown in Fig. 8, the total number of citations to articles related to the subject gradually increased over time from the 1960s, and since 2000 the slope of the growth has been markedly increased and could reach its peak in 2011-2013, but after it has fallen quickly.

Fig. 8. Reference publication year spectroscopy
Fig. 9 shows the highest number of repetitive words in the articles studied after "social media" and "decision making" with the sequence of words "human", "humans", "social networking (online)", "gender", "internet", etc.

Fig. 9. Highest number of repetitive words in articles
The repetition trend of each these words over time suggests that almost all of these repetitive words will be maximized in 2016, after which they will be downtrend, according to Fig. 10.

Conceptual Structure, Factorial Analysis
Based on the results of factor analysis, 3 major clusters have been identified, each containing close and related keywords used in our studies.Fig. 11 shows the results of this analysis.On the other hand, the clusters of "social media", "ethics", "tends" have preserved a low density and centrality, in other words, they are emerging or neglected.

Intellectual Structure
A citation network survey shows that the article Asur (2010) has received the most centrality and citation relationship with other documents followed by Golder (2011).Table 6 shows the number of co-citation between the documents and the articles studied.

Conclusion
This study has tried to provide a comprehensive view of scientific papers between 2008 and the first month of 2019 in social media and decision-making fields.This research has shown the United States, United Kingdom, and Australia have been the most productive countries in this area.The thematic map has identified that clusters of "human", "priority journal", "interpersonal communication" have been well developed.On the other hand, the clusters of "social media", "ethics", "tends" have been emerged or neglected.The result of this research has shown "Health" and "Disaster/Crisis Management" are popular among scientists.We hope the present study may help scholars identify gaps in their researches.

Fig. 13
Fig.13shows that most collaborations have been co-authored by authors from the United States to four countries in the United Kingdom, China, Canada, and Australia.

Table 1
Most cited articles and their relevant themes

Table 2
Articles statistics

Table 4
Top affiliations

Table 5
Total number of citations referenced to articles

Table 5
Most cited papers

Table 6
Intellectual Structure