• No results found

Mood effects on emerging adults’ Facebook use

N/A
N/A
Protected

Academic year: 2021

Share "Mood effects on emerging adults’ Facebook use"

Copied!
57
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Mood effects on emerging adults’ Facebook use

Bram Theunissen 10118012 Master’s thesis

Graduate School of Communication Master’s programme Communication Science

Rinaldo Kühne Date: 24 June 2016

(2)

Abstract

Emotional appeals, in the form of videos and other content, are part of the social media landscape as it exists today, which is arguably due to the fact that these appeals have been proven effective in enhancing interaction with users. Thus far it is unclear however, what effect these appeals have on social media use following the consumption of these messages. This study attempted to address this issue by conducting a 3x1 factorial online experiment. Results showed that while there were some mood effects on commenting, liking, and intention to chat, possible remaining moods did not seem to affect general Facebook use following the consumption of mood evoking content.

(3)

Introduction

Social media has received a lot of attention in recent years, Facebook in particular. A substantial amount of research looked into how and why Facebook is used (e.g., Pempek, Yermolayeva & Calvert, 2009; Smock, Ellison, Lampe & Wohn, 2011; McAndrew & Jeong, 2012) as well as who uses certain features of the medium (e.g., McAndrew & Jeong, 2012; Nadkarni & Hofmann, 2012). Furthermore, there are several studies looking into individual determinants that can result in different positive as well as negative effects of social media use, such as social capital, Facebook depression, and social support (e.g., Ellison, Steinfield & Lampe, 2007; Frison & Eggermont, 2015; Nabi, Prestin & So, 2013; Tandoc, Ferrucci & Duffy, 2015; de Vries & Kühne, 2015). The role that mood and emotion might play in a social media setting has received relatively little attention however, despite the fact that mood effects have been found for several other media platforms. Mood can, for example, play a role in what type of music a person selects (e.g., Knobloch & Zillmann, 2002) as well as influence how this person perceives the music (Hunter, Schellenberg & Griffith, 2011). At the same time mood can influence what television show an individual chooses to consume (Zillmann, 1988). This selection process is often perceived as an attempt to regulate certain moods and emotions (Knobloch & Alter, 2006; Zillmann, 1988).

There has been research into the influence of messages containing emotion on

sharing, commenting, and liking behaviour however, which found that emotion being present in a message increases the chances of it being shared, liked, and commented on (Malhotra, Malhotra & See, 2013; Stieglitz & Dang-Xuan, 2013). At the same time, the type of emotional experience a user has as a result of consuming such a message also plays a vital role. Berger and Milkman (2012) found, among others, that content that evokes more intense or arousing emotions are often more likely to be shared (Botha & Reyneke, 2013). Both

(4)

emotional responses towards a message as well as the emotionality of a message can thus play a role.

It remains unclear however, what happens with that experienced mood or emotion after the mood-evoking content is consumed. Does it alter a person’s Facebook use like it does for other media (see Knobloch & Alter, 2006; Knobloch & Zillmann, 2002; Zillmann, 1988)? This study will build on the findings from those previous studies (Berger & Milkman, 2012; Botha & Reyneke, 2013; Malhotra et al., 2013; Stieglitz & Dang-Xuan, 2013), by not only looking into sharing, commenting, and liking but also incorporating the possible effects of mood on more general Facebook use like chatting, wall posts, and content selection. This leads to the following research question:

RQ: To what extent does the experience of a positive or negative mood affect emerging adults’ Facebook use?

Theory Facebook Dimensions

This study focuses on Facebook use amongst emerging adults, who are among the heaviest users, as was shown by Pew. 82% of the people between 18 and 29 years old use Facebook (Duggan, 2015), this percentage becomes gradually lower in the older age categories. It is thus not surprising that most research on Facebook is done amongst this younger age group (e.g., Pempek et al., 2009; Ross, Orr, Sisic, Arseneault, Simmering & Orr, 2009; Smock et al., 2011).

As this age group are relatively heavy users, studying mood effects on their Facebook use is relevant. Furthermore, while several other social media platforms are on the rise, Facebook is still the largest platform (Duggan, 2015)

(5)

Secondly, it is important to define dimensions of Facebook use to assess which features are most important to consider when studying mood effects. Generally research agrees that overall Facebook use revolves around three main dimensions, social interaction or relationship maintenance, expressive information sharing, and entertainment (Pempek et al., 2009; Sheldon, 2008; Smock et al., 2011, Voorveld, in press). According to Smock et al. (2011), social interaction takes the form of writing a post on someone else’s wall,

commenting, and chatting with others, while expressive information sharing is often done through posting status updates and posting in groups (Smock et al., 2011). While the sharing of content was not part of the study by Smock et al. (2011) Baek and his colleagues did find that sharing links is often done to share entertainment, information, news, and to express oneself (Baek, Holton, Harp & Yaschur, 2011) Lastly, entertainment often relates to more passive behaviours, like scrolling through your news feed, or other people’s profiles, and photos (Sheldon, 2008).

These different features will be defined as they are interpreted in this study. Posting on someone’s wall or his or her own wall can be defined as leaving a typed message, which can contain a picture, video or a link on the profile page of a friend, other Facebook page, or his or her own profile page. Secondly, commenting can be defined as leaving a response in the form of a typed message on someone else’s content or one’s own post. Thirdly, chatting is conversing with a friend using the built-in chat function of Facebook. Which can be used to send private messages to each other. Sharing, can be seen as posting content posted by someone else again on one’s own, or on someone else’s page. Consuming content on one’s own news feed or others profiles can be defined as reading status updates, articles, and watching photos and videos.

While this is not an exhaustive list of all of Facebook’s features, it does represent the most prevalent activities on Facebook. These core features will thus be at the centre of the

(6)

definition for Facebook used in this study. Which is: Facebook is a platform used for social interaction, expressive information sharing, and entertainment.

Mood Effects on Facebooks Entertainment Features

First, possible mood effects on entertainment features will be discussed. These

features can be characterized by more passive uses of Facebook, like consuming content from the news feed (Sheldon, 2008) as well as more active forms like sharing entertaining content (Baek et al., 2011). Mood effects on these two uses will be discussed in the following

paragraphs.

Content selection. First, consuming content from the news feed will be discussed, part of consuming content is the selection of that content, which is what this paragraph will focus on.

One of the most well-known theories regarding mood regulation through content selection is the mood management theory by Zillmann (1988). Moods being feelings that are generally low in intensity, high in longevity, but do not refer to a specific object (e.g., feeling happy) (Forgas, 1995).This theory posits that the consumption of messages can alter

prevailing mood states, as well as that the selection of these messages often serve mood regulation purposes.

Mood management theory focuses on hedonic motives for media choice. It posits that individuals will always strive to perpetuate good moods as well as maintaining its intensity (Zillmann, 1988). At the same time, individuals in a bad mood will strive to rid themselves of this mood or decrease its intensity. This can be achieved by consuming different stimuli, which can be any type of media message, such as the news or a comedy. The person

consuming the message does not have to be aware of his or her current mood state to attempt to alter it.

(7)

This hedonic premise has been contested in more recent research however. While there is numerous research that provides empirical evidence for the mood management theory (e.g., Bryant & Zillmann, 1984; Knobloch & Zillmann, 2002; Meadowcroft & Zillmann, 1987), the fact that there is also evidence that more sombre entertainment seems to appeal to individuals who feel sad (Gibson, Aust, & Zillmann, 2000; Mares & Cantor, 1992) seems to imply that there are other factors that play a role in media selection.

One explanation, that plays into these seemingly contradictory findings (eg., Gibson et al., 2000; Mares & Cantor, 1992), is offered by the social comparison theory. Specifically the theory of downward comparison by Wills (1981). This theory posits that someone can increase their well-being by comparing themselves to someone who has it worse or is in a similar situation. Which can remind the consumer of the fact that he or she does not have it that bad. Through this mechanism an individual in a negative mood can attend to this mood by consuming negatively valenced content.

Secondly, there is an explanation which takes a developmental perspective on this issue. A recent study by Mares and her colleagues (Mares, Oliver & Cantor, 2008) showed that age plays a role in preferences for more positive or negative content. Emerging adults (see, Arnett, 2000), ages 18 to 25, seem to have a higher preference for negatively valenced content than older adults. Mares et al. (2008) suggest that this could be due to the age of identity exploration, in which emerging adults find themselves (Arnett, 2007). In this age of identity exploration, emerging adults attempt to explore themselves. Part of this, is exploring different emotions, as well as their personality and other life experiences (Arnett, 2007). As such, experiencing negative affect might sound appealing to emerging adults. This does not necessarily mean that emerging adults in a negative mood will select negative content while their counterparts in a positive mood will select positive content however. Experimental evidence does seem to imply this.

(8)

An experiment conducted by Forgas and Bower (1987) showed that individuals tend to spend more time on examining and encoding mood congruent information. They argue that this is due to three reasons. First, affect can prime “the availability of mood-congruent

interpretations and the salience of congruent stimulus materials for learning” (Bower, Gilligan, & Monteiro, 1981, p. 451). Secondly, Forgas and Bower (1987) argue that a

selective exposure effect can contribute to longer processing of mood congruent information. Generally, selective exposure is defined as a process in which individuals attempt to avoid cognitive dissonance by consuming content that is consonant with their views and avoiding challenges to that view (e.g., D’Alessio & Allen, 2007; Frey, 1986), Forgas and Bower (1987) argue, that this also appears to apply to moods. As such mood congruent information might appeal more to an individual, since selective exposure states that individuals generally favour information that caters to their pre-existing views, or moods in this case. Finally, Forgas and Bower (1987) mention that mood consistent information is more likely to remind subjects of relevant episodes from their past. Which can lead to slower and deeper processing of that information.

To summarize, a selection process can take a motivational approach, like wanting to experience certain feelings (Arnett, 2007; Mares et al., 2008), a hedonic approach like the downward social comparison theory (Wills, 1981), as well as more unconscious mechanisms like the fact that a mood enhances salience of mood congruent stimuli (Forgas & Bower, 1987). All in all, both the downward social comparison theory (Wills, 1981) and empirical evidence (Forgas & Bower, 1987; Gibson et al., 2000; Mares & Cantor, 1992) seem to imply that mood congruent stimuli are preferred over their counterparts, appear more salient and can be used to improve mood. Furthermore, emerging adults, unlike older adults have less of an inhibition towards negative content (Mares et al., 2008). It is thus expected that emerging

(9)

adults will show a preference towards mood congruent content, which leads to the following hypotheses:

H1: Emerging adults will select Facebook posts that are congruent to their current mood

Sharing. The second Facebook feature in the entertainment dimension is the sharing of links, which is a core part of social media as it exists at this point in time. The content of a message can influence the likelihood of it being shared. As stated previously, research has found that content is more likely to be shared if it evokes a strong emotional reaction (Berger & Milkman, 2012; Botha & Reyneke, 2013), which they found to be especially pronounced for positive content. Emotion being present in the message also increases the chances of it being shared (Malhotra et al., 2013; Stieglitz & Dang-Xuan, 2013). Emotions, in this case are characterized as a feeling that is high in intensity, low in longevity, while referring to a specific object (e.g., being angry at or afraid of) (Forgas, 1995).

As such it is expected that Facebook content that is both positive in tone while also evoking a positive mood or emotion will be most likely to be shared, if compared to content that is neutral or negative in affective tone.

H2a: Emerging adults in a positive mood will be most likely to share the content that evoked that mood

While the research cited above suggests that positive content is most likely to be shared, it does mention that content containing emotion in general is more likely to be shared. As such, it is to be expected that Facebook content that has a negative affective tone and evokes a negative mood should be more likely to be shared than neutral Facebook content. The negative content should be less likely to be shared than positive content however.

H2b: Emerging adults in a negative mood will be more likely to share the content that evoked that mood than emerging adults who are in a neutral mood

(10)

Mood Effects On Facebooks Social Interaction Features

As stated previously the main Facebook uses associated with the social interaction dimension are chatting, commenting, and writing wall posts. Mood effects on the first two elements will form the focus of the following paragraphs, as the writing of wall posts is expected to be mostly influenced in the expressive information sharing dimension.

Mood is expected to have an effect on how frequently these features are used as well as the valence of the comments and chat messages that are sent. Frequency will be discussed first, followed by valence.

Frequency.

Chatting. Starting with chatting, it is expected that the experience of a certain mood

will be followed by social sharing, by individuals in a positive as well as individuals in a negative mood. Rimé (2009) argues that the experience of mood and emotion, if intense enough (see, Luminet IV, Bouts, Delie, Manstead & Rimé, 2000), leads to sharing of this state and the situation that evoked it. This experience is usually shared with friends or family among emerging adults, rather than professionals or strangers (Rimé, 2009).

According to Rimé (2009) social sharing following the experience of a negative emotion can happen in various ways, most important of which for this study: social comparison and narration, (Rimé, 2009). The sharing of emotions through the social comparison mechanism builds on Festinger’s classic theory of social comparison (1954), which states the individuals have a continuous need to assess their opinions. This can be done by comparing their views with individuals in their environment through which individuals can arrive on common worldviews and opinions. This is due to the fact that an emotional experience towards a certain object is not always something that can be compared through simply observing, especially when individuals are not in the same location. Furthermore, the experience of emotion can stimulate the production of narration (Rimé, 2009). Bruner (1990)

(11)

argues that narration can only emerge when common sense expectations are violated. The experience of a negative emotion arguably fits this description, as there is no reason to

experience a negative emotion if everything goes exactly as planned. For example, one can be angry if his or her laptop is stolen, at the same time this makes for a good story. Still having your laptop on the other hand, does not make for an exciting story nor should it evoke a strong emotional reaction.

Positive emotions lead to sharing through a different avenue. Re-accessing past positive experiences can lead to reliving past feelings and sensations (Rimé, 2009). Positive emotional episodes can thus be seen as an event an individual can take advantage of, besides ruminating on the event one can share this event through his or her social network to relive these positive feelings and sensations (Langston, 1994). Which according to more recent research also contributes to increased life satisfaction (Quoidbach, Berry, Hansenne & Mikolajczak, 2010).

If this perspective is applied to someone consuming content on Facebook that evokes a certain mood or emotion, it is expected that this individual will feel the need to reach out to someone. Since the content that he or she consumed either gave the opportunity to relive positive experiences, makes for a good story, or stimulated a need to assess their opinions of that content. This is most likely to occur through the use of Facebook chat, as this most direct form of communication possible on this type of medium, allowing for the most direct contact with friends and family, who are generally the target audience for sharing such an experience (Rimé, 2009). These individuals are the target for social sharing as sharing also is an

opportunity to strengthen the relationship between the sharer and the sharing targets (Rimé, 2009). Moreover, the sharing of a certain emotion can be a relatively private concern for most people, depending on the emotion. Indicating that tie strength with the sharing target is an important factor.

(12)

While it is not unlikely that individuals in general want to use the chat function on Facebook, individuals who recently experienced a mood-altering event should have an additional motivation to do so, as it allows them to share this experience with a specific person, rather than their entire social network.

H3: Following the experience of a positive or negative mood, emerging adults will attempt to (a) share the event that elicited this mood by (b) talking to individuals close to them (c) using Facebook chat more than those in a neutral mood.

Commenting & liking. Smock et al. (2011) found that commenting, besides being a

tool that is used to interact with one another, is also seen as a form of relaxing entertainment (which the authors define as features that allow someone to unwind, relax, and enjoy

something). At the same time it is not seen as a feature that stimulates companionship (Smock et al., 2011). Viewing this from Rimé’s (2009) perspective, this means that

commenting is not something that an individual in a negative mood will want to spend too much time on, as Rimé states that someone in a negative mood is looking for someone close to talk to. Which is in a sense a form of companionship. The relation with relaxing

entertainment suggests that commenting is something that most individuals don’t spend too much cognitive processing power on, as a demanding cognitive task is hardly relaxing. As such there should be plenty cognitive processing power to ruminate on other feelings. According to Quoidbach and his colleagues (Quoidbach et al., 2010), this allows an individual to be present in the moment and ruminate on positive thoughts. A strategy the authors found to be applied frequently by individuals in a positive mood, as well as being an effective strategy to enhance positive affect.

Moreover, as stated in the introduction and the sharing of links section, empirical evidence has shown that mood and emotion can influence how often a post is shared (Berger

(13)

& Milkman, 2012; Stieglitz & Dang-Xuan, 2013). The same appears to be true for commenting and liking, as was shown by Malhotra, Malhotra and See (2013), who did a content analysis of over 1000 wall posts distributed by companies. Their results showed that emotion being present in a message can be one of the factors that increases commenting and liking. While emotion being present in a message is expected to enhance interaction

(Malhotra, Malhotra & See, 2013; Stieglitz & Dang-Xuan, 2013) it should be noted that it is likely that there will be a difference between the positive and negative moods.

To summarize, while emotion being present in a message will likely lead to increased interaction (Berger & Milkman, 2012; Malhotra et al., 2013; Stieglitz & Dang-Xuan, 2013), individuals in a positive mood will be most likely to comment. This leads to the following hypotheses:

H4a: Emerging adults in a positive mood are more likely to comment on content that evoked that mood than emerging adults in a negative mood

H4b: Emerging adults in a negative mood will be more likely to comment on content that evoked that mood than emerging adults in a neutral mood

Liking. Liking, similarly to commenting is expected to change due to mood.

Empirical evidence by Malhotra et al. (2013) showed that emotion being present in a message does increase the amount of likes on a post, like it does for commenting and sharing.

Malhotra and her colleagues (2013) argue that this is due to the fact that showing emotion in a post can humanize communication, which has proven to be a successful communication strategy in previous research (Solis, 2010).

In line with these findings it is thus expected that:

H5: Emerging adults will be more likely to like the content that evoked a positive or negative mood than they are to like one that evoked a neutral mood

(14)

Valence.

Chatting. Besides frequency it is expected that mood will alter the affective tone of

messages sent on Facebook. Assuming that the third hypothesis is correct, emerging adults should share their experience in Facebook chat, which should affect the affective tone of that message. Positive experiences are often shared to relive certain sensations and feelings experienced as a result of that event. As the shared message should contain a positive event, the tone of that message should generally be positive. Negative experiences are shared to allow the individual to make sense of his or her experience by comparing it to one from someone else or through conversing about that experience. As a result the message should have a more negative tone, as it contains a negative experience.

Besides being the topic that is conversed on, affect can influence communication (among others) in other ways. The Affect Infusion Model (AIM) (Forgas, 1995) and the previously mentioned affect priming mechanisms by Forgas and Bower (1987) attempt to explain this. According to the AIM, affect can influence evaluations through two ways of processing, the model does recognize two alternative ways of processing in which affect does not alter evaluations (Forgas, 1995). Heuristic and substantive processing do allow for affect infusion to occur however, these two processing strategies both require some form of open and constructive thinking (Forgas, 1995). The first being a relatively quick, yet open form of processing. This process is often used when the target is simple or highly typical, there are no motivational objectives, and when the individual has limited cognitive processing power available (Forgas, 1995). While substantive processing is a more elaborate form of processing. This process is used when the target is complex or atypical, the judge has no specific motivational objectives, sufficient cognitive processing power, and is motivated to be accurate (Forgas, 1995).

(15)

Substantive processing can be influenced indirectly through the previously mentioned affect-priming mechanisms, which state that mood enhance the availability of mood

congruent interpretations, memories, and the salience of mood congruent stimuli (Forgas & Bower, 1987; Forgas, 1995; Forgas, 1999). Direct mood influences can also occur through the affect-as-information principle. According to this principle individuals can use their feelings as a shortcut to decide on their evaluative reactions to a target (e.g., “How do I feel about this?”) during the relatively fast, heuristic processing.

Generally, research has shown that individuals in a positive mood prefer to rely on a more superficial form of processing, basing their judgment on heuristics (Mackie & Worth, 1989, as cited in Kühne, 2012). Individuals in a negative mood prefer more systematic processing however (Chartrand, van Baaren & Bargh, 2006; Forgas & Bower, 1987; Schwarz, 1990, as cited in Kühne, 2012). As such, it is expected that emerging adults in a positive mood will process the question “What shall I write in this chat message?” through the heuristic process. This allows for the affect-as-information to occur, which should lead to individuals in a positive mood answering “What do I feel like writing in this chat message?” rather than “What shall I write in this message?”. As a result the written message is likely to be mood congruent. Individuals in a negative mood are more likely to answer the question “What shall I write in this chat message” through the more elaborate substantive processing method, which allows for mood influences through the affect-priming principle. Since mood primed mood congruent interpretations and memories (Forgas & Bower, 1987) it is likely that messages by individuals in a negative mood will be mood congruent as well.

While AIM is not a predictor of behaviour but rather judgment, its principles, like affect-priming, have been proven useful in explaining fluctuations in communication style as a result of mood (e.g., Forgas, 1999). Since mood can be used to help inform an individual about his or her evaluative judgement towards an object, as well as priming mood congruent

(16)

memories and experiences it is expected that individuals will communicate differently as a result of their mood. Either in a more positive manner due to a positive mood, or in a more negative manner due to a negative mood.

It is thus expected that the participants will show mood congruent sentiment in their chat messages.

H6: As a result of their mood, emerging adults will show mood congruent sentiment in their chat messages

Commenting. Similar to the chat messages it is expected that the comments will differ

in their sentiment, either more positive or negative. Since Facebook users often directly comment on content it is expected that comments like these are related to the content. As such, the sentiment displayed in the comments are likely to be consistent with the content they viewed.

Following the same reasoning as was described in the chatting section, comments are expected to show mood congruent sentiment. Emerging adults in a positive mood will process information through the heuristic process, which allows for the affect-as-information effect to occur (Forgas, 1995). Emerging adults in a negative mood will process information through the substantive process, allowing for affect-priming to occur (Forgas, 1995). This should result in mood congruent sentiment in the respective comments. This leads to the following hypothesis:

H7: As a result of their mood, emerging adults will show mood congruent sentiment in their comments

(17)

Mood Effects On Facebooks Expressive Information Sharing Features

The main Facebook uses associated with the expressive information sharing

dimension are the wall posts and the sharing of links. Possible mood influences on the writing of wall posts will be discussed in the following paragraphs. As the sharing of links has

already been discussed in the entertainment dimension.

Wall posts. The writing of wall posts can take multiple forms, as stated previously. It can take the form of a message of one’s own Facebook page or someone else’s page. It is expected that mood will influence the affective tone of such a shared message. This is again based on the previously mentioned affect-as-information and affect-priming mechanisms by Forgas and Bower (1987; Forgas, 1999) and the Affect Infusion Model (AIM) (Forgas, 1995).

It can be argued that the sentiment of such a wall post will be mood congruent similarly to the previously discussed comments and chat messages. Emerging adults in a positive mood will process the question “What shall I write in this wall post” through the heuristic process, which allows for the affect-as-information effect to occur (Forgas, 1995). Emerging adults in a negative mood will process the same question through the substantive process, allowing for affect-priming to occur (Forgas, 1995). This should result in mood congruent affective tone in the written wall posts, which leads to the following hypotheses. H8: As a result of their mood, emerging adults will show mood congruent sentiment in their wall posts

Moderators

Besides the main effects it is expected that some processes will be moderated by gender, incremental and entity beliefs, and extraversion.

(18)

Gender. First, gender. Research has shown that women tend to be more forthcoming with personal information than men (Dindia & Allen, 1992). While this seems to imply that women might be more prone to social sharing, Rimé (2009) stated that women are equally likely to share their moods and emotions as men. A medium like Facebook however, generally revolves around sharing and consuming experiences. As a result women’s higher tendency to disclose personal information (Dindia & Allen, 1992) could result in higher rates of social sharing.

If women are to be more prone to social sharing, it is likely that they would show a greater preference for the chat function of Facebook. Which is most likely the outlet were social sharing is to occur (see chatting section). It is thus predicted that while the experience of mood will lead emerging adults to use the chatting function, it will be especially

pronounced for women. To test this assumption the following hypothesis has been drafted: H9: Gender is expected to moderate the effect that mood has on the intention to chat

Incremental and entity beliefs. The second moderator in this process is incremental versus entity beliefs. People with incremental beliefs see emotions as the kind of thing that can be changed, people with entity beliefs see emotions as something you cannot change (Tamir, John, Srivastava & Gross, 2007). Individuals with incremental beliefs are more likely to attend to their mood and in general more proficient in emotion regulation than people with entity beliefs (Tamir et al, 2007). This could lead to individuals who have incremental beliefs to attend to their mood states more often than people who have entity beliefs, since they believe that they can alter their mood state (Tamir et al., 2007). According to Rimé (2009), this should be done through social sharing. To test if individuals with incremental beliefs are more likely to attend to their moods, the following hypothesis has been drafted:

(19)

H10: Emerging adults with increased incremental beliefs are more likely to attend to their mood through chatting than emerging adults with entity beliefs

Extraversion. The final moderator is believed to be extraversion. Research has shown that people who are more extraverted use Facebook more as a social tool, have more Facebook friends, and use social network sites more in general (Nadkarni et al., 2012). It is thus not a stretch to state that extraverted people might use more communicative features than introverted people While mood is expected to affect several types of Facebook use, like chatting, commenting, and liking, it is expected that higher and lower levels of extraversion will strengthen or lessen this effect. To test this, the following hypothesis has been drafted: H11: Extraversion is expected to moderate the effect of mood on the use of communicative features on Facebook

Method Sample

119 people participated in this study. Participants using a mobile phone were not allowed to participate since viewing the survey on a PC ensures a more involving viewing experience than a mobile platform. 19 participants only filled in part of the survey or stopped after watching the mood induction video. These cases were removed before further analysis. As a result 100 cases were left for the full analysis.

The sample had an average age of 24 years (M= 23.98, SD= 2.27), with the youngest participant being 19 and the oldest being 30. The sample was relatively skewed towards higher education, with 80% of the participants indicating University as their highest level of education (N= 80) and only 12% indicating HBO as their highest level of education (N=12). HBO is the second highest level of education in the Netherlands.

(20)

Procedure

The experiment took the form of a 3x1 factorial online experiment. The participants were told that the study was on the relationship between online videos and Facebook use. Afterwards they viewed either a positive, neutral, or negative stimulus which differed in their emotional content. All three mood conditions were induced using audio-visual material and text (see appendix B).

After watching the stimulus they answered a questionnaire on what they would like to do with the video (e.g., commenting, liking, and sharing) they saw followed by questions on what they would like to do most on Facebook at that moment. These questions encompassed all features described in the theory section. Participants were also asked what exactly they would write in a comment, chat message, or wall post.

The experiment then proceeded with a selection task. The participant was presented with two Facebook posts at a time regarding the same subject, while similar in topic and length their valence was opposite from each other.

The last questions related to the extraversion trait followed by questions on incremental versus entity beliefs and control variables regarding general Facebook use.

Finally, the experiment ended with a short debriefing on the intended effects of the video to ensure that any residual moods were eliminated. See appendix A for the full survey.

Stimulus

According to previous research on mood induction procedures movies and stories are most effective (Westermann, Spies, Stahl & Hesse, 1996) in inducing a certain mood. These movies can be as short as three minutes and still be effective in inducing intense emotion (Luminet IV et al., 2000). The material used in this study is of a similar length.

(21)

A commercial for business cards has been used as a basis for all three conditions, ensuring that all three conditions view similar stimuli with the exception of the mood it is inducing. The three videos end in a text describing the ending in a narrative manner. This is the only point in which the videos differ. The story is about a father and son who start a bakery together and work together while having a lot of fun. At a certain point the son sees a job opening in the paper. The video is cut off at that point and finished in text.

The fact that the videos are similar in characters and setting can be of vital importance for the results, since mood is the only fluctuating factor, the true effects of mood on Facebook use can be identified. For example, different music or a different setting between conditions might induce a slightly alternate mood or might prime different memories. These small differences could have an effect on how an individual responds. As a result the conditions and the results are less comparable to each other, preventing the identification of true mood effects. See appendix B for the full stimulus.

The edited video fulfils the role of mood induction as well as the independent variable on which most hypotheses have been based.

The stimulus was pre-tested to avoid a questionnaire in the main experiment on the effectiveness of the mood induction, which could limit the effectiveness of the mood

induction by making the participants aware of the irrelevance of that mood to the task at hand (Schwarz & Clore, 1983). The study by Schwarz & Clore (1983) showed that the impact of a bad mood can be eliminated if the individual realizes that the mood is irrelevant to the task/evaluation at hand.

The pre-test used the PANAS X scale to measure the induced moods as described and tested by Watson and Clark (1999), specifically the general positive (ten items), joviality (eight items), general negative (ten items), and sad scales (five items). The joviality and sadness scales were chosen because of the fact that the stimuli are likely to elicit these

(22)

emotions because of their content. To account for the possibility that the stimuli might induce other affective reactions as well, the general scales were included.

The items from each scale consisted from single words such as happy for joviality, attentive for general positive, sad for sadness, and upset for the general negative scale. The respondents had to rate to what extent they felt that way at the moment using a 5 point Likert scale, 1 being “Very slightly or not at all”, 2 “A little”, 3 “Moderately”, 4 “Quite a bit”, and 5 “Extremely”. Mood scales were constructed by averaging the items, creating a mean index. For the full pre-test survey, all the items from the scales, and reliability scores see Appendix C & D and E.

All four scales gave a clear indication from the difference between positive, neutral and negative condition (see appendix E, table 1). With participants in the positive condition scoring 3.65 (M= 3.65, SD= .56) on the joviality scale for example, indicating a moderate presence of that mood. Participants in the negative mood condition on the other hand showed little to no presence of that mood, with a score of 1.84 (M= 1.84, SD= .65)on the joviality scale.

The effect sizes in table 2 (see, appendix E) show a similar picture, with the general negative scale showing smaller effect sizes. While the other scales show medium to large effect sizes in the expected directions. This thus suggests that the stimulus videos differ in emotional content.

Measures

This study measured the effects of videos that differ in emotional content on different aspects of Facebook use. To measure this, different types of Facebook use and its scales have been used. The scales have been tested using factor analyses as well as a Cronbach’s Alpha calculation. For a full list of measures see appendix A.

(23)

All scales used a 5 point Likert scale ranging from strongly disagree (1) to strongly agree (5) unless indicated otherwise.

Content selection. In the content selection section the participant was asked to choose between two posts with the question: Which of the two posts do you feel like reading most? These posts were stripped from details about the Facebook page that posted the content, number of likes, comments, and shares. The posts dealt with the following topics: climate change, education, MMA, dogs, and a wedding. A full list of the posts can be found in appendix A.

The five posts were recoded to ensure that 1 was a positive selection and 0 a negative selection. Those were then added to each other into a content selection sum index in which 5 meant all positive selections and 0 all negative selections (M= 2.94, SD= 1.34).

Sharing. Sharing was operationalized as spreading the video through a social network. The items measuring sharing were based on a single item by Ross and his

colleagues (2009), expanded by two additional items to ensure a more complete coverage of the concept. The scale for sharing consisted of three items, an example being: “I would like to show this video to someone”.

A principal component analysis showed that the three items formed a single uni-dimensional scale. After a Varimax rotation both the eigenvalue (eigenvalue 2.47) and the scree plot indicated that there was one component explaining 82,29% of the variance. The Cronbach’s Alpha was more than sufficient at .89. The three items were averaged to create a mean index (M= 2.37, SD= .994). Observed scores ranged from 1 to 4.33.

Chatting. Chatting was operationalized as responding, sending and reading private messages, as suggested by Ross et al. (2009). Chatting consisted of three items. For example: ”I would like to send private messages”.

(24)

A principal component analysis showed that these three items form a uni-dimensional scale. After a varimax rotation both the eigenvalue (eigenvalue 2.24) and the scree plot indicated that there was one component that explained 74.63% of the variance. Chatting proved reliable as well with a .83 Cronbach’s Alpha. The three items were averaged to create a mean index (M= 3.17, SD= .90). Observed scores ranged from 1 to 5.

Commenting. Commenting is operationalized as participating in the conversation on the video. The scale was based on a single item by Pempek et al. (2009), to ensure that the whole concept would be covered two items were added. The scale for commenting consisted of three items such as: “I would like to comment on the video”.

A principal component analysis showed that the three items formed a single uni-dimensional scale. After varimax rotation both the eigenvalue (eigenvalue 2.24) and the scree plot indicated that there was only one component, which explained 74.63% of the variance. The scale proved reliable with a Cronbach’s Alpha score of .83.The three items were

averaged to create a mean index (M= 2.09, SD= .87). Observed scores ranged from 1 to 4.33. Liking. Liking the video was operationalized as liking the video. As such the liking scale consisted of a single item: “I would like this video”, M= 3.11, SD= 1.11. Observed scores ranged from 1 to 5.

Social sharing, and sentiment. As stated in the procedure section, the participants had to answer questions on what they would write in a comment, chat, or wall post, which they could then fill in in a text box. These answers were coded to allow for statistical analyses. They were coded for type of sentiment, using SentiStrength (see the following studies for detailed analysis of this tool, Stieglitz & Dang-Xuan, 2013; Thelwall, Buckley, Paltoglou, Cai & Kappas, 2010; Thelwall, Buckley & Paltoglou, 2012), which gives a text a positive score (maximum of 5) and a negative score (maximum of -5). Besides sentiment they were also coded for social sharing following the definition set by Rimé (2009). Which is:

(25)

sharing the conditions portrayed in the video which evoked the mood state as well as/or the state itself. By this definition a message like: “Have you seen this video? It’s really sad.” is social sharing, while “Who chose the music, it’s horrible” is not. While stating discontent of the video is, in a way, sharing of emotion it is not the type of sharing this study is looking into. This was coded as either 0 for not sharing or 1 for sharing. To ensure that the definition was correctly interpreted a secondary coder coded 30 cases, results showed that both coders interpreted the definition correctly, Kalpha= .84.

Tie strength. The third hypothesis also deals with tie strength, as in who they would want to talk to when using Facebook chat, to this end a question dealing with this concept has been added. Based on a scale from Gilbert & Karakhalios (2009) the following question has been implemented: “Imagine yourself using Facebook chat, who would you chat with? Please indicate how close you are with the person(s) you would chat with.” Followed by a slider, like in Gilbert & Karakhalios’ study (2009), from 0 “Barely know them” to 100 “We are very close”.

Moderators.

Incremental and entity beliefs. The scale measuring incremental and entity beliefs

consists of 4 items using a 5 point likert scale, as designed by Tamir et al. (2007), an example being: “Everyone can learn to control their emotions”.

A principal component analysis showed that these 4 items formed a uni-dimensional scale. After a varimax rotation both the eigenvalue (eigenvalue 2.61) and the scree plot indicated that there was one component, explaining 65.27% of the variance, as well as being reliable, Cronbach’s Alpha= .82. After recoding the two reversed entity items a higher score means a higher level of incremental beliefs while a lower score indicates entity beliefs, with 3 being the middle ground. These four items were averaged to create a mean index (M= 3.36, SD= .87). Observed scores ranged from 1.25 to 5.

(26)

Extraversion. The 12-item extraversion scale developed by Francis and his colleagues

(Francis, Brown & Philipchalk, 1992) was used to measure the extraversion trait. This scale consists of questions like “Are you a talkative person?” and “Do you like mixing with people?”, which could be answered with no (0) or yes (1). As two items were reversed, these items were recoded to ensure that a higher score meant higher levels of extraversion in an individual.

A principal component analysis showed that these 12 items did not form a uni-dimensional scale. After a varimax rotation both the eigenvalue and scree plot indicated 4 components, these items were not removed however. The 12 items explained 60.21% of the variance. This scale proved reliable with a .85 Cronbach’s Alpha. The 12 items were

averaged to create a mean index (M= .76, SD= .25). Observed scores ranged from 0 to 1. Control variables. The last two scales, acting as control variables, were all measured on a 5 point Likert scale ranging from “Not at all” to “Exactly” as designed by Sheldon (2008). These scales contained statements related to the described features of Facebook described in this study, such as chatting, wall posts, as well as more general measures such as I use Facebook to pass time. The statements were preceded by: ”I use Facebook for the following reasons”.

Entertainment. The entertainment scale consisted of four items, like: “To see other

people’s pictures”.

A principal component analysis showed that these four items formed a

uni-dimensional scale. After varimax rotation both the eigenvalue (eigenvalue 2.51) and the scree plot indicated that there was one component, explaining 62.79% of the variance, with a Cronbach’s Alpha of .80. The items were averaged to create a mean index (M= 3,65, SD= .76). Observed scales ranged from 1 to 5.

(27)

Social interaction. The social interaction scale consisted of six items, such as: “To

send a message to a friend”.

A principal component analysis showed that these six items formed a uni-dimensional scale. After a varimax rotation both the eigenvalue (eigenvalue 2.77) and the scree plot indicated that there was only one component, explaining 46.09% of the variance. This scale proved reasonably reliable with a Cronbach’s Alpha of .75.These six items were averaged to create a mean index (M= 3.56, SD= .77). Observed scales ranged from 1 to 5.

Results

As most hypotheses compared the difference between mood states, and an interval, dependent variable, the result section mainly consists of one-way ANOVA’s, unless indicated otherwise. All ANOVA’s were accompanied by a Bonferroni post hoc test. Furthermore, control variables were added to most ANOVA’s, social interaction and entertainment (see method for description). These control variables consisted of items relating to the

aforementioned dimensions of Facebook use, and how the participant used those in their daily lives. As a result the found variances are most likely due to the mood induction, instead of certain preferences a participant might have to certain Facebook features. To give some indication of effect size, partial eta squared has been reported as well.

Each set of hypotheses is discussed separately along with the analyses that have been conducted to answer them.

Randomisation Check

To check if the sample was evenly distributed based on important demographic and dispositional characteristics, several randomisation checks were done with: sex, incremental and entity beliefs, extraversion, and age. The first three characteristics were chosen as they

(28)

are expected to moderate some mood effects, it is thus important that these factors are equally distributed amongst the conditions. Age was selected as age differences have been shown to play a role in media selection and appreciation of that selected media (Mares et al., 2008). While the sample should consist of emerging adults only, it is possible that some participants are further along in their development than others. As such it is important to ensure that age is equally distributed amongst the conditions.

Sex has been tested with a Chi-square test with mood condition as the second variable. The results showed no significant differences between the three mood conditions, X2(2, N = 98) = .66, p= .719.

Age, incremental and entity beliefs, and extraversion were tested using three separate one-way ANOVA’s with the aforementioned as dependent variables and the mood condition as independent variable. The results showed no significant difference for age, F(2, 97)= .94, p= .395, η2= .02 (positive: M= 24.39, SD= 2.6, neutral, M= 23.91, SD= 2.15, negative: M= 23.64, SD= 2.04) , incremental and entity beliefs, F(2, 97)= .11, p= .894, η2= .00 (positive: M= 3.33, SD= .87, neutral: M= 3.33, SD= .95, negative: M= 3.42, SD= .87), or extraversion, F(2,97)= .15, p= .859, η2= .00 (positive, M= .78, SD= .27, neutral: M= .77, SD= .25,

negative: M= .75, SD= .23).

Furthermore, to ensure that both social interaction and entertainment could be used as covariates, independence from the mood condition has been tested. Entertainment showed no significant differences between positive (M= 3.7, SD= .59), neutral (M= 3.65, SD= .73), and negative (M= 3.59, SD= .95) conditions, F(2, 97)= .18, p= .836, η2= .00, neither did social interaction, F(2, 97)= .63, p= .538, η2= .01, between the positive (M= 3.22, SD= .67), neutral (M= 3.16, SD= .74), negative (M= 3.37, SD= .92) conditions.

(29)

Main Analyses

The first hypothesis predicted that emerging adults would be more likely to select mood congruent content. Mood condition was the independent variable and content selection the dependent variable. Higher scores on the content selection scale indicated a preference for positive content while lower scores indicated a preference for negative content. A score of 5 meant all positive selections while a score of 0 meant all negative selections. The test did not show a significant result however, F(2,95)= .41, p= .662, η2= .01. Overall participants in all three conditions favoured slightly more positive content (positive: M= 2.76, SD= 1.39, neutral: M= 3.03, SD= 1.34, negative: M= 3.03, SD= 1.31).

The second set of hypotheses predicted that participants in a positive mood would be most likely to share the content that evoked that mood. As well as that participants in a negative mood would be more likely than participants in a neutral mood to share the content that evoked that mood. Mood condition was the independent variable and sharing the

dependent variable. Sharing was measured on a 5-point scale, a score of 5 indicated willingness to share while 1 meant unwillingness to share, 3 was a neutral score. Results showed that no significant differences existed between the positive (M= 2.53, SD= 1.19), neutral (M= 2.36, SD= .91), and negative (M= 2.22, SD= .86) conditions however, F(2, 95)= .89, p= .416, η2= .02.

The third prediction stated that participants in a positive as well as negative mood would be inclined to chat about the event that elicited the mood by talking to individuals close to them using Facebook chat. To test the social sharing hypothesis three analyses have been conducted, first a one-way ANCOVA with a LSD test, to test if there was a difference between mood conditions (independent) and wanting to use the chat function (dependent). Intention to chat was measured on a 5-point scale, a score of 5 indicated willingness to chat while 1 meant unwillingness to chat, 3 was a neutral score. Secondly, a one-way ANCOVA

(30)

with a Bonferroni test, mood condition as independent variable and tie strength as dependent variable. Tie strength was measured on a 0 to 100 scale, 100 meaning very close and 0 not close at all. And finally a Chi-square test if any differences between amount of social sharing (ordinal variable with 2 categories) existed in the different mood conditions. Tie strength and intention to chat were analysed separately as they are conceptually quite different, and they did not show any significant correlation (r= -.04, p= .728). Social sharing was coded as an ordinal variable with two categories and could thus not be incorporated into an ANOVA.

Results showed a significant difference between the conditions for wanting to use the chat function, F(2,95)= 3.47, p=.035, η2= 0.07. Levene’s test was not significant, F(2, 97)= .12, p= .901, so LSD’s post hoc test was interpreted. Participants in a negative (M=2.89, SD= .92) mood were significantly less likely to use the chat function than participants in the neutral (M= 3.26, SD= .85) condition, p= .034, and positive (M= 3.34, SD= .89) condition, p= .018. No significant differences existed between the positive and neutral conditions, p= .791.Tie strength, F(2, 95)= .19, p= .831, η2= .00, did not show significant differences between groups, the mean indicated high tie strength however for positive (M= 79.76, SD= 18.23), neutral (M= 77.41, SD= 16.81), and negative (M= 79.15, SD= 21.50) conditions. Finally, social sharing did not show significant differences either, X2(2, N = 100) = 4.82, p = .09, between the positive (M= .30, SD= .08), neutral (M= .21, SD= .08), and negative

conditions (M= .46, SD= .08).

The fourth hypothesis predicted that participants in a positive mood would be most likely to comment on the mood evoking content, participants in a negative mood would still be more likely to comment on the mood evoking post than participants in a neutral mood. Mood condition was the independent variable and commenting the dependent variable. Intention to comment was measured on a 5-point scale, a score of 5 indicated willingness to comment while 1 meant unwillingness to comment, 3 was a neutral score. No significant

(31)

differences between positive (M= 2.14, SD= .99), neutral (M= 2.04, SD= .80), and negative conditions (M= 2.09, SD= .87) were found however, F(2, 97)= .11, p= .893, η2= .002.

The fifth hypothesis predicted that participants would be more likely to like the post that evoked a positive or negative mood than one that evoked a neutral mood. Mood

condition was the independent variable and liking the dependent variable. Liking was measured on a 5-point scale, a score of 5 indicated willingness to like while 1 meant

unwillingness to like, 3 was a neutral score. Results showed a significant difference between the groups, F(2, 95)= 5.03, p= .008, η2= .10. The Bonferroni indicated that emerging adults in a negative mood were significantly less likely to like the stimulus (M= 2.61, SD= 1.10) than emerging adults in a positive (M= 3.39, SD= 1.20), p= .016, and a neutral mood (M= 3.32, SD= .88), p= .028. No significant differences existed between the neutral and positive condition, p= 1.000.

The sixth prediction related to valence. Participants in a positive mood would send more positive messages than participants in the other two conditions. Mood state was the independent variable and positive sentiment in chat messages (higher scores mean more positive sentiment) the dependent. The results show no significant difference in how positive each group communicated (positive, M= 1.83, SD= .47, neutral, M= 1.61, SD= .63, negative, M= 1.75, SD= .79), F(2,76)= .78, p= .461, η2= .02.

Similarly to the previous hypothesis it was expected that participants in a negative mood would send more negative messages than participants in the other two conditions. As such negative sentiment in chat messages was the dependent variable (higher scores meant higher negative sentiment) and mood condition was the independent variable. It should be noted that the negative sentiment variable was not normally distributed.1 The ANCOVA for

1 In an attempt to address this issue both log transformations and square root procedures have been attempted. The distribution remained skewed however.

(32)

this hypothesis showed a significant difference between conditions in how negative each group communicated, F(2,78)= 3.26, p= .044, η2= .08. The Bonferroni post hoc test indicated that significant differences existed between negative (M= 1.75, SD= 1.36) and neutral (M= 1.11, SD= .31) conditions, p= .026. No significant differences existed between negative and positive (M= 1.37, SD= .82) mood conditions, p= .282, or neutral and positive, p= .825.

The seventh prediction stated that participants would show mood congruent sentiment in their comments. This hypothesis has been tested using two one-way ANCOVA’s. Mood condition was used as an independent variable, sentiment of the comment they would write (positive and negative) were the dependent variables. The results showed a significant difference in negative sentiment in the comments, F(2, 79)= 8.90, p< .001, η2= .18.

Bonferroni’s post hoc test indicated that participants in a negative mood wrote significantly more negative comments (M= 2.90, SD= 1.57) than individuals in a positive mood (M= 1.44, SD= 1.16), p< .001, as well as those in a neutral (M= 2, SD= 1.29) mood, p= .037. No

significant differences were found between the positive and neutral conditions, p= .249. Results also showed a significant difference in positive sentiment in comments, F(2, 79)= 15.40, p< .001, η2= .28. Bonferroni testing showed that individuals in a positive (M= 2.24, SD= .83) and neutral mood (M= 2.17, SD= .79) wrote significantly more positive messages than individuals in a negative mood (M= 1.24, SD= .58), (positive and negative, p<.001, neutral and negative, p< .001). No significant difference existed between the positive and neutral condition however, p= 1.000.

The final hypothesis predicted that participants would show mood congruent

sentiment in their wall posts, similarly to their comments and chat messages. This hypothesis was tested using two one-way ANCOVA’s with Bonferroni, both with mood condition as independent variable and either positive sentiment in wall posts or negative sentiment in wall posts as dependent variables. The tests did not show significant results for positive sentiment

(33)

in wall posts, F(2, 50)= .24, p= .789, η2= .01, between the positive (M= 1.78, SD= .60), neutral (M= 1.85, SD= .81), and negative conditions (M= 1.83, SD= .72). The same applied to the positive (M= 1.04, SD= .21), neutral (M= 1.1, SD= .31), and negative conditions (M= 1.5, SD= 1.17) with regards to negative sentiment in wall posts, F(2,50)= 3.14, p= .052, η2= .09.

Moderators

Analyses for the moderating hypotheses have been done using the Process tool for SPSS (Hayes, 2012). This tool uses a regression analysis to which you can add a moderator variable. As such the mood condition variable had to be altered, a three-group independent condition variable is not compatible with a regression. Instead it was recoded to a dummy variable, with positive or non-positive (i.e., negative or neutral) as options. To ensure that the control group was also part of this analysis another dummy variable was created with the options neutral or non-neutral (i.e., positive or negative). This has been done as adding a control group dummy allows the interpretation of differences between the mood conditions and the control variable. It should be noted that the Process tool does not allow for two independent variables in a model at a time, as such each regression model will be estimated twice. Once for the positive dummy and once for the neutral dummy. Furthermore, to prevent high multicollinearity with the interaction term, the variables were centered (Aiken & West, 1991)

The first moderator hypothesis predicted gender to moderate the effect that mood has on intention to chat. In the first step intention to chat was the dependent variable and the (positive) dummy variable of condition the independent variable. The model was not significant, F(3,96)= 1.21, p= .312. Next the moderating variable, gender (0= male, 1= female), was added. The interaction was not significant either, b= -.01, t(96)= 1.1, p= .977.

(34)

When the control group dummy was used as independent variable, results were similar, F(3,96)= 1.13, p= .339, also after adding the moderator, b= .23, t(96)= .57, p= .569.

The second moderator hypothesis predicted that incremental and entity beliefs would moderate the effect mood has on intention to chat. To test this hypothesis the scale for incremental and entity beliefs was added to the regression with the positive dummy variable as independent, and intention to chat as dependent variable. The model proved to be

insignificant, F(3. 96)= 1.09, p= .357, neither did the interaction between incremental and entity beliefs and intention to chat, b= .06, t(96)= .25, p= .800. The results with the control group dummy were similar for the model, F(3, 96)= 1.36, p= .26, as well as the interaction b= -.26, t(96)= -1.03, p= .303.

The final moderator hypothesis stated that extraversion would moderate the effect mood has on the use of communicative features of Facebook. Multiple moderation analyses have been performed to test this hypothesis. The dummy variables for the positive and neutral mood condition were the independent variables and extraversion the moderating variable. The dependent variables varied between: chatting, commenting, sharing, and liking.

Most of the analyses turned out to be insignificant (see Appendix E, table 3 for all analyses), with the exception of three variables. The model with the positive dummy variable and chatting, F(3, 96)= 4.54, p= .005, was significant. Meaning that extraversion and the positive dummy variable accounted for a significant amount of variance in intention to chat (R2= .34). The interaction between extraversion and the positive dummy variable was not significant however, b= 1.47, t(96)= 1.69, p= .093. The same applied to commenting with the neutral dummy as independent variable, F(3,96)= 3.11, p= .030, the neutral dummy and extraversion accounted for a significant amount of variance in the commenting variable (R2= .03). The interaction between extraversion and the neutral dummy was insignificant however, b= 1.03, t(96)= 1.57, p= .120. Finally, a significant interaction effect with the neutral dummy

(35)

and extraversion was found, b= 1.48, t(96)= 2.03, p= .046, it should be noted that the model was not significant however, F(3, 96)= 1.41, p= .088. When looking into the conditional effects however, it becomes clear that none of the levels of extraversion show a significant effect. Low levels of extraversion, p= .113, moderate levels of extraversion, p= .928, and high levels of extraversion, p= .253.

Conclusion & Discussion

The aim of this study was to build on previous research on the influences of mood on Facebook use and explore further effects that mood might have on this medium. To this end ten hypotheses have been drafted to answer the following research question: To what extent does the experience of a certain mood affect emerging adults’ Facebook use? Generally results showed very little effect of moods on social media use. Mood did not show the

hypothesized effects on content selection, sharing, tie strength to contacted individuals, social sharing, and wall posts. The same applies to the moderating hypotheses for gender,

incremental and entity beliefs, and extraversion. As such previous findings of mood effects on sharing and commenting (e.g., Berger & Milkman, 2012; Malhotra et al., 2013; Stieglitz & Dang-Xuan, 2013) have not been replicated. There were three exceptions to this however, mood did appear to affect the intention of chat use, liking of the mood evoking content, and the valence of the comments posted as a reply to the mood evoking content.

Contrary to the predictions emerging adults in a negative mood were actually less likely than their counterparts in the neutral condition to use Facebook chat. It is possible that a moderate negative mood does not stimulate a need for conversation while a moderate positive experience does. Luminet IV and his colleagues (2000) found that moderate negative experiences indeed do not stimulate social sharing. It should be noted however that they did not find indications of social sharing after a neutral experience either. As such it is plausible

(36)

that the neutral condition in this experiment was relatively positive. The scores in the pre-test and the high positive sentiment found in the comments on the neutral mood induction do seem to imply this.

The hypothesis on liking wasn’t confirmed either, as emerging adults in a negative mood were actually less likely than participants in a neutral or positive mood to like the mood evoking content. This contradicts findings in earlier studies (e.g., Malhotra et al., 2013). A possible explanation is that liking of such negative content might “feel” contradictory for some (e.g., liking a break up post). Which is one of the reasons Facebook recently expanded the options one can use to respond to a post (e.g., love, sad etc.) (see, Bell, 2016, February 24). The hypothesis on commenting was confirmed however. Participants did show mood congruent sentiment in their comments on the mood evoking content.

Another explanation is for the unexpected findings in the chatting and liking sections of this study is offered by Berger & Milkman (2012). They found that when a message induces a deactivating emotion (such as sadness) people are actually less likely to interact with the stimulus. As such a similar effect could have played a role in this study. While this was not necessarily found in all analyses, the results for liking and chatting as well as most means for other variables did show this trend.

It is possible that the lack of findings can be attributed to the video that perhaps was not intense or involving enough. As multiple studies found or stated, intensity of the

experienced emotion is an important predictor for social sharing (Luminet IV et al., 2000; Rimé, 2009). Due to the nature of moods, which are generally low in intensity and higher in longevity (Forgas,1995), the focus of the video on evoking a certain mood state rather than an emotion could have lead the participants to not feel the need to attend to this mood as it was not intense enough. As such it did not make for an interesting story or was not confusing enough to stimulate social comparison, both not stimulating social sharing (Rimé, 2009) .

(37)

Furthermore, besides the intensity of the mood the type of mood also plays a vital role. Research has shown that more activating emotions such as anxiety, awe, and anger result in higher interaction rates, happiness and sadness do not have this effect (Berger & Milkman, 2012).

There are several theoretical explanations as well however. The most plausible one relates to the previously mentioned AIM model, specifically the affect-as-information principle. Forgas (1995) argues that current mood can completely inform judgement or can have no predictive influence at all. The latter happens when the mood is already attributed to another cause. For this to happen it is essential that the participant realises what the cause for their mood was however (Schwarz & Clore, 1983). Schwarz and Clore (1983) interviewed participants on sunny (positive condition) and rainy (negative condition) days. One condition was made aware of the cause of their mood by stating: “We are interested in how the weather affects a person's mood” (Schwarz & Clore, 1983, p. 519) at the start of the interview, in contrast to the less salient introduction: “By the way, how's the weather down there?" (p. 519). As a result participants who were aware of the cause of their mood did not attribute that mood to the follow up questions on life satisfaction in the interview, while participants who were unaware of the cause of their mood did. This applied only to participants in a negative mood, participants in a positive mood scored similar scores on life satisfaction in both the aware and unaware conditions. While it was not the intention of this study to make the

participant aware of the cause of their mood, the sentence in the introduction might have done so. As can be read in Appendix A, the introduction of the survey states: “the contents of this video might affect your current mood”. The assumption that the participants attributed their mood to the video is further strengthened by the comments they made on the video, such as: “aah this is so sweet:)”, and “What a depressing movie!”.

(38)

Following this line of reasoning it is to be expected that one of the first questions, dealing with commenting on the video which evoked the mood, showed significant results, as it dealt directly with mood evoking content. Allowing the participants to voice their feelings towards the mood evoking content, which could have lead them to discount the mood for the following questions, as the mood was already attributed to the video.

This does not mean however that participants often realise the source of their mood unless it is explicitly stated, as is indicated by studies using rather obvious mood induction procedures (Forgas, 1995), and other studies using videos to induce moods (e.g., Brader, 2005, Westermann et al., 1996)

This leads to several implications for future research. First, if mood effects on the use of general Facebook features are to be measured it might be best to study related affect and unrelated affect separately. Meaning that it might be best to separate studies on what an individual would do with mood evoking content (e.g., commenting or liking) from the studies relating to mood effects on general Facebook use (e.g., chatting, scrolling through the

newsfeed). As questioning what they would do with the content might lead them to attribute the mood to that particular section, if the participant realises

that that section is the source of the mood, thus leaving the rest of the Facebook features unaffected by mood. Doing so will allow research to specify mood effects on features directly related to the mood evoking content as well as features that might not be related, without interference from one another. As a whole this will give a clearer indication of specific conditions required for mood effects to occur.

It is thus recommended that these findings are replicated in future research, while splitting up related and unrelated affect. This will support a

Referenties

GERELATEERDE DOCUMENTEN

However, given the randomization, the observation that mood changes did not vary between the two groups, and the fact that the two alcohol groups did not differ in mean weight,

Abbreviations: TRP, tryptophan; TD, tryptophan depletion; MDE, major depressive episode; MDD, major depressive disorder; BPD, bipolar disorder; HDRS, Hamilton Depression Rating

Other included variables were: free and total plasma Trp levels (% of reduction after ingestion of the AA mixture relative to baseline), gender, number of prior depressive

Effect of solvent, monomer structure, and pressure on reactivity in radical copolymerization : kinetic investigation within the monomer series ethylene - vinyl acetate - vinyl ester

Ze vallen samen als beide lijnen in het grondvlak getekend worden; hebben een snijpunt als P en S in het grondvlak en Q en R in het bovenvlak liggen; lopen evenwijdig als PQ in

Much research is done using experiments aiming at a reduction of depressive symptoms with aerobic as well as anaerobic exercise interventions driven by an

Het raakvlak van de planologie en de politicologie is te vinden bij de maatschappelijke factoren. Waar deze in de politicologie centraal staan, zijn deze in de planologie in

Tweede aspect wat er speelt zijn ruimtelijke ontwikkelingen. Er zijn initiatieven om buitendijks te gaan, een aantrekkelijk gebied voor het ontwikkelen van havens