JEL Classification M31
In this era of the Internet, marketers clearly recognize that online user reviews are vitally important for spreading the word about their products. Among a variety of products and services, word of mouth (WOM) is highly effective with experiential products, including movies (Tsao, 2014). As consumers cannot try the product before purchase and return it after purchase, even if they do not like it, reviews and ratings serve as great references for them to reduce uncertainty (Bae and Lee, 2011; Robinson, Goh, and Zhang, 2012). Potential moviegoers used to only refer to professional reviews and ratings published on traditional media such as newspapers and magazines. But now, nonprofessional reviews and ratings can be referred to on online media as well. According to previous studies, nonprofessional reviews influence moviegoers more than professional reviews when they choose movies to see (Desai and Basuroy, 2005; Niraj and Singh, 2015; Tsao, 2014). On the other hand, it has been revealed that review texts have a greater impact on box office performance than ratings (Duan, Gu, and Whinston, 2008; Gu, Park, and Konana, 2012; Lee, Jung, and Park, 2017). This means review texts written by nonprofessional reviewers influence potential moviegoers the most. However, most of the extant studies have paid little attention to review quality. Some studies from various countries—such as the US, Taiwan, and South Korea—have investigated the kind of reviews effective to consumers. For example, review valence is not so important for American consumers (Kim, Park, and Park, 2013), while negative reviews influence consumers the most when choosing movies in Taiwan (Tsao, 2014). These differences can be derived from the uncertainty-avoidance (UA) tendencies of the country where each study took place. Following Hofstede’s (1984, 2001) theory, countries can be scored according to UA tendencies. The higher the score, the higher UA tendency the country has. Considering America’s UA score is 46 and Taiwan’s is 69, the reasons for the different results can be understood. American consumers do not care whether reviews are positive or negative based on their low UA nature. Only review frequency is effective for American moviegoers (Kim et al., 2013). As Godes and Mayzkin (2004) suggested, American moviegoers may choose a movie they often see discussed in the media. Meanwhile, consumers in Taiwan seem to prefer checking possible risks before purchase based on their medium–high UA nature. Taiwan’s potential moviegoers may avoid a movie with negative WOM, which they cannot ignore.
Japan is widely known for frequent UA behavior—that is, residents are highly sensitive to the perceived risk of experiential products (Robinson, Goh and Zhang, 2012). Japan’s UA score (92) is much higher than Taiwan’s. It is possible that different results will be found in a high UA country from those in lower UA countries. Therefore, this study examined text selections from online movie reviews that appeared on a movie website in Japan and analyzed them using a text-mining method. The Yahoo! Japan Movies website’s reviews were chosen for the study because this website provides a usefulness counter for each review. Former studies did not remove useful reviews from useless reviews, possibly because most movie review sites do not have usefulness counters. However, useless reviews should be excluded from the analysis, as these reviews are obviously not effective to consumers. This study reveals differences between useful reviews and useless reviews, and the kinds of topics or expressions in reviews that can draw risk-avoiders to movie theaters are discussed. The results of this study will practically contribute knowledge to practitioners from the film industry in high UA countries and theoretically contribute to the field of cross-cultural communication.
2. Literature Review
Research on online user reviews is a growing trend among scholars in marketing and consumer related fields, and the bulk of academic articles in published in recent years. There are mainly three directions of research streams: 1) the influence of reviews on consumers making purchasing decisions from the perspective of WOM effects (e.g., Eslami and Ghasemaghaei, 2018; Hu, Pavlou, and Zhang, 2017; Liu and Karahanna, 2017), 2) the product providers’ responses to reviews (e.g., Eslami, Ghasemaghaei, and Hassanein, 2018; Jia, 2018; Liu and Dong, 2018; Tang, 2017; Wang and Chaudhry, 2018; Zhou et al., 2018), and 3) how to detect and deal with untruthful/fake reviews (e.g., Clare, Wright, Sandiford, and Caceres, 2018; Moriuchi, 2018; Zhang, Zhou, Kehoe, and Kilic, 2016). Some of the articles focus on specific e-commerce sites, such as Amazon.com, or reviews for specific products/services—including restaurants, hotels, and entertainment shows (e.g. Jia, 2018; Kim and Im, 2018; Lee, Trimi and Yan, 2018; Roozen and Raedts, 2018). Many researchers choose online movie reviews for their study because consumers strongly rely on reviews due to the uncertainty nature of experiential products. For example, Niraj and Singh (2015) investigated how consumer reviews and critical reviews influence moviegoers when choosing a movie as well as evaluating the movie after seeing it. It was found that consumers generally consider customer reviews are more important to them than professional critics’ reviews as references. On the other hand, it is indicated that movie ratings do not have a significant impact on box office performance (Duan et al., 2008; Gu et al., 2012; Lee et al., 2017). This means review texts written by nonprofessional reviewers influence potential moviegoers the most. However, only a few researchers focus on the quality of nonprofessional review texts.
Kim et al. (2013) examined the impact of both online user reviews and expert reviews on box office figures in the US domestic market. Using a dataset of 169 movies released in 2008 in the US, the researchers found that, in the US, the frequency of online WOM and the valence ratings of expert reviews have positive impacts on moviegoers regardless of the reviews’ contents. It seems consumers in the US do not care about the review sentiment. They are more favorable to the movies they have heard of, even if the tone of WOM is mainly negative.
Meanwhile, Tsao (2014) investigated the influence of consumer expectation and online user reviews on moviegoers’ movie selection and evaluation in Taiwan. As a result, it was revealed that moviegoers in Taiwan are influenced by negative reviews rather than positive reviews when they select a film, while positive reviews are more influential than negative reviews when they evaluate the film after seeing it. Consumers in Taiwan may not be willing to take risks, so they may calmly estimate the possibility of wasting money by referring to negative reviews.
Lee et al.’s (2017) study, in Korea, focused on the review text sentiment to reveal the relationship between online reviews and sales. Analyzing the entropy of the review text sentiment, they concluded Korean consumers are more likely to trust WOM information when they equally encounter positive, negative, and neutral review texts. This comes about because consumers consider it is natural that people have different preferences regarding the same movie; thus, there should be various types of the review text sentiments about a movie. Korean consumers are usually suspicious of reviews’ credibility and will not simply accept reviewers’ comments about a movie.
As referred to above, these former studies found different results. It can be considered that it depends on the consumers’ cultural background. This study follows Hofstede’s (1984) theory to explain these differences.
3. Hofstede’s Cultural Dimensions Theory
Developed by Dutch social psychologist Geert Hofstede, the cultural dimensions theory is commonly used in the literature to understand different cultures and advance cross-cultural communication. By using factor analysis, Hofstede found six values that affect a society’s culture and relate to its members’ behaviors: 1) the power distance index, 2) individualism, 3) the uncertainty avoidance (UA) index, 4) masculinity, 5) long-term orientation, and 6) indulgence versus restraint (Hofstede, 1984, 2001). Numerous researchers and practitioners used his theory for their studies or practices covering various fields, such as cross-cultural issues and international business matters. However, many arguments and criticisms have also been presented against Hofstede’s work. For example, many researchers mentioned that surveys, the method that Hofstede chose to measure cultural differences, are not appropriate for this kind of study (Schwartz, 1999). Other researchers stated that cultures cannot be divided exactly by borders of countries (Jones, 2007) or many countries have multiple ethnic groups domestically (Nasif et al., 1991). Hofstede’s theory itself has also been considered as obsolete in relation to recent global society (Jones, 2007). Although such arguments are understandable and persuasive in a sense, many marketing researchers continued to explain consumers’ purchase intentions or advertising preferences using the concept of UA index from Hofstede’s cultural dimensions theory (e.g., Hoeken et al., 2003; Kim et al., 2017; Mori et al., 2010). Thus, the current study also uses the UA index to analyze the consumers’ online reviews.
UA is the extent to which people prefer to avoid uncertainty or ambiguity rather than accept risks and enjoy adventure. If the UA score is low, the society is more flexible and its members are more relaxed and easy-going, even in new and unknown situations. If the UA score is high, the society is stricter and its members prefer following plans and keeping to time, not willing to jump into new situations without knowing anything.
According to Hofstede (1984, 2001), the UA score of the US is 46. The US can be described as a lower UA country. Thus, it seems natural that American consumers do not care about valence or sentiments of review texts and are not as suspicious of reviews’ credibility.
Meanwhile, Taiwan can be categorized as a moderate UA country because its UA score is 69.
As consumers in Taiwan need to know the risk in advance, they focus more on negative reviews than positive reviews based on their moderate UA nature.
Korea’s UA score (85) is higher than Taiwan’s. As there are so many fake reviews and stealth marketing advertisements on the Internet, Korean consumers are basically suspicious of reviews’ credibility. They never want to be cheated by setups; thus, they may look for the evidence of truly believable reviews—for instance, the variation of review sentiments.
As described above, consumers from higher UA countries more strictly assess the quality of review texts to rely on them as references than consumers from lower UA countries. Then what kind of reviews are effective for consumers from extremely high UA scored countries? The 10 highest-ranked UA score countries are shown in Table 1. These countries are European Union member states or countries located in Central/South America, except Japan. This study focuses on Japan as one of the highest UA countries.
Table 1. 10 highest-ranked UA countries referred to Hofstede (2001)
|5 El Salvador||94|
|10 Costa Rica||86|
Movie consumers in Japan use the movie-information website Yahoo! Japan Movies (movies.yahoo.co.jp)—which includes usefulness counters with reviews, as shown in Figure 1. Not all online reviews are useful for consumers naturally. Whether a review is truly influential or not so influential is dependent on the quality of the review text. Lee et al. (2018) realized the importance of such usefulness of reviews and examined one of the biggest e-commerce sites Amazon.com, which has the usefulness counter on the review pages. However, the usefulness of movie reviews has never been focused on by researchers, probably because there are only a few movie review sites with usefulness counters. Hence, it is reasonable to use online reviews published on Yahoo! Japan Movies as data for analysis.
Among more than 63,000 movies reviewed on Yahoo! Japan Movies, two Japanese movies released in 2016 were examined for this study: (A) Your Name (Kimi no nawa, an anime movie) and (B) Godzilla Resurgence (Shin Godzilla, a monster thriller). These movies were two of the top three box office draws in Japan in 2016. In addition, reviews of these movies drew the most readers ever on Yahoo! Japan Movies, and the reviews scored the highest “useful” count totals ever (as of April 2020). Reviews with 100 or more useful counts (described as “useful reviews”) and reviews with no useful counts (described as “useless reviews”) from each movie were collected for analysis. Although the usefulness counter changes moment by moment and it is possible that some reviews with no useful counts could get a useful count just after data collection, the two movies selected for this study were released in 2016 and both movies’ road shows had already been closed by 2018. Not so many moviegoers read reviews of movies no longer available in theaters. Even if some reviews with no useful counts get useful counts after the data collection, it should be just a few. Those reviews still can be categorized as useless reviews, compared to reviews with 100 or more useful counts.
On the other hand, newly written reviews should naturally have no useful counts, as few people read them. Although not so many people may read them afterward, as the road shows were already over, this study collected reviews written until the end of August 2018 and still obtained no useful counts as of the end of 2019.
Eventually, this study found 254 useful reviews (2.9%) and 1,201 useless reviews (13.7%) among 8,796 overall reviews of movie (A) and 353 useful reviews (6.2%) and 564 useless reviews (9.9%) among 5,684 overall reviews of movie (B) on Yahoo! Japan Movies. The total number of useful counts were 171,808 for movie (A) and 122,977 for movie (B). Each movie’s useful reviews got 88,540 (Movie[A]) / 70,161 (Movie [B]) useful counts; thus, more than 50% of the overall useful counts were earned by only 3% or 6% of the reviews (see Table 2).
After careful readings, words of review texts were extracted with Tiny Text Miner (Matsumura and Miura, 2009) and evaluated using the statistical software R. Based on Tsao’s (2014) study, this study selected 36 nouns and adjectives regarding 1) a plot of the movie, 2) staff and actors, 3) personal views, and 4) recommendations/evaluations among frequently used words in reviews of both movies for comparison purposes. Verbs were excluded because many of the most frequently used verbs—such as do, be, become, and can—did not make sense alone. Hence, word clouds were created with them.
Figure 1. Review format of Yahoo! Japan Movies
Table 2. Basic data on selected two movies
|Movie Title||Box office sales (currency: JPY)||Total number of reviews||Total number of useful counts||Number of reviews with 100 and up useful counts and their total number of useful counts||Number of reviews with no useful counts|
|A. Your Name (Kimi no nawa.)||250.3 million||8796||171808||254 (2.9%) 88540 useful (51.5%)||1201 (13.7%)|
|B. Godzilla Resurgence (Shin Godzilla)||82.5 million||5684||122408||353 (2.9%) 70161 useful (51.5%)||564(10%)|
Note: As of August 31, 2018
As shown in Table 3, the average length of useful reviews is more than four times longer than that of useless reviews for both movies. Furthermore, useful reviews have more variations of words than useless reviews. Although word variations of movie (A)’s useful reviews and useless reviews seem close when compared to movie (B)’s, movie (A) has about 5 times more useless reviews than useful reviews, while Movie (B)’s useless reviews were only 1.6 times more than useful reviews. Thus, useful reviews of movie (A) have a rich variety of words.
The characteristics of each movies’ reviews are described below.
Table 3. Basic data on selected two movies
|Movie||Useful or useless||Number of reviews||Average length of reviews (number of letters)||Nouns||Verbs||Adjectives|
5.1.(A) Your Name.
Surprisingly, many useful reviews did not seem to criticize the movie but, instead, criticized other reviews that gave the movie with an opposite rating. It seems the reviewers just wanted to write their own opinions on other reviews and not on the movie itself. For example, one useful review read, “Most low-rated reviews’ texts were basically poor and childish, saying things like ‘I just don’t like this film.’ I was disappointed while reading those reviews.” Another useful review more straightforwardly read, “Please excuse my review because it’s my comments on other reviewers who rated this movie with a low score, rather than a movie review.” As many reviews gave the movie low ratings, the reviewers who loved this movie tried to protect the movie’s reputation and gained more than 100 useful counts. Some useful reviews strongly criticized low-rated reviews. There are also useful reviews that criticized high-rated reviews. For example, one review read, “I’m one of the reviewers who were too honest to speak highly of this movie. I don’t understand why so many people were moved by the film with this kind of quality.” Such arguments can be seen so often in that movie’s review pages. Few useless reviews mentioned other reviews.
In comparison, useless reviews’ descriptions were much more general and simple than those of useful reviews: “It was good,” “Very interesting,” and “I enjoyed it from the beginning to the end.” Yet, even if their review texts positively described the movie, consumers seemed to ignore such reviews, possibly because of the boring and common expressions with which they were presented.
Figure 2 below provides word clouds for useful reviews and useless reviews of this film. The words review and evaluation in the useful reviews’ word cloud are much bigger than those in the useless reviews’ word cloud. In other words, many authors of useful reviews wrote the words review and evaluation in their review texts. These words can be used to refer to other authors’ reviews.
Meanwhile, an adjective, good, is much bigger than other words in the useless reviews’ word cloud. That is, several useless reviews used a very easy adjective: good. Although good is prominently featured in the useful reviews’ word cloud, it is not significantly bigger than other words. Furthermore, other simple adjectives, such as interesting, wonderful, and great, are bigger in the useless reviews’ word cloud than those in the useful reviews’ word cloud. Hence, Japanese potential moviegoers may not be interested in such impressions of reviewers even though they positively assess the movie.
Tsao (2014) suggested that useful reviews should comment on the characters, the storyline, and the quality of the acting. However, as seen in Figure 2, the words storyline and acting are obviously bigger in the word cloud of useless reviews than that of useful reviews. Moreover, the word characters is almost the same size in the useful and the useless reviews’ word clouds. This may be because the movie (A) is an anime film, while the two movies Tsao (2014) analyzed are both live-action films. It is possible that potential consumers of anime films are interested in different topics compared to the potential consumers of live-action films.
Meanwhile, more technical words, such as screenplay, scenes, and theme, are bigger in the useful reviews’ word cloud than those in the useless reviews’ word cloud. Those words are supposed to arise from film enthusiasts’ perspectives. Anime fans often mention the director’s style and explain it with their deep knowledge about anime. It may be that Japanese consumers may want to read such “otaku” (geek) commentary and not just a summary of the movie. Anime otaku, grown-ups who loved anime shows, were considered weird people in Japan until the early 2000’s because they patronized movies usually made for teenagers or lower-aged children (Harada, 2015, p.18). Today, however, the otaku are now widely thought of as cultural trend leaders (Harada, 2015, p.32). Thus, reviews with otaku perspectives may be considered useful for consumers, or consumers may think that reviews made by the otaku are simply interesting as articles even if they present a highly negative assessment of the movie.
Describe the procedures for selecting participants, how you attracted your data and who are your subjects.
Figure 2. Word clouds for the movie Your Name.
5.2.(B) Godzilla Resurgence
Godzilla fans, or monster-movie fans in general, seem to have written many useful reviews that strongly recommended this movie. Like movie (A), this movie’s useful reviews also mentioned other reviews but not as critically as the few low-rated reviews that were written. Reviewers basically respected other reviews, and some just added their opinions to previous reviews. One author even started his/her review by writing these words: “Many excellent reviews for this movie already exist, so I will just write about two things regarding this movie that impressed me.”
As shown in Figure 3, the words review and evaluation in the useful reviews’ word cloud are bigger than those in the useless reviews’ word cloud. In addition, the words ending, scenes, and story in the useful reviews’ word cloud are much bigger than those in the useless reviews’ word cloud. Useful reviews also use the words director and direction more frequently than the useless reviews do. Useful reviews’ authors may have tried to explain this revolutionary new concept (the Godzilla movie) with its storyline or impressive scenes as well as the director’s style who created it. As this movie was created by a famous anime director, Hideaki Anno, who also created “Neon Genesis EVANGELION,” an iconic Japanese TV anime series broadcasted in the 1990’s, many useful reviews indicated Anno’s talent to make this live-action monster panic movie very realistic. Such reviews cannot be written if the reviewer does not have a comprehensive knowledge about anime.
Moreover, the words storyline, acting, and characters are bigger in the useful reviews’ word cloud than those in the useless reviews’ word cloud, unlike movie (A)’s word clouds. Hence, the suggestion this study stated previously may be true, that is, consumers tend to focus on different points depending on the form of the movie: live-action or anime.
In addition, in the useful reviews, so many reviewers mentioned their generation’s Godzilla films in their texts. Godzilla was born in 1954. By the end of 2019, a total of 29 films based on the character have been produced in Japan, including movie (B) in 2016. Thus, several Godzilla fans had their own notions of memorable Godzilla film(s) and compared it/them with movie (B). Furthermore, there seemed to be so much of “Godzilla otaku” on movie (B)’s review pages. Such reviewers exhibited their deep knowledge of Godzilla films and gained several useful counts. This is very similar with movie (A)’s useful review, which were written by anime otaku.
Meanwhile, useless reviews often described simple feelings, just like the useless reviews of movie (A): good and interesting were the first and second most frequently used words in useless reviews, as indicated in useless reviews’ word cloud. Furthermore, useless reviews used the word disappointment much more than the useful reviews did. Potential moviegoers in Japan seem to ignore negative and positive expressions while potential moviegoers in Taiwan focus on negative reviews when they choose a movie to see (Tsao, 2014). In other words, Japanese consumers are not interested in the reviewer’s personal feelings about the movie—whether it is positive or negative. They simply want to read enthusiastic commentaries based on otaku knowledge or the authors’ personal experiences. Moreover, discussion among reviewers in the review site may be considered worth reading for consumers for fun or for reference. It can be assumed \that useful review authors themselves enjoy reading other authors’ reviews, so they cannot help but mention other reviews.
Figure 3. Word clouds for the movie Godzilla Resurgence
6. Discussion and Conclusion
This research demonstrates three principal characteristics that are common to useful reviews: a) a longer text (i.e., useful reviews are four times longer than useless reviews, on average), b) a high level of verbal expressiveness (i.e., the movie descriptions in useful review texts are rich, whereas those in useless reviews are basically poor and full of hackneyed words and phrases), and c) references to other reviews (i.e., useful reviews refer more or less to other authors’ reviews). Considering the high level of Japan’s UA tendencies, longer reviews should be more favorable to potential moviegoers in Japan than shorter ones, because longer reviews can simply be considered more informative. Likewise, rich text descriptions should be more favorable to Japanese customers than simple text descriptions, as the Japanese audience can imagine the quality of the movie more easily with a good text than with a poor text.
Potential moviegoers, however, may not view a review as useful just based on length and verbal expressiveness. Furthermore, in many cases, useful reviews criticize or make references to other reviews based on authors’ geeky knowledge or personal experiences. This could mean that people may simply enjoy reading about disagreements between reviewers. Such “review battles” arouse readers’ interest and influence how they evaluate a movie, depending on whether the authors’ reviews of the film were positive or negative. Analyzing the data from the viewpoint of Hofstede’s cultural dimensions theory, Japanese consumers should be more careful about risk in accordance with their high UA score. Japanese moviegoers, however, seem to be willing to gamble with the experience when they are intrigued by controversial reviews. Rather than those of Taiwanese or Korean moviegoers, it seems the characteristics of Japan’s potential moviegoers are similar to those of American moviegoers, although the UA scores of Taiwan and Korea are much higher than that of the US and closer to Japan’s.
Some former researchers have claimed that Hofstede’s study is outdated and has no modern value (Jones, 2007). Hofstede refuted these, because culture cannot be changed so easily (Hofstede, 1998). Hofstede’s opinion should be right. However, the current study suggests that the different culture may exist in the online world apart from the real world. Normally, Japanese people do not voice their opinions clearly and prefer to use vague expressions (Hayashi and Hayashi, 1995) based on their high context communication culture (Hall, 1976). However, in the online world, opinions with strong expressions—sometimes too extreme—can be seen very often, even in Japan, just as in the US (Yamaguchi, 2018). Although Japanese people have high UA tendencies in the real world, they may have lower UA tendencies in the online world. Possibly, two kinds of cultures may exist in a country: the historical culture and the emerging online culture. According to the research by Trend Lab. (2013), more than 40% of Japanese people answered that they use different personalities between the real life and the social networking services. Assuming that people have different characteristics on the Internet, it is understandable that Japanese potential moviegoers would prefer clear descriptions of whether or not the author liked the movie and if he/she supports or is against other reviews. Such obvious arguments or discussions could have attracted Japanese consumers and made them consider seeing the movie to confirm the reviews. In other words, the purchase intentions of Japanese moviegoers are derived from moviegoers’ desire to evaluate a movie themselves, not just to enjoy watching it. Some people write review articles on the movie review site after watching a movie and express their own opinions for or against former review articles. Those people may be motivated to see the movie and join the community of the review pages, whose reviews are reviewing other reviews. On the one hand, reviews for movie (A) often used the phrase in my own way, such as “Here, I discuss this movie in my own way to counter negative reviews” or “Some reviewers still negatively evaluated this movie and said they couldn’t understand the story very well. So I try to interpret the essence of this masterpiece in my own way.” On the other hand, reviews for movie (B) often described the reviewer’s personal experience and preferences regarding Godzilla films. One review read, “Since other reviewers have already fully reviewed this movie, I will introduce my personal experiences, preference, and topics, which have been rarely mentioned in former reviews based on this film’s concept.”
If Japanese consumers’ purpose of watching a movie is to evaluate it and occasionally to join the community of reviewers, a movie that has several long and expressive reviews assessing other reviews can be a correct choice for Japanese consumers. Even when the movie was not fun, consumers can at least evaluate it themselves and even post a review to support other negative reviews. In that sense, Japanese consumers can already avoid the risk of wasting money when they find a movie with many useful reviews.
6.1. Limitations and Future Research
This study analyzed online movie reviews for two movies that were certified box office hits in 2016 in Japan. Then, it gathered the most and second most popular reviews on Yahoo! Japan Movies website, which were still the most and second most popular as of April 2020. Through the analysis, certain trends among useful reviews were found, and Japanese consumers’ movie selection intentions were discussed. However, more movie reviews should be analyzed to ensure the study’s validity. Box office successes and actively reviewed movies, as well as different kinds of movies, such as movie (A) and movie (B) (e.g., Hollywood movies), are appropriate for future studies.
Furthermore, if it is true that people have two different cultures: one in the real world and another on the Internet, Hofstede’s theory should be verified by studying people in the online world. Such a study may find that countries have different scores in all six dimensions presented by Hofstede. Moreover, additional dimensions can be included to characterize the cultures. For instance, the level of different personality usages could be scored to understand cross-online cultural differences. Through the Coronavirus experience, human activities are rapidly shifting to online platforms, including working, studying, and even social drinking. This phenomenon can continue for a while even if the pandemic has ended. In relation to this, multi-nation studies of online cultures will be increasingly necessary to prepare for the day when people would live in the online world rather than in the real world.
- Balagué, C. and Fayon, D., 2010. Facebook, Twitter et les autres... Paris, France: Pearson Education.
- Barnes, J., 1954. Class and committees in a Norwegian Island Parish. Human Relations, 7, pp.39-58.
- Boyd, D.M. and Ellison, N.B., 2007. Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication, 13(1), pp.210-230.
- Briard, E. and Bontemps, A., 2011. Entrez dans la toile de mon reseau. Paris, France: Edipro.
- Dagnon S., 2018. Using Chatbots for Social Media Marketing. [online] Available at: https://mavsocial.com/chatbots-social-media-marketing/. [Accessed on 27 February 2019].
- Dakouan C. and Benabdelouahed. R., 2018. Content shared on social networks: What effect on the buying intentions of Moroccan consumers?. International Journal of Scientific & Engineering Research, 9(7), pp.315-323.
- Ertel W., 2017. Introduction to Artificial Intelligence. Second Edition. London, UK: Springer
- Frankenfield J., 2018. Chatbot. [online] Available at: https://www.investopedia.com/terms/c/chatbot.asp. [Accessed on 02 March 2019].
- Graham, J. and Havlena, W., 2007. Finding the "Missing Link»: Advertising’s Impact on Word of Mouth, Web Searches, and Site Visits. Journal of Advertising Research, 47 (4), pp.427-435.
- Haugeland J., 1989. Artificial Intelligence: The Very Idea. Cambridge, Massachusetts, United States: Bradford Book, MIT Press.
- Kotler, P. Kartajaya, H. Setiawan, I. and Vandercammen, M., 2017. Marketing 4.0. Paris, France: Edition Nouveaux Horizons.
- Kreimer I., 2018. How to Get Started with AI-Powered Content Marketing. [online] Available at: https://www.singlegrain.com/artificial-intelligence/how-to-get-started-with-ai-powered-content-marketing/. [Accessed on 03 March 2019].
- Lenhart, A., 2009. Adults and Social Network Websites. [online] Available at: https://www.pewresearch.org/internet/2009/01/14/adults-and-social-network-websites/ [Accessed on 03 March 2019].
- Li, D. and Du, Y., 2017. Artificial Intelligence with Uncertainty. Second Edition. Boca Raton, USA: CRC Press, Taylor and Francis Group.
- McCarthy, J., 1955. A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. AI Magazine, 27(4),
- McCarthy, J., Minsky, M. L., Rochester, N. and Shannon, C. E., 2006. A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, August 31, 1955. AI Magazine, 27(4), p.12. doi:10.1609/aimag.v27i4.1904
- Newberry, C., 2017. Social Selling: What it is, Why You Should Care, and How to Do It Right. [online] Available at: https://blog.hootsuite.com/what-is-social-selling/ [Accessed on 01 March 2019].
- Nilsson, J., 1998. Artificial Intelligence: A New Synthesis. Burlington, Massachusetts, United States: Morgan Kaufmann Publishers.
- Rissoan, R., 2011. Les Réseaux Sociaux: Facebook, Twitter, Linkedin, Viadeo, Google+, Comprendre Et Maîtriser Ces Nouveaux Outils De Communication. Saint-Herblain, France: Eni Editions.
- Stelzner, M., 2018. Predictive Analytics: How Marketers Can Improve Future Activities. [online] Available at: https://www.socialmediaexaminer.com/predictive-analytics-how-marketers-can-improve-future-activities-chris-penn/ [Accessed on 02 March 2019].
- Tjepkema, L., 2018. Why AI is Vital for Marketing with Lindsay Tjepkema. [online] Available at: https://www.magnificent.com/magnificent-stuff/why-ai-is-vital-for-marketing-with-lindsay-tjepkema. [Accessed on 02 March 2019.]
- Ziryeb, M., 2011. Les Réseaux Sociaux Numériques D’entreprise. Paris, France: Éditions L'Harmattan.