Persuasive Recommender System is a relatively new research direction that emphasizes on using persuasive approaches to increase user's acceptance of the recommendations. Recent studies have demonstrated the feasibility of deploying the six persuasive principles of Cialdini as explanations besides the recommended items. These principles, however, should not be treated in a onesize-fits-all approach. Instead, they should be deployed in a personalized manner. The factors that help to personalize these principles for the users of recommender systems have not been fully explored. To fill this gap, this paper investigates one of these factors, which is personality traits. In particular, it explores the influence of the Big Five Personalities on recommender system users' susceptibility to Ciladini's persuasive principles. The study contains two parts; the impact of personality traits as an independent variable, and the impact of personality traits in conjunction with the application domain. We explored these factors through a questionnaire disseminated online. The analysis of the data received from (279) responses shows that personality traits are an important factor that influences the efficiency of the six principles of Cialdini. Moreover, the effect of personality traits becomes more significant if they are considered in combination with the application domain.
• Human-centered computing ~ Interaction design ~ Interaction design theory, concepts and paradigms.
We can define Persuasion as influencing people’s
behaviours or attitudes without any deception or coercion
[
At the beginning of the current decade, Yoo et al. [
There is a wide range of persuasive principles introduced
by the psychologists. Among these principles, the six
principles of Cialdini [
The rest of this paper is organized as follows: Section 2 provides a brief introduction about persuasive principles and personality traits, and it discusses the related work. Section 3 talks about the study design. Then the results are analyzed in section 4. Discussion and design guidelines are provided in section 5. Finally, section 6 concludes the paper.
In this section, we introduce the main concepts used in this work, which are persuasive principles and personality traits. Then we discuss the related work.
In the literature of social sciences, there are various
persuasive principles. For instance, the forty (40) principles
of Fogg [
Reciprocity means that people have an obligation to give back to others what they have received first or to return favours. For instance, if your friend sent you a birthday gift, then you owe that friend a future gift. This principle is used in many computerized systems where they give new users a gift (such as a voucher or a free service for a limited time). The second principle, Scarcity, states that people want more of scarce things, and they consider these things as more valuable. Displaying numbers of availability for products is an example of implementing this principle. eCommerce websites (e.g., Amazon) use this principle by providing a limited-time-only promotion, which is often presented as a countdown timer showing the time remaining before the offer expires. The next principle, Authority, indicates that people are more inclined to follow others who have legitimate authority. For instance, it is more likely to give change for a parking meter to a stranger if she wears a uniform rather than casual clothes, and it is more likely to buy a toothpaste if a well-known dentist recommends it. The fourth principle, Social Proof, means that people tend to do what others do. When people are uncertain, they look to the actions of others to decide. Widespread deployment of this principle is checking the reviews of others. For instance, when people want to book a reservation in a resort, they usually check out the reviews of that resort. This principle is widely implemented in the computerized system, and it is implemented in different ways, such as showing ratings, emphasizing the number of followers or fans, or presenting testimonials. The fifth principle is called Liking, and it states that people are most likely to accept the request made by those that they like. For instance, people may prefer one store over the other only because they like the employees in that store. In online communication, you also need your customers (or users) to enjoy your service. So, the service or product should be presented attractively. The last principle is Commitment. It indicates that people tend to be consistent with the things they have previously said or done. The basis of this principle is that if a person committed to do small requests, it will be easier to persuade them to do larger requests. Commitment is implemented online by different means, such as asking the customer to test a new feature in your application for free and to write a review about it. If the customer used and wrote a good review of the service, then she is more likely to continue using this service as a kind of commitment.
These principles provide approaches that cause one person to say yes to another one. That is, appropriately implementing these principles can increase the acceptance of your requests, or your products.
Psychologists have extensively studied humans’
personalities and their characteristics. The Big Five Model
(a.k.a the Five-Factor Model, FFM) [
Original, and Wide interest. • Conscientiousness: Efficient, Organized, Planful, Reliable, Responsible, and Thorough. • Extraversion: Active, Assertive, Energetic, Enthusiastic,
Outgoing, and Talkative. • Agreeableness: Appreciative, Forgiving, Generous, Kind,
Sympathetic, and Trusting. • Neuroticism: Anxious, Self-Pitying, Tense, Touchy,
Unstable, and Worrying.
These five factors are known as relatively stable; they are
stable throughout individuals’ lives, with some slight
exceptions. A study by Soto & John [
This section highlights the most recent studies related to our study. Thus, the section does not focus on how to incorporate persuasive capabilities to RSs. Instead, it discusses the works that investigate the relationship between the six weapons of influence on one side, and the big five personalities and the application domain on the other side.
Some researchers discussed the susceptibility of different
groups to the Cialdini’s principles. For instance, Oyibo et
al. [
Up to our knowledge, the feasibility of using persuasive
statements along with the recommendations has been
introduced by Gkika and Lekakos [
As a response to their call, Alslaity and Tran investigated
the effect of application domain on the susceptibility of
RS’s users to the six weapons of influence [
Despite the existence of some research in the area of personality and persuasiveness in RS, the literature still has the following limitations: first, most of the work focus on a single factor (mainly users’ characteristics) while ignoring other important factors. Second, they discuss a single factor in isolation of other factors that may have a significant effect if combined (i.e., if the interaction effect is considered). Our work aims to fill this gap by investigating the influence of personality traits and the application domain of RSs on the performance of the six principles of Cialdini. The next section presents the design of this study.
Our study is based on a questionnaire that consists of two
main sections: personality test, and persuasion test. In the
personality test, we deployed the Big Five Inventory (BFI),
which is one of the most popular questionnaires for the
FFM. It was introduced in the late 1980s [
It is noteworthy to mention that we adopted the persuasive
explanations designed by Sofia et al. [
eComm
A friend of you, who bought the item that you suggested to him/her in the past, recommends you this item! A Facebook friend, who saw the movie that you suggested to him/her in the past, recommends this movie! Giving you something for free (e.g., samples, gift, or free delivery) The recommended item will be available for two months only! The recommended movie will be available for two months only! Display a countdown, beside an item, indicating the time remaining for an offer on that item The recommended item won 3 prizes as the best-manufactured product! The recommended movie won 3 Oscars! Presenting an image of an expert uses the recommended item (ex: a doctor suggests particular exercises for his/her patients, or a security guard uses the recommended security lock) 87% of users rated the recommended item with 4 or 5 stars! 87% of users rated the recommended movie with 4 or 5 stars! Presenting the “best sellers” or the “most watched” items.
This item belongs to the kind of items you usually buy. This movie belongs to the kind of movies you enjoy watching.
Using “add to wish list” option.
in [
Each cue is followed by a seven-point Likert scale. The seven scales are distributed as follows: 1 to 5 options are scaled from very low to very high effect. Zero (0) indicates no effect, and (-1) means a negative impact. The users were asked to give a rating for each cue. The rating reflects the impact of the cue on their acceptance of a recommendation. Particularly, the participants were asked to imagine that they use an RS. Then they need to rate each cue based on the question: “To what extent do you think that each of the following statements will influence your decision to buy an item (or to watch a movie) recommended to you?” We built and disseminated the questionnaire online using the “SurveyMonkey1” website. We followed a withinsubject design, where we asked each participant to complete all parts of the questionnaire. The participants were recruited through different means, including paper posters posted and online channels (such as emails and social media). We received 329 responses. After filtering the responses by removing incomplete records, we retained a total of 279 responses. Table 2 summarizes the demographic information of the participants.
(count, percentage) 16-25 (53, 19%), 26-35 (129, 46%), 36-45 (59, 21%), 46+ (38, 14%) Male (177, 63%), Female (98, 35%), preferred not to mention (4, 2%) Asia (75, 27%), Europe (13, 5%), North America (186, 67%), South America (2, 1%), Oceania (3, 1%)
This section discusses the results of our study. It is divided into two subsections according to our research questions; first, it discusses the effect of personality traits in isolation of other factors. Then, it examines the interaction between personality traits and the application domain.
For the results significance test, we deployed the Analysis Of Variance (ANOVA), which is a set of statistical models. ANOVA analysis is mainly used to investigate the differences between groups in a sample. For ANOVA analysis, the persuasive principles are considered as the dependent (or within-subject) variable, while the other two factors (i.e., Personality traits and application domain) are included as independent variables. The significance level (α) is set to be (0.05) for all ANOVA analysis.
The reliability of the results was measured by McDonald’s
omega (ω) reliability test. The McDonald’s omega
reliability test is the non-parametric equivalent of
Cronbach’s alpha (ρ) reliability test [
This section answers our first research question; it discusses the relationship between personality traits of RS’s users and the six persuasive principles. Also, it shows which persuasive principle is more effective for each personality trait.
Figure 1 depicts the average ratings for the persuasive principles grouped based on personality traits. The figure shows that all means are larger than the neutral value (Zero), which means that all personalities are susceptible to the six principles. The degree of susceptibility, however, 3.5 3 2.5 n 2 a eM1.5 1 0.5 0 Extraversion Agreeableness Conscientiousness Neuroticism Openness varies from one personality to another. The figure also shows that Agreeable people are the most susceptible to three persuasive principles, which are Reciprocity, Social Proof, and Commitment. Extrovert people are the most vulnerable to Authority and Liking principles.
The ANOVA analysis shows that the differences between the five personalities regarding the Reciprocity principle are significant [F = 4.378, P = 0.002]. A Tukey posthoc test revealed significant pairwise differences between Agreeableness and each of Conscientiousness and Openness. The test also shows that the mean rating of Agreeable people is 0.7 and 0.85 more than Conscientious and Open people, respectively, which means that Agreeableness is more vulnerable to the Reciprocity principle. Regarding the other principles, the data did not provide enough evidence that the differences are statistically significant (as illustrated in Table 4), but it shows that there are differences. This conclusion is not surprising, taking into consideration that the personality traits represent continuums; Individuals may fall anywhere in the continuum for each trait.
Table 3 depicts the persuasion profile for each personality. By persuasion profile, we mean the order of the persuasive principles. That is, a persuasion profile shows the order of the persuasive principles based on their ability to persuade the corresponding personality; Each row in Table 3 represents the persuasion profile of the corresponding personality. The principles are ordered descendingly from the most influential principle (order 1) to the least influential (order 6). For instance, the first row of Table 3 depicts the persuasion profile of the Extraversion personality. It shows that Liking is the most influential principles, followed by Reciprocity, Authority, Scarcity, Social Proof, and finally, Commitment (which is the least influential one).
As general observations, Table 3 indicates that Reciprocity and Liking are the most influential principles, where Reciprocity is more influential than Liking. On the other side, Scarcity and Commitment are the least influential for Agreeableness and Extraversion, respectively, where Authority is the least influential strategy for the remaining three personalities. Besides, the table depicts some similarities between the persuasion profiles. For instance, four out of six principles occupy the same order in each of the following pairs of personalities: Agreeableness & Conscientiousness, Conscientiousness & Neuroticism, and Conscientiousness & Openness.
F-value
P-value The results presented in this section suggest that users’ responses to the Cialdini’s principles diverge based on their personality traits. This divergence becomes more significant for users’ responses to the Reciprocity principle as the ANOVA analysis showed. The next section adds up into this section by discussing how users’ personalities affect their reactions to Cialdini’s principles in different application domains.
The previous section shows that there are some differences
in the effect of personality traits on users’ responses to the
persuasive principles. This section answers the second
research question; It adds another dimension to the analysis,
which is the persuasive context (presented by the
4
6
5
4
5
1.627
0.168
application domain). As we have mentioned above, recent
researchers have found that the application domain is an
essential factor that may affect the persuadability of the
persuasive principles [
Figure 2 depicts a comparison between users’ responses to the persuasive principles in the eCommerce domain compared to the movie domain. The figure contains five charts; each chart represents the results regarding one personality. The y-axis shows the mean values, while the x3.5 3 2.5 an 2 eM1.5 1 0.5 0 eCommerce Movie (a) 3.5 3 2.5 an 2 eM1.5 1 0.5 0 eCommerce Movie (c) eCommerce Movie (d) eCommerce Movie (e) axis shows the persuasive principles. For each principle, there are two columns; one represents the mean rating in the eCommerce domain while the other is for the movie domain.
The common observation between the five charts in Figure 2 is that the means of all persuasive principles are different. For instance, in regard to the Agreeableness personality (Figure 2-b), the means for all principles vary from one domain to the other; Reciprocity and Social Proof were rated slightly higher in the eCommerce domain, while the other four principles were rated higher in the movie domain. The only two exceptions to the observation mentioned above are related to the Extraversion personality (Figure 2-a); the figure shows that Scarcity and Authority have similar means in both domains; the other four principles have different means, though.
Table 5 depicts the persuasion profiles for each personality over both domains (eCommerce and Movie). This table has a similar structure to Table 3. The only difference is that the table is divided horizontally into five parts based on personality factors. Each of these five parts is divided into two rows that represent the eCommerce (eComm) and the movie domains. For instance, the first row shows that Social proof is the most effective strategy for Extraversion personality in the eCommerce domain, and Commitment is the most influential in the movie domain.
Several points can be inferred from Table 5; first, it shows that in most of the cases, the persuasion profiles in the eCommerce domain are different than that for the movie domain, although they are for the same personality traits. More precisely, the table shows six congruent cases only (for readability purposes, we distinguished these cases by the bolded text). Second, the persuasion profiles for Agreeableness and Conscientiousness personalities are entirely different. That is, none of the persuasive principles occupies the same order in both profiles. For the other three personalities, the table shows that their profiles are different
eComm Movie eComm Movie eComm Movie eComm Movie eComm 3 4 2 4 2 4 3 2 1 5 5 5 6 5 6 5 5 5 2 3 4 3 4 3 4 4 4 to a high extent. Third, in the eCommerce domain, Liking and Scarcity are the least and the second least influential strategies for all personalities. Forth, Commitment is the most influential strategy for all personalities in the Movie domain.
To test the significance of the interaction between personality traits and the application domain, we deployed the Repeated Measure ANOVA (RM-ANOVA). As Table 6 depicts, we found that the interaction between personality traits and the domain of RSs have a statistically significant effect on users’ susceptibility to three persuasive principles; namely Reciprocity (F= 2.296, p = 0.05), Scarcity (F = 2.897, p = 0.023), and Liking (F = 2.305, p = 0.049). Comparing to the results of the previous subsection, we can say that the impact of personality traits, in term of their responses to the six principles, become more significant if it is studied in combination with application domains. This section shows that the differences between personalities are statistically significant for three principles, while the previous section shows that the results were statistically significant for one principle only. These results suggest that there is a more significant effect of the personality traits when they are considered in combination with the application domain.
Personalized recommendations have shown great success in the RS area; Giving suggestions that are tailored to every user, have increased users’ acceptance of the recommendations. An important question that may arise here is, “would persuasion still be useful if the recommended list of items is already personalized?” the answer to this question would be “yes.” In other words, the persuasive principles focus on “how to recommend” instead of “what to recommend.” Persuasive principles could stand as a cutting-edge to reduce users’ hesitation toward making a final decision. 4 1 3 1 3 1 2 1 3 1
1.018 0.398 Our results show that RS users are vulnerable to all of Cialdini’s principles. The results also show that personality traits can be an essential factor that may influence the decisions of RS users. That is, users’ personality is an important factor that should be considered when we design a persuasive RS. It is noteworthy to mention that the ANOVA results did not give enough evidence to reject the null hypothesis for some cases. Besides, our study considered two recommendation areas only. Accordingly, we suggest that wider studies that considers more recommendation domains are still required to generalize these results.
The following subsection provides general tips for designing persuasive RSs. We inferred these tips based on our analysis.
The previous section answered the research questions by presenting the results obtained from our study. It shows that users’ personalities may affect their responses to the persuasive principles, and the effect of personality traits becomes more significant if it is considered in combination with the domain of the recommendation. Based on these results, this section provides general guidelines to design persuasive RSs. • •
Personality traits are an essential factor that should be considered when we design a persuasive RS. The results show that the influence level of the six principles varies from one personality to another. So, a one-size-fits-all approach should not be used when we design a persuasive RS. Instead, an RS designer should consider users’ personalities in order to select the correct influencing approach to the right person.
Personalities are not treated in a black-and-white
basis. As the results shown, our analysis did not show
clear evidence that differences between personalities
are statistically significant for most of the cases. Thus,
we do not recommend treating users’ personalities on a
binary basis if your design depends on the FFM.
Instead, we suggest two solutions; first, consider the
combination between the traits, such that categorizing
the users as a combination between the existence and
the absence of traits. Promising work in this direction
is introduced by Sofia et al., [
The use of persuasive principles has been recently introduced to the RS area, and it has shown promising results in terms of increasing users’ acceptance of the recommendations. These principles should be personalized in order to improve their persuasiveness. As a first step to personalize them, we need to explore the factors that may help in this personalization process. Our work explores the effect of personality traits and the application domain on RSs users’ susceptibility to Cialdini’s principles of influence. It also explores the persuasive profiles for each personality trait under two application domains. The analysis of the results received from (279) responses to our questionnaire revealed that personality traits affect users’ responses to the persuasive principles, and this effect becomes more significant if we considered the interaction between personality traits and the application domain. Based on these results, we suggested some general guidelines that should be considered for designing a personalized persuasive RS.
As future work, we are working on exploring other factors that are expected to affect RS users’ responses to the Six weapons of influence. These factors include, but not limited to, users’ age, gender, and culture. Also, we should investigate the interaction between all (or part of) these factors, and how this interaction may affect the influence of persuasive principles. Besides, it is necessary to study the impact of the persuasive principles in other RSs domains, such as music and education RSs.