Relational affordances of information processing on Facebook

G Model INFMAN 2954 No. of Pages 13

Information & Management xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Information & Management journal homepage: www.elsevier.com/locate/im

Relational affordances of information processing on Facebook Ksenia Korolevaa,* , Gerald C. Kaneb a b

Rotterdam School of Management, Erasmus University, Burgemeester Oudlaan 50, 3062 PA Rotterdam, Netherlands Carroll School of Management, Boston College, 140 Commonwealth Ave, Chestnut Hill, MA 02467, USA

A R T I C L E I N F O

Article history: Received 22 July 2015 Received in revised form 6 September 2016 Accepted 27 November 2016 Available online xxx Keywords: Facebook Social network sites Relational affordances Broadcasting Tie strength Information processing Heuristic cues “likes” Comments Post length Perceived usefulness of information Perceived likeability of information

A B S T R A C T

Facebook is increasingly considered as a trusted medium for obtaining news, but the abundance of information on the platform often leads users to experience information overload. Consequently, users need to develop strategies to process information. A survey conducted through a Facebook application reveals that the tie strength of Facebook friends influences how users perceive information on Facebook. Specifically, they appear to rely on heuristic cues (e.g., Facebook “likes” and comments) to process information from weak ties, but these cues are not used when processing information from strong ties. These so-called relational affordances have significant implications for platform design and marketing. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Recent years have seen the emergence of a new form of digital communication technology, commonly referred to as social network sites (SNSs). These have evolved into two main types: the ones which are person based, i.e., centered around the individual user’s personal profile and network (e.g., Facebook and LinkedIn), and content based, where the content is of primary importance (e.g., Pinterest, Instagram; [92]. Although personbased SNSs such as Facebook have been originally viewed only as tools to support interpersonal connections [44], presently, the platforms are used for a variety of purposes. Recently, they have become popular as sources of news: 30% of US adults use Facebook as a source of news and 78% report being exposed to news when using the platform for other purposes [76]. The reason why Facebook has become popular as a news source is that friends act as information gatekeepers, vetting the significance and relevance of content [83]. In the same way as users rely to their friends for product recommendations [33], they also trust them to provide

* Corresponding author. E-mail addresses: [email protected] (K. Koroleva), [email protected] (G.C. Kane).

credible news when they are browsing on Facebook, especially in an age when the status of traditional news sources is declining [3]. The popularity of Facebook as a news source can be in part attributed to the particular form of communication, which the platform offers; it provides a form of many-to-many communication, which enables people to broadcast information simultaneously to their entire network [84]. In the literature, broadcasting of this kind is referred to as communal continuous conversations, because they can be seen by all participants in the network, and are available for asynchronous communication, even though the players involved may change [61]. Broadcasting allows users to communicate more quickly and effectively: they can engage with twice as many people than they can through directed or reciprocal communication [29], and can make contact particularly with those outside their usual circle of friends. Broadcasting also creates trustworthy environments for information exchange [67], provides a starting point for conversations, and allows users to stay in regular contact with a large circle of friends [13]. One result of broadcasting, however, is that users have to process large amounts of information, as the networks of users grow and users become more engaged on SNSs [31]. Most of today’s consumers of news feel overloaded by the amount of news they encounter [42]. Moreover, broadcasting on Facebook can be associated with low-quality interactions, self-promotion, and mendacious behaviors [84]. Thus, the increased quantity and

http://dx.doi.org/10.1016/j.im.2016.11.007 0378-7206/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

2

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

low quality of information can lead users to experience information overload on Facebook, making them feel stressed and anxious [90], which may result in them becoming less active on the platform or even quitting it altogether [91]. SNSs platform designers therefore try to provide users with features that allow them to process information more effectively. We argue that several features of Facebook can help users identify relevant information on the platform: “likes,” comments, post length, and post type. However, how users use these features to process information on Facebook has not yet been sufficiently explored – a research gap we want to fill in this study. In this paper, we consider how the technological features provided by the platform seem to imply a certain homogeneity of use, yet because these features are perceived differently by different users, they are in fact used in a variety of ways [47]. Thus, we would like to uncover the ways in which these features are used. The main reason behind the popularity of Facebook (or any other person-related SNS) as a news source is that information is tied to the person who is sharing it, who acts as trustworthy source of this information [14]. Thus, tie strength with the person sharing the information might influence how a user perceives that information. Although many previous studies have found that SNSs are most valuable for communicating with weak ties [20,72], more recent studies state that users can communicate with similar degrees of efficiency with both their weak and strong ties [2]. As the Facebook platform provides no way of distinguishing between ties of different strengths, we argue that users use the various features differently when communicating with strong vis-a-vis weak ties. As such, they will interact with different frequency, and with different types of information, and will also perceive the platform features differently, depending on the nature of the relationship that they have with the person sharing the information – and we refer to this as relational affordances. Platform features, such as “likes” and comments, can be especially helpful when evaluating information from weak ties as with ties of this type, there is insufficient basis for a trusting relationship, unlike with strong ties. The research question we answer in this paper is: How does the interplay between platform features and tie strength influence the likeability and usefulness of information on Facebook? In order to answer this question, we used a specially designed Facebook application to examine a sample of 810 pieces of content from 135 Facebook users. Our findings suggest that, as expected, Facebook users use platform features to process information from their weak ties, where comments and “likes” play different roles in terms of their impact on perceived usefulness and likeability of information. At the same time, no features are needed to process information from their strong ties. These findings suggest that practitioners need to consider this heterogeneity of platform uses, and that researchers should also consider the concept of relational affordances when theorizing about SNSs. 2. Theoretical background Person-based SNSs such as Facebook are commonly defined as web-based services, which allow individuals to construct a public profile, list their connections to other users, and view and traverse this list of connections and those made by others in the system [11]. This definition focuses solely on the networking functionality, which constitutes the backbone of any SNS. However, the creation and processing of digital content have now become a more important part of any SNS [47]. Merely maintaining a profile on SNSs is not enough: only by actively sharing information can a user exploit the affordances of this technology effectively in order to stay in regular contact with a broad network of friends [44]. To benefit from the information shared on the platform, users have to process that information effectively. However, the

frequency and ease with which information can be shared on the platform results in the sharing of trivial or over-detailed information, often of little interest to others [84]. Moreover, much of the information that users share on SNSs is ambiguous unless one understands its context [51]. With the increasing quantity and varying quality of information, users have to develop strategies on how to process this information in order to avoid information overload. To reduce the cognitive burden, users are known to rely on heuristic cues and bypass certain content [70,16]. Heuristic cues provide shortcuts and allow users to obtain an impression of the information more quickly and effectively [24]. Platform designers therefore build in functionality that can act as heuristic cues, such as “likes” and comments from other users. Our aim is to uncover how users use these cues to evaluate information on Facebook. To develop our conceptual model of information processing on Facebook, we draw mainly on two theories: strength of weak ties [38] and the theory of affordances [57]. The original concept of affordances is rooted in ecological psychology, which states that animals do not perceive what a particular object is, but rather what kinds of uses it “affords” [35]. As such, the same object can be perceived differently by a different set of users completely or in different contexts. Information systems researchers have widely adopted this concept to describe the relationship of users to technology, and, specifically, to explain why different users might use technologies in different ways, and not as originally intended by the designers [57]. Affordances refer neither solely to the material properties of a technology nor to the personal qualities of the people who use it, but to people’s perceptions of the technological artifacts [81]. Technological artifacts stay the same, but because people’s perceptions vary, the affordances of those artifacts can vary depending on the person or context. For example, comments on Facebook can be used to provide support or voice disagreement with the information shared, depending on the context of the communication and how the platform is being used. However, on Facebook as two users are using the same technological feature from two different perspectives, e.g., one providing and another viewing the comment, it is not just about an individual user’s perception of the technology, but it is a combination of affordances of these users. We argue that this relational affordance depends on the strength of relationship between the users and their patterns of communication on the platform; it also determines how the information is perceived, together with other technological features, such as post type and the number of comments and “likes.” The role of relationships in information processing and knowledge acquisition has been widely discussed in the social sciences since Granovetter’s hypothesis on the strength of weak ties. As with the earlier forms of information technology (IT)enabled communication [20,72], SNSs have been regarded as being of greatest value with weak ties as in these environments a tie of this kind requires less effort to maintain [26] and can provide the user with an unprecedented amount of novel information [14]. However, the high frequency of communication with strong ties means that similar amounts of novel information can be transferred on SNSs as can be achieved with weak ties [2]. Overall, we can conclude that SNSs can be used effectively to support both weak and strong ties, and that, on SNSs weak ties play a slightly more important role than they do in face-to-face interactions [31]. One problem that arises, however, is that common SNS functionality does not allow any distinction to be made between the ties of different strengths: on Facebook, one is either a friend or not. In real life, however, relationships are much more nuanced and differentiated by their underlying qualities [10]. Relationship strength is a very dynamic concept, so it cannot be incorporated into the design of a static platform. Therefore, we propose that in

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

order to differentiate among ties of different strengths, users might use platform features differently. As such, depending on the user’s relationship with the person who is sharing information, platform features such as “likes” and comments might differ in terms of their impact on the user’s attitude to that information – uncovering the concept we have called relational affordances. How these different features are perceived for the different types of ties, we will explore in the next section. 3. Conceptual model In our model of relational affordances of information processing on Facebook, depicted in Fig. 1, we explore how users’ attitudes toward information on the platform are affected by both platform functionality and tie strength between the users who are processing and sharing information, and by the interaction between these two elements. The dependent variable in our model is the attitude of a platform user toward information that he or she is evaluating. Most authors differentiate between hedonic (likeability) and utilitarian (usefulness) dimensions of attitude (cf. [1,6], and these distinctions also apply in SNSs [50]. As Facebook is a hedonic information system [86] where users become immersed in the social experience [82], and are therefore likely to evaluate information on the basis of its likeability. Moreover, because users are increasingly using Facebook to obtain news, and news processing is regarded as a cognitive task [42], we want to explore the perceived usefulness of information as well. In our context, usefulness is defined as: “the degree to which the information people encounter is useful for expanding their horizon.” Much as warmth is regarded as a more important factor than competence in our judgments of people [32], likeability will be the primary factor in evaluating information on Facebook. We suggest that a set of platform features will influence the perceptions of likeability and usefulness of information on Facebook. First, in line with the original concept of affordances, we want to explore the impact of platform features on the attitudes of users toward information shared on the platform. Authors generally differentiate between the content and structural features

3

of information [45]; content covers topic, valence, intent, etc., whereas structural features include cues such as the number of words and type of information’ [58]. In the context of SNSs, it is the structural features that play a more important role [5], as users are increasingly more influenced by their interpersonal communication rather than by the actual content of the news (cf. [49,83]. In terms of structural features, various authors have explored the impact of source trustworthiness, expertise and attractiveness, amount of information, number of links, multimedia content, and mentions of other users [59]. In our model, we consider the main features available on Facebook, including the number of “likes,” the number of comments, post length, and post type. In our model, we also include tie strength between the user who shares and the one who processes the information. In almost all studies on information processing, user perceptions of information are influenced by the status of the message poster as a friend [55] or opinion leader [83], and by the trustworthiness of the source [79] or the attractiveness of the person sharing information [59]. We argue that tie strength encompasses all these qualities as it correlates with likeability and depth of knowledge exchanged [65] as well as with the frequency of communication between parties [36]. In order to test the proposed theoretical concept of relational affordances, we add interaction effects between tie strength and the platform features in our model. We suggest that tie strength is the main cue that people use to evaluate the information on Facebook, and this then determines how they perceive the functionality provided by the platform. Specifically, we suggest that, with strong ties, platform features such as “likes,” comments, and post length play much less of a role; however, users increasingly use these features to evaluate information from weak ties. In both models, we also control for the impact of post type on user evaluations. In the following section, we develop our hypotheses. 3.1. Tie strength We have argued that on Facebook both strong and weak ties can be maintained, and that tie strength can help users to evaluate the

Fig. 1. Conceptual Model.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

4

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

information on the platform. Researchers state that in situations with abundant information [15] or in uncertain information environments [66], people give higher priority to information from their strong ties [67]. First, information from strong ties can be processed more efficiently. Users are usually familiar with the context and the subject of communication with strong ties [69]. Second, they regard the information coming from strong ties as being more trustworthy [67], as strong ties are formed in embedded networks where everyone can mutually verify information and thus prevent opportunistic behavior [85]. Third, the homophily theory [64] suggests that users have more in common with their strong ties and are thus more interested in information from them [55]. Fourth, the media multiplexity theory shows that people with a stronger relationship use more functions to communicate and as such will engage with the information from their strong ties more often than that from their weak ties [56]. Moreover, users prefer information from strong ties, despite its content. The balance theory states that people aim to maintain their attitudes, so that the affect valence between any three possible elements multiplies to a positive result [40]. Balance theory is applied to study network triplets [43], for example: if a person likes a celebrity and perceives that this celebrity likes a product, the person will like the product too, to achieve psychological balance [68]. We assume that the same process occurs when evaluating information on Facebook: if a user encounters information from a strong tie, she will tend to form a positive attitude toward this information despite its content or other cues contained in the message, in order to maintain the balance in the attitude formation process. Formally, we hypothesize: Hypothesis 1. The strength of the relationship is positively related to perceived likeability (1a) and perceived usefulness (1b) of information on Facebook.

Facebook: the more “likes” the information receives, the more “socially normative” the information will be and thus the better will be the users’ attitude toward it. We hypothesize: Hypothesis 2. There is a positive relationship among the number of “likes” and the perceived likeability (2a) and perceived usefulness (2b) of information on Facebook. As information on Facebook is always relational, i.e., shared by someone in the user’s network, “likes” are never experienced apart from tie strength with the user who shared the information. Thus, when a user who evaluates the information encounters information that has attracted “likes,” she must evaluate these two cues (even implicitly) in order to form an attitude toward information on Facebook. In the conditions of increasing information flow, Facebook users are more likely to behave as “cognitive misers”: if one cue delivers sufficient information to enable them to form an attitude, other cues might not matter [9]. For example, when processing news online, people are mainly guided by source credibility and they will only consider other cues if the source is not sufficiently credible [78]. We propose a similar effect for Facebook: the stronger is the relationship with the person who shared the information, the less is the impact of “likes” on the resulting attitude. Users can easily evaluate information from strong ties, as they are the people with whom they have shared context, the opinions of other people regarding this information are of little value to them. At the same time, when the person sharing information is a weak tie, “likes” provide normative guidance regarding the status of this information, leading to an increase in the attitude toward it. Therefore, we hypothesize: Hypothesis 2. (continued): The positive effect of “likes” on perceived likeability (2c) and perceived usefulness (2d) decreases the stronger the tie between the users on Facebook. 3.3. Comments

3.2. “Likes” In the online environments, various rating mechanisms are used such as affirmations (e.g., Facebook’s “Like”; Google + ’s +1), binary decisions (e.g., Digg’s “up” or “down”), or continua (Amazon’s 1–5 stars) to help users evaluate products, services, and information. Many SNSs use one-sided ratings in order to prevent the spread of negative feedback, which could discourage users from interacting on the platform. For example, Facebook reports the number of people who have “liked” the information, providing each user with a summary of how many other users positively responded to the information that she has shared. As “likes” are easy to process, they can make certain information more salient to a user within the general information flow [71]. Feedback from others has been found to be valuable for ranking, filtering, and retrieving content [8]. Moreover, “likes” have been found to correlate with positive emotions expressed in a post [74] and indicate the popularity of a certain news item [58]. Reflecting the number of people from the user’s social network who are in consensus with the information shared, “likes” provide an impetus for normative social influence – the user’s desire to adopt a certain behavior in order to be accepted by the social group to which she belongs [19]. Studies have found that user attitudes may be shaped simply by the number of people who support a particular position rather than by the objective quality of information [23]. In online forums, when users encounter information that is consistent with the opinions of other forum users, they are likely to agree with it [18]. On daily deal platforms, the number of people who bought a certain deal has a positive impact on the attitude other people take toward what is being offered [52]. We propose that this process will also apply to

Comments are open-ended feedback mechanisms that enable users to share their opinions on the information shared. Unlike “likes,” which can only be given once on Facebook, comments allow for multiple exchanges between users and thus can increase both the breadth and depth of shared information. If comments all originate from the same person, they increase the depth of communication and thus contribute to the development of shared meaning [65]. If comments all come from different people, they indicate popular concerns and a wider range of social influence, at least in the context of news sharing [58]. Out of the two feedback mechanisms provided by Facebook, we propose that “likes” account for breadth of influence, whereas comments increase depth of sharing. Moreover, in contrast to “likes” – which display the opinions of others that are in consent
comments can reveal that others disagree with the information being shared. Indeed, on Facebook, many comments are related to negative emotions [74] and controversial news [58]. A central premise of our analysis is that processing comments on Facebook requires certain demands, as comments are verbal in nature, can be added multiple times, and cannot be summarized as effectively as “likes.” Moreover, the demands increase with the cognitive complexity of information – i.e., the intensity of information exchanged and the number of people leaving comments in response [80]. Thus, in order to evaluate information that has attracted comments, the user must first evaluate the original information, and then assess the significance and validity of each subsequent comment. Thus, users might experience information overload when they encounter a lot of comments [75]. A greater number of contributions tends to lead to decreasing marginal value [4]: that is, additional comments require similar levels of

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

information processing but produce less insight. Because of time and motivational constraints [77], users might find it too effortful to assess comments and simply consider them as having a negative impact on their attitude. We hypothesize: Hypothesis 3. There is a negative relationship among the number of comments, perceived likeability (3a), and perceived usefulness (3b) of information on Facebook. Similarly, comments cannot be experienced apart from tie strength with the person who shared the information (and potentially with all the people who commented on it). We argued that tie strength is the primary cue that users use when encountering the information on Facebook. Specifically, we propose that the stronger is the relationship with the person who shared the information, the less is the negative impact of comments on the attitude toward information due to several reasons. First, people are more interested in information, which comes from their strong ties and are ready to process more information from them. Second, strong ties are more likely to share a common context for information sharing, and thus any information relating to that context, including comments, will have more meaning for such users. Users are more likely to read a comment thread when the comments are from close friends as these give more insight into those individuals, but they care much less when they do not know the people who are commenting [55]. Therefore, we propose: Hypothesis 3. (continued): The negative effect of comments on perceived likeability (3c) and perceived usefulness (3d) decreases as the strength of tie between the users increases.

5

from weak ties as they do not require a lot of words to understand tacit cues involved in a message [39]. Tacit information may be particularly relevant on Facebook, where people frequently share thoughts that contain implied information (e.g., identities, relationships, and locations) and can convey more information to those users with whom the user shares strong ties. Moreover, due to more frequent interactions that lead to the development of shared meaning [65], strong ties can understand each other’s feelings and intentions easily, without the necessity to process large amounts of information. Therefore, we propose: Hypothesis 4. (continued): In addition, this inverted U-shaped relationship of post length with perceived likeability (4c) and perceived usefulness (4d) will be weaker: the stronger is the tie with the person sharing the information. 3.5. Control variables 3.5.1. Post type Apart from status updates in text form, Facebook offers various multimedia formats, such as pictures, videos, and links. For example, on Facebook picture sharing generates twice as much traffic as the three next largest photo-sharing websites [13]. Compared to text, pictures transfer a great deal of tacit information and allow us to make inferences, for example, about the location of a person. In general, these multimedia features enhance the presentation of information, give it more context, and make the experience of using the site more interactive and rich [21]. Therefore, users might perceive these multimedia features differently and we want to control for them in our model.

3.4. Post length

4. Research method and setting

Post length is a direct measure of information quantity in a post – information input that a user has to process in order to form the attitude toward information on Facebook. Facebook does not have a limit on the number of words that can be used in a post, which results in some users sending very lengthy messages. Scholars across disciplines have found that performance of individuals, e.g., quality of decisions or reasoning in general, initially positively correlates with the amount of information [28]. Higher amount of information might decrease the ambiguity present in the message or give more information about the feelings or the intent of the person who is sharing it. In the context of Twitter, the amount of information has a positive effect on information retweeting [59]. As opposed to Twitter that has a limit on the maximum number of characters that can be used in a message, we expect that on Facebook after a certain point increases in the amount of information are likely to result in decreases in the value of information. This is in line with the information overload hypothesis [73] where information quantity has an empirically verified inverted U-shaped relationship with decision accuracy [17]. Information overload is known to occur in hedonic contexts as well: users of online forums are more likely to respond to shorter messages [46] and users experiencing social overload tend to quit Facebook [62]. Formally, we hypothesize:

To test our hypotheses, users evaluated the information on their Facebook Newsfeed in real time. For this, we designed a specific application that retrieved the information from the Facebook database using Facebook query language [30]: an application programming interface provided by Facebook that uses a structure similar to SQL. To access the study, users logged into their Facebook accounts and installed the application, after which they had to grant explicit permission to access their posts. From the information available on each user’s stream from the preceding 72 h (this is the maximum time frame that the information is available on the Facebook database for third party applications), the application selected six pieces of content at random (the researcher did not have control over this process) and presented them to users for evaluation, one at a time, using an integrated survey tool. Users were asked about the value of this information (likeability and usefulness) and their relationship with the person who had shared it. They were also asked for basic demographic data. In the background, the application collected the type of post (status update, picture, or link), the number of words, and the number of comments and “likes” the post had received up to that point. The responses were collected using snowball sampling. A small group of initial users was recruited through a request to fill out the survey on various Facebook groups and through university mailing lists. Those who responded to the survey also shared the survey request with friends in their network. The participants were motivated to take part in the survey by nonmonetary rewards in the form of scores about the relative value of information on their Newsfeed, which were calculated based on the answers they gave during the survey. In total, 158 people completed the survey, producing 930 observations. After removing respondents who received fewer than six posts (as the panel regression method requires a balanced number of evaluations from each participant), we obtained 810

Hypothesis 4. The length of post is positively related to perceived likeability (4a) and perceived usefulness (4b) of information on Facebook; however, this effect marginally diminishes in the number of words. Again, as the information on Facebook relates to the person who shared it, we suggest that tie strength will play an important role when a user is evaluating the length of post as well. Users can evaluate information from their strong ties much faster than that

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

6

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

observations from 135 respondents for our analysis. Of these, 49% were men and 51% women. Other descriptive statistics for our sample are presented in Table 1. We note that the sample is quite balanced and that the mean values represent an average SNS user: 20–30 years of age, using the network at least once a day, and staying approximately 30 min on the site.

Table 2 Valuation Frequencies. Perceived Likeability Valuation

Description

Freq. Share

Description

Freq. Share

0

Dislike very much Dislike Slightly dislike

31

3.8%

Very useless

200

24.7%

79 126

9.8% 15.6%

177 134

21.9% 16.5%

Slightly like Like Like very much

301 190 83 810

37.2% 23.5% 10.3% 100.0%

Quite useless Slightly useless Slightly useful Quite useful Very useful

180 82 37 810

22.2% 10.1% 4.6% 100.0%

1 2

5. Operationalization of model variables We capture usefulness and likeability dimensions of users’ attitudes on Facebook. In line with [87], in our study, likeability was operationalized by measuring the extent to which people “liked” or “disliked” a post, and the usefulness was measured in terms of how far they perceived it to be either “useful” or “useless.” We measured one item per desired construct because we wanted to make the experience of using our application as close to the actual Facebook experience as possible. Use of single items is a common practice in social network research in order to minimize respondent fatigue from answering highly similar questions [37] about numerous actors [22]. In fact, authors have demonstrated that there is no difference in predictive validity between singleitem and multiple-item measures [7], especially if the attribute object can be conceptualized as concrete and singular. Hedonic and utilitarian dimensions of consumer attitudes can be considered concrete constructs and the items taken (usefulness and likeability) have a similar predictive validity to the multi-item construct [25]. Moreover, using a bipolar scale further reduces the necessity to use multiple-item scales [89]. Each item used a six-point ordinal scale, in which no neutral response option was offered, so users had to make a choice in a particular direction. If given the possibility to answer neutrally, users often prefer this option because they can avoid engaging in a complex information evaluation process [34]. We present the sample response frequencies in Table 2. We considered tie strength, post length, “likes,” comments, and type of post, all of which we hypothesized, would influence user judgments about information on Facebook. For tie strength, we measured the strength of the relationship between the respondent and the user who had provided the information. Therefore, we asked respondents to evaluate how well they knew the person providing the information, using a fivepoint ordinal scale from “very well” to “don’t know at all.” For similar reasons as described above, a single-item scale was used. From Table 3 one can see that, of the information evaluated by users, 34% came from strong ties (people they knew either quite well or very well), whereas 66% came from weak ties (all other contacts). We drew post length as well as the comments and “likes” directly from Facebook, together with every post extracted from the participants’ Newsfeed. Post Length is operationalized as the number of words used by participants are to describe what they were conveying, be it link, picture, or status update. “Likes” were operationalized as the total number of people on Facebook who rated the information positively. Comments are operationalized as the total number of comments that the information attracted at the time of evaluation. The descriptive statistics are given in Table 4.

Table 1 Descriptive Statistics.

Perceived Usefulness

3 4 5 Total

Table 3 Valuation Frequencies of Tie Strength. Description

Frequency

Share

Don’t know (Weak) Hardly know (Weak) Slightly know (Weak) Quite well (Strong) Very well (Strong)

47 154 332 205 72 810

5.8% 19.0% 41.0% 25.3% 8.9% 100%

Table 4 Descriptive Statistics for “Likes” and Comments.

“Likes” Comments Length (number of words)

Mean

SD

Min

Max

2.39 7.01 15

3.87 5.61 26

0 0 0

29 30 270

Moreover, Facebook allows users to share information not only in text form but also multimedia content, such as pictures, links, and videos. As Table 5 shows, on Facebook, there are three different types of posts: status updates, photos, and links, which were randomly drawn from the Newsfeeds of users. 6. Empirical operationalization To test our hypotheses, we construct a model of attitude toward information on Facebook using random-effect ordered probit specification, which allows us to treat respondents’ valuations as ordinal and control for respondent-specific characteristics by including a random effect term. We assume that respondent i’s attitude toward a specific post j, measured on a Likert scale (yij ), depends on the latent variable yij , which is a linear combination of post-specific characteristics, respondent-specific characteristics, and standard normally distributed error term: 0

yij ¼ xij b þ ji þ eij

ð1Þ

0

xij is a K-dimensional row vector, which represents the postspecific characteristics and includes explanatory variables such as tie strength (0–4), “likes” (number), comments (number), post length Table 5 Valuation Frequencies of Post Types.

N = 135

Mean

Min

Max

Number of friends Age Frequency of use Duration of use

236 27 once a day 11–30 min

40 21 almost never up to 10 min

804 55 always in the background >2 h

Status Updates Photos Links Total

Frequency

Share

404 135 271 810

50% 17% 33% 100%

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

(number), its quadratic term, length2, post type_photo (1 = post is a photo, 0 = otherwise), and post type_link (1 = post is a link, 0 = otherwise), as well as the interaction terms: “likes”* tie strength, comments*tie strength, length*tie strength, and length2*tie strength. To minimize potential multicollinearity, variables are meancentered before creating the interaction terms. In order to avoid potential bias in parameter estimates due to multicollinearity, we have checked the variation inflation factors (VIFs) of each item used for analysis. The VIFs across all models (including models with interaction effects) are <4.03, significantly below the common rule of thumb, according to which multicollinearity is considered an issue if VIF > 10 [53]. The relationships between the variables and the attitude toward the information is represented by K-dimensional column vector of coefficients b. In order to control respondent-specific influences, panel data method is applied by including individualspecific random effects ji . The assumed relationship between the observed Likert scale valuation of post j by respondent i, yij 2 f0; 1; 2; 3; 4; 5g, and the latent variable yij is characterized by a set of unobserved cutoff points, fm0 ; . . . ; m4 g. As yij moves beyond a cutoff point, the

observed ordinal variable yij moves up one category. Formally, this relationship is represented as follows: yij < m0 m0 < yij < m1 m1 < yij < m2 m2 < yij < m3 m3 < yij < m4 5 m4 < yij 0

1 2 yij ¼ f 3 4

ð2Þ

Maximum likelihood method is used to jointly estimate the set n o of parameters b; m0 ; . . . ; m4 ; s 2j , where s 2j is the variance of the individual-specific random effects. 7. Results The results of the regression analysis are presented in Table 6. In all the models, our main variable – tie strength – is positively and significantly related (1%) with both perceived likeability and usefulness of information on Facebook (row 1 in Table 6). Thus, we confirm hypotheses 1a and 1b.

7

Results of the regression analysis on control variables (row 6 in Table 6) reveal that photo posts are significantly (1% level) preferred to status updates for both dimensions of attitude. Links (row 7 in Table 6), however, are consistently preferred (at 1% level) to status updates for perceived usefulness, but not perceived likeability of information on Facebook. Examining the second row of Table 6, we see that the number of “likes” correlates positively and significantly (at 1% level) with perceived likeability and perceived usefulness of information on Facebook. Therefore, we confirm hypotheses 2a and 2b. Exploring the interaction effect (row 8, columns 3–4 in Table 6), we find that greater tie strength diminishes the positive relationship of “likes” with attitude, according to the negative point estimates of “likes” * tie strength, significant at the 5% level for perceived usefulness, and 10% for perceived likeability of information on Facebook. This enables us to confirm hypotheses 2c and 2d. Let us explore the impact of each level of interaction among “likes,” tie strength, and user attitude toward information on Facebook. The combined effect of tie strength (s) – measured on a scale from “0” (don’t know a person) to “4” (know very well) – and “likes” (x) on the attitude toward information j by respondent i can be given by:   bx xij þ bxs xij sij ¼ bx þ bxs sij xij ð3Þ With this equation, we can test the composite relationship between “likes” and attitude toward information for different levels of tie strength, using:   H0 : bx þ bxs s ¼ bs ¼ 0   H1 : bx þ bxs s ¼ bs 6¼ 0 for s ¼ 0; . . . ; 4 If H0 = 0, the Wald test statistic follows a x2 distribution with one degree of freedom. Similarly, the impact of each level of interaction between the number of comments and tie strength on the attitude toward information on Facebook are explored. The results of the tests of hypotheses for both likeability and usefulness are presented in Table 7. Tie strength is a significant moderator of the impact of “likes” on both perceived likeability and usefulness of information on Facebook. The last row in Table 7 suggests that the impact of “likes” on the attitude toward information is not statistically significant

Table 6 Regression Results (***p < 0.01, **p < 0.05, *p < 0.1).

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19)

Tie Strength Number of “likes” Number of comments Post Length Post Length2 Photos (w.r.t. status) Links (w.r.t. status) “likes”*Tie Strength Comment*Tie Strength Length*Tie Strength Length2*Tie Strength _cut1 _cut2 _cut3 _cut4 _cut5 rho PRE Pseudo R2

(1) Likeability

(2) Usefulness

(3) Likeability

(4) Usefulness

0.36(0.04)*** 0.05 (0.01)***
0.01 (0.01) 0.01 (0.003)**
0.00 (0.00) 0.37 (0.11)***

0.36 (0.04)*** 0.05 (0.01)***
0.02 (0.01)** 0.01 (0.03)***
0.00003 (0.00002)* 0.44 (0.12)***

0.35 (0.05)*** 0.05 (0.01)***
0.01 (0.01) 0.01 (0.03)*
0.00 (0.00) 0.375 (0.11)***

0.34 (0.05)*** 0.05 (0.01)***
0.02 (0.01)** 0.08 (0.003)**
0.0001 (0.00004)* 0.45 (0.12)***

0.04 (0.09)

0.48 (0.09)***

0.04 (0.1)

0.48 (0.09)**


0.87 (0.07)***
0.13 (0.07)* 0.39 (0.07)*** 1.28 (0.08)*** 2.08 (0.11)*** 0.23 (0.04)*** 15.3% 0.082


0.02 (0.01)* 0.005 (0.007)
0.005 (0.003) 0.00 (0.00)
2.11 (0.11)***
1.32 (0.08)***
0.68 (0.07)*** 0.46 (0.07)*** 1.46 (0.08)*** 0.16 (0.04)*** 12.36% 0.065


0.02 (0.01)** 0.01 (0.008)*
0.006 (0.003)* 0.00 (0.00)
0.91 (0.08)***
0.16 (0.07)** 0.37 (0.07)*** 1.26 (0.08)*** 2.08 (0.11)*** 0.24 (0.04)*** 16.06% 0.085


2.07 (0.11)***
1.29 (0.07)***
0.66 (0.06)*** 0.47 (0.06)*** 1.47 (0.08)*** 0.16 (0.04)*** 11.8% 0.056

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

8

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

Table 7 Results of the tests of hypotheses for different levels of tie strength (***p < 0.01; **p < 0.05; *p < 0.1). “Likes”

Comments

Likeability

0 1 2 3 4

(don’t know) (hardly) (slightly) (quite well) (very well)

Usefulness

Likeability

Usefulness

bs

x2 (1)

sig.

bs

x2 (1)

sig.

bs

x2 (1)

bs

x2 (1)

sig.

0.091 0.071 0.051 0.031 0.011

n.a. 17.83 22.42 5.24 0.25

*** *** *** *

0.1 0.078 0.056 0.034 0.012

n.a. 19.59 24.46 5.68 0.20

*** *** *** **


0.017
0.014
0.011
0.008
0.005

n.a. 1.68 2.39 0.82 0.17


0.06
0.047
0.034
0.021
0.008

n.a. 15.06 17.35 4.1 0.25

*** *** *** **

for the strongest ties. But as tie strength decreases, the impact of “likes” on attitude increases and becomes statistically significant. To illustrate this effect, in Fig. 2 we depict the total estimated composite effects of tie strength and “likes” for perceived likeability of information on Facebook (effect on perceived usefulness would look similar). When we hold everything else constant, the perceived likeability of information is consistently high when the posts come from the respondent’s strongest ties; the number of people who “liked” the information does not change this significantly. However, as tie strength decreases, the marginal impact of “likes” on the perceived likeability increases. At the lowest value of tie strength, the slope of the total effect is at its steepest, revealing the strongest impact of the number of “likes” on perceived likeability of information on Facebook. We find that the number of comments (row 3 in Table 6) is negatively related only with the perceived usefulness of information on Facebook (at 5% level) and insignificant for perceived likeability. We can only confirm hypothesis 3b and reject hypothesis 3a. Exploring the interaction effect of the number of comments and tie strength on attitude (row 9 columns 3–4 in Table 6), we see that increasing tie strength mitigates the negative relationship of comments on attitude, though this moderating effect is significant (10%) only for perceived usefulness of information on Facebook. We can only confirm hypothesis 3c and reject hypothesis 3d. The results of the tests of hypotheses for each level of tie strength presented in Table 7 show that the number of comments barely affects the perceived usefulness of information from users whom the respondent knows very well. However, for those users whom the respondent does not know, perceived usefulness decreases as the number of comments increases. This effect is depicted in Fig. 3. When we hold everything else constant, perceived usefulness is consistently high when the posts come from the respondent’s strongest ties; the number of comments does not change this significantly. However, as tie strength

1.4

Estimated composite coefficient, βs

Tie strength

1.2 1 0.8 0.6

s = 4 (very well)

0.4

3 (quite well)

0.2

2 (slightly)

0

-0.2

1 (hardly)

0

1

2

3

4

5

6

7

8

9 10

0 (don’t know)

-0.4 -0.6 -0.8

Number of Comments

Fig. 3. Interaction Effect of Comments with Tie Strength on the Perceived Usefulness of Information on Facebook.

decreases, the marginal negative impact of comments increases. At the lowest value of tie strength, the slope of the total effect is at its steepest, revealing the strongest (negative) impact of the number of comments on perceived usefulness of information on Facebook. Post length (row 4 in Table 6) initially positively and significantly correlates with perceived usefulness (at 5% level) and likeability (at 10% level) of information on Facebook. However, the fifth row in Table 6 reveals that the quadratic term is negative and significant (at 10% level) only for perceived usefulness of information, thus confirming the inverted U-shaped relationship between post length and information usefulness, confirming hypothesis 4b. The quadratic term is not significant for perceived likeability of information, rejecting hypothesis 4a. This effect is visually illustrated in Fig. 4.

1.8 1.6 1.4 1.2

s = 4 (very well)

1

3 (quite well)

0.8

2 (slightly)

0.6

1 (hardly)

0.4

0 (don’t know)

0.2

Perceived Usefulness

Estimated composite coefficient, βs

6 5 4 3 2 1 0 0

0 0 1 2 3 4 5 6 7 8 9 10

50

100

150

200

250

300

Post Length (Number of Words)

Number of "Likes" Fig. 2. Interaction Effect of “Likes” with Tie Strength on the Perceived Likeability of Information on Facebook.

Fig. 4. Concave Relationship between Post Length and Perceived Usefulness of Information on Facebook.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

Moreover, the interaction effect between post length and tie strength (row 10, Table 6) reveals that tie strength diminishes the initial positive relationship of post length and perceived usefulness of information, significant at 10%; however, the interaction effect with the quadratic term is not significant (row 11, Table 6). It means that the stronger the tie, the less is the impact of post length on the attitude, and after a certain point, the decrease is similar to that of a weak tie. Thus, we can only partially confirm hypothesis 4d. The interaction effect of post length and tie strength on perceived likeability of information is not significant, thus we reject hypothesis 4c. In addition, we find that personal characteristics – as measured by rho (row 17 in Table 6), which indicates the percentage of unexplained variance accounted for by the respondent-specific error component – are more important in determining the perceived usefulness of information than the perceived likeability. The last two rows in Table 6 depict the PRE (proportional reduction in error) and pseudo R2 both of which are used to estimate the fit of the ordered probit model [60]. PRE shows that our model reduces the error by 12% of perceived likeability of information on Facebook and 15% of perceived usefulness of information; by adding the interaction variables, we are able to reduce the error by an additional 0.5%. The last row of Table 6 depicts the pseudo R2 values, which are calculated on the basis of the log likelihoods (difference of log likelihood of the constantonly model and the log likelihood of the full model as the percentage of the log likelihood of the constant only model) and not percentage of variance explained, and thus can only be used for model comparison and not as a measure of fit [63]. We conclude that our model is better at explaining the perceived usefulness than the perceived likeability of information on Facebook. 8. Discussion The objective of this research was to explore how people process information on Facebook. We developed a framework, which explores the impact of platform functionality, tie strength, and their interactions to explain two dimensions of attitude toward information on Facebook: perceived likeability and

9

perceived usefulness. The main finding of our study is that users use platform functionality differently depending on their relationship with the person who shared information on Facebook, which provides support for the theory of relational affordances discussed in our theoretical section. An overview of our findings is presented in Fig. 5. Our findings enable us to formulate various theoretical, managerial, social, and methodological implications. 8.1. Theoretical contributions With our findings, we make significant contributions to several streams of literature for the context of Facebook. We also believe that our findings can be generalized to other person-based SNSs, because of there being similar dynamics in their use. Given that our findings are anchored specifically on the significance of the person sharing the information and his or her relationship with the other user evaluating that information, it may be that with contentbased SNSs different dynamics might play a role. Our first contribution to the literature lies in uncovering the impact of cues on information processing strategies of Facebook users. Our results show that the way in which users process information on Facebook seems to coincide with heuristic processing [70]. Heuristic processing occurs when users bypass the content of information and their attitudes are shaped by other cues present in a message as an attempt to reduce their cognitive burden when processing information [16]. Our results reveal that some of the variance in the perceived likeability, and especially in the perceived usefulness of information on Facebook, are explained by cues such as tie strength between the two people sharing and viewing a given piece of information, the number of “likes,” and comments this information received from other users, post length, and post type. These can be seen as “heuristic cues,” because they serve to reinforce attitude toward information on Facebook without considering other content-related information. These cues might serve as anchors [27], allowing Facebook users to form an instant impression of the information being shared. Specifically, tie strength alone already functions as a heuristic cue: Facebook users prefer information that comes from their strong ties rather than their weak ties, despite its content. This is

Fig. 5. Overview of Results.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

10

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

intuitive, as strong ties have an already established relationship; thus, the information that comes from them, among the plethora of items on the platform, can be processed more easily and is regarded as being more trustworthy. Furthermore, we find that when a piece of information has received a large number of “likes” on Facebook, the user will tend to evaluate this information positively, regardless of its content. “Likes” are easy heuristics as they reflect how many others in the network share the same view, thus increasing the status of the information being shared. Status has been identified as the main determinant of information propagation in networks [31]. However, we find that the impact of heuristic cues is not always positive. For example, when there are more comments attached to a post, Facebook users consider this information as being useless. This is probably because users feel overloaded by all the information available on the platform, they do not have – or want to use – the cognitive resources needed to evaluate all the information contained in a post, and when encountering comments, which need extensive processing, they simply evaluate that information negatively. The increased cognitive complexity involved in processing comments and the difficulty of ascertaining the precise intent of those who post comments can explain why user’s perceptions of these cues are often negative. The main theoretical contribution of the paper lies in uncovering the impact of tie strength on the information processing strategies of Facebook users. The main finding of our paper is that users perceive platform features differently, depending whether information comes from either their strong or weak ties – something which we term relational affordances. Our findings show that if the tie is weak, “likes” have a positive effect on how a user evaluates the information on Facebook (both the likeability and usefulness dimensions), but the effect of “likes” diminishes when ties are stronger, and with the very strongest ties “likes” have no effect at all. This effect can be explained by a lack of trust that usually characterizes weak ties and the absence of a shared context for communication, and by the fact that “likes” may compensate for this missing trust and allow users to process information easier. This also explains why the strength of weak tie hypothesis holds on SNSs such as Facebook [38]. However, due to a large number of weak ties in their networks, the available platform functionality helps users to decide which weak ties are the most relevant to focus on. We see an opposite effect, albeit less strong, when we examine how the number of comments affects perceptions of the usefulness of information on Facebook: comments evoke negative responses if the information comes from a weak tie but there is no such effect with strongest ties. This can be explained by the fact that, on Facebook, users want to obtain as much information as possible about their strong ties, whereas the reverse is true for their weak ties, as has already been found in some research [55]. Users know that finding out a lot of interesting details about strong ties will offset the effort they need to invest in order to process the comments. However, they do not have the same confidence in information they receive from weak ties. One potentially useful contribution to the information overload literature is the interesting effect we observed in relation to how post length influences the perceived usefulness of information on Facebook. We see that post length is initially positively correlated with perceived usefulness, as supported by previous studies [58], but we also show that, after a certain point, the marginal value of each additional word diminishes. This relationship has been previously identified in the literature [73], but to our knowledge, it has not yet been empirically verified in the context of Facebook. At the same time, this inverted U-shaped relationship between post length and perceived usefulness is typically observed with weak ties, again confirming the theory of relational affordances. The

initial positive effect of post length is less for stronger ties (and the quadratic term is not significant), indicating that people require fewer words to understand the intentions behind a message from strong ties. Furthermore, users seem to perceive different types of multimedia content in different ways on the Facebook platform. Our regression results show that users like photos more than status updates and find them more useful – this had already been confirmed in the news processing context [58] and our results now show this to be true also for Facebook. This is intuitive as photos can convey contextual information, including nonverbal cues, which is of more value to the user. We also show that users find links more useful than status updates (but not more likeable). Links usually contain more objective and informative content and thus can be easily understood than status updates that can be quite cryptic. Finally, our results reveal that there are slightly different mechanisms driving the hedonic and the utilitarian dimensions of user attitudes toward information on Facebook – another potentially valuable contribution to the literature. Our results reveal that contrary to our expectations that it is the hedonic dimension of attitude that is more important on Facebook, we are better able to explain the perceptions of the usefulness than the likeability of information on Facebook: (i) users find links more useful, but not necessarily more “likeable” than status updates on Facebook; (ii) comments are perceived to diminish the usefulness of a post, but will not necessarily make people like it any less; (iii) we confirmed the U-shaped relationship between post length and usefulness of information. This might be because likeability evaluations are more subjective, and it is rather the usefulness perceptions that are rather affected by the heuristic cues we have studied. Alternatively, these results might also suggest that there are different types of Facebook uses: the sociable and the utilitarian. For sociable type, the user just wishes to pass time, to get distracted on Facebook, and therefore is not bothered by people posting too much information or engaging in very deep discussions, is less likely to experience information overload and will explore more information. For utilitarian type, the user is targeted at finding useful information from weak ties (as these are more valuable on the network) with least possible effort, and therefore is more prone to heuristic processing. 8.2. Social implications A relational affordance is tacit information that is not readily available on the platform but can be inferred from the interactions that users have. Relational affordances imply that the information that people share, and the relationships they develop, have a reciprocal effect. The stronger the existing relationship between the users concerned, the more positive will be their attitude toward the information obtained through the Newsfeed and the more likely they will be to respond to it. Conversely, by analyzing the interactions people have on Facebook, one can determine the nature of their relationship with one another. This has several social implications. First, using platform functionality differently when interacting with the ties of different type is the means Facebook users use to compensate for there being no way within the platform itself of differentiating between ties of different strengths and signaling to other users is the nature of their relationship with particular members of their network. SNSs have been criticized for being artificial environments for information exchanges where users can construct any image of themselves that they desire [26], and as a result many authors study the dynamics of online and off-line behaviors separately [88]. Our study shows that users online use

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

similar strategies they do in face-to-face encounters, by distinguishing between ties of different strengths and adjusting their communication based on this relationship. This is in line with the finding that people do in fact, which represent themselves truthfully in social media [12]. In our opinion, this also constitutes a new way in which SNSs such as Facebook are contributing to the integration of online and off-line contexts, which was not previously recognized in the literature. Second, responding to this information allows the user to reconfirm the relationship he or she has on the network or even to develop a new one [54]. The information users share on Facebook serves as glue that connects users with each other by constantly reminding them about the presence of others, and in this way allows to activate latent ties on the network: the added value of SNSs such as Facebook. In real life, users are not able to remember all of their weaker acquaintances; however, on Facebook they are constantly exposed to the information from their weak ties, and the platform functionality allows them to determine which ones are more valuable for them. This also confirms the findings that with the help of SNSs such as Facebook, people are able to maintain more relationships than in real life [41]. 8.3. Managerial implications First, we can say that Facebook users aimed to support existing ties rather than to develop new ones. Although this platform delivers a lot of information about weak ties, the processing of this information requires a cognitive effort that users are not always willing to make. Therefore, our recommendation to Facebook platform developers would be to develop functionality that allows users to process information easily. This might require a limit to be set on the number of contacts in an individual user’s network, as is done with the social network Path, or providing an easy-to-use way of communicating with groups of people, as is done on Google+. It might also involve limiting the number of words that can be shared in a post, as is done on Twitter. Moreover, a lot of ties maintained by users always remain latent, and it is hard for users to identify which ties have the most potential from the overall information flow. Developers could develop algorithms that provide recommendations as to which ties are most useful from their weak connections based on the individual preferences of the user. Second, the findings regarding relational affordances provide useful insights on how to improve information filtering algorithms. The current filtering algorithm EdgeRank of Facebook prefers information that comes from users with whom they have already interacted or the one with which other users have interacted, e.g., through “likes.” However, when selecting information to be presented, the algorithm does not consider the relationship between the users (as this relationship is not embedded in the platform). Our study shows that a user might be interested in information from the strong tie, irrespective of the number of “likes” or comments, or in information from a weak tie, but only if it has received a lot of “likes” and few comments. Only showing information with a high number of “likes” might exclude some information coming from strong ties, and only showing the information from those with whom the user has interacted before, might end up presenting only the information from strong ties. On the basis of our findings, we propose a two-stage algorithm: first identifying the tie strength between users, and then using the platform functionality, such as the number of “likes” and comments to select which information to present to the user. 8.4. Methodological contribution The main methodological contribution stems from the fact that our survey was administered using a Facebook application.

11

From a research standpoint, Facebook and other SNS platforms provide substantial benefits in terms of the sheer volume of data available on user activity [48]. However, researchers have to infer the meaning of some of the data, and those inferences may or may not be associated with real-world constructs. With our application-based survey, we can capitalize on the strengths of the voluminous data, which social media platforms provide and also minimize some of the weaknesses. In particular, by obtaining data from the platforms, we can minimize respondent fatigue or respondent inaccuracy. With this kind of approach, we can also ask specific questions that are not covered in the existing data, such as “How well do you know this person?” or “How helpful do you find this information?” With application-based methods, we can combine objective data presented to users for evaluation with subjective responses to some variables that are not available on the network. 8.5. Limitations and further research Our study is subject to several limitations. First, although we believe our application-based survey method is a strength and a key contribution of our research, novel methods do inherently suffer from some drawbacks. For example, we know little about potential response bias: we do not know who has chosen not to take our survey or whether our results may be biased as a result. Perhaps only committed, active Facebook users decided to respond. In addition, those who had stronger ties to the person inviting them to participate in the survey may have been more likely to respond, and people with strong concerns over privacy might not have responded. Although all surveys face similar challenges, because our target is a “typical” Facebook user, we think that this potential bias does not substantially undermine our results. Further research is needed to develop a greater understanding of how to conduct application-based surveys and of the dynamics of snowball sampling. Our second limitation concerns the number of comments. In our study, we do not identify who provided comments, but just use their total number, which might of course have biased the results that we obtained. We also do not consider the valence of the comments, or whether they agreed with or contradicted the original post in some way. Differentiating between the network and the content component in the analysis of comments might allow us to come to more fine-grained conclusions. In order to estimate this, however, one would need to conduct a content analysis of the comments to determine their valence as well as a network analysis to estimate the breadth and depth of the participants’ networks – a possible avenue for further research. Third, we do not account for the content of the posts in our study, although this could increase the variance explained in user attitudes toward information on Facebook, but accounting for this requires more effort, especially as we would also need to develop frameworks to analyze links and pictures as well. However, this is an interesting venue for further research. 9. Conclusion Our study shows that on Facebook, information is the glue that connects users to one another and influences the development of their relationships. We find that Facebook users are adapting the technological functionality to their own relational needs. Specifically, we have coined the term relational affordances to mean that users perceive the platform functionality differently, depending on the strength of the ties they have with the people whose information they are exposed to. This has lead us to formulate several theoretical, social, and managerial implications about the impact of Facebook in our everyday lives.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

12

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx

Acknowledgments We thank Dominik Kunze for designing the Facebook application without which this research would have not been possible. We also thank Vid Stimac, Rodrigo Belo, and Narine Yegoryan for providing guidance on the methodological part of the paper, as well as Hanna Krasnova and Oliver Guenther for their valuable feedback during the initial stages of this research. References [1] I. Ajzen, Attitudes, Personality and Behavior, 2nd edition, McGraw Hill, New York, 2005. [2] S. Aral, M. van Alstyne, The diversity-bandwidth trade-off, Am. J. Sociol. 117 (1) (2011) 90–171. Retrieved from http://www.jstor.org/stable/10.1086/661238. [3] K. Arceneaux, M. Johnson, C. Murphy, Polarized political communication, oppositional media hostility, and selective exposure, J. Pol. 74 (01) (2012) 174– 186. [4] A. Asvanund, K. Clay, R. Krishnan, M.D. Smith, An empirical analysis of network externalities in peer-to-peer music-sharing networks, Inf. Syst. Res. 15 (2) (2004) 155–174. [5] P. Bao, H.W. Shen, J. Huang, X.Q. Cheng, Popularity prediction in microblogging network: a case study on sina weibo, Proceedings of the 22Nd International Conference on World Wide Web, New York, NY, USA : ACM, 2013, pp. 177–178. [6] R. Batra, O. Ahtola, Measuring the hedonic and utilitarian sources of consumer attitudes, Mark. Lett. 2 (2) (1991) 159–170, doi:http://dx.doi.org/10.1007/ bf00436035. [7] L. Bergkvist, J.R. Rossiter, The predictive validity of multiple-item versus single-item measures of the same constructs, J. Mark. Res. 44 (2) (2007) 175– 184, doi:http://dx.doi.org/10.1509/jmkr.44.2.175. [8] J. Bian, E. Agichten, Y. Liu, H. Zha, A few bad votes too many? Towards robust ranking in social media, Proceedings of AIRWeb, Beijing, China, 2008. [9] G. Bohner, G. Moskowitz, S. Chaiken, The interplay of heuristic and systematic processing of social information, European Review of Social Psychology, John Wiley and Sons, 1995. [10] A. Borgatti, D.J. Brass, G. Labianca, Network analysis in the social sciences, Science 323 (5926) (2009) 892–895. [11] D. Boyd, N.B. Ellison, Social network sites: definition, history and scholarship, J. Comput. Mediated Commun. 13 (2008) 210–230. [12] L. Bullingham, A.C. Vasconcelos, The presentation of self in the online world: Goffman and the study of online identities, J. Inf. Sci. 39 (1) (2013) 101–112. [13] M. Burke, C. Marlow, R. Kraut, Social capital on facebook: differentiating uses and users, Proceedings of the Conference on Human Factors in Computing Systems, Vancouver, BC, Canada, ACM Press, 2011. [14] R.S. Burt, Structural Holes, Harvard University Press, Cambridge, MA, 1992. [15] D. Carpenter, K. Esterling, D. Lazer, The strength of strong ties: a model of contact-making in policy networks with evidence from US health politics, Ration. Soc. 15 (4) (2003) 441–440. [16] S. Chaiken, Heuristic versus systematic information processing and the use of source versus message cues in persuasion, J. Pers. Soc. Psychol. 39 (5) (1980) 752–766. [17] Y.C. Chen, R.A. Shang, C.Y. Kao, The effects of information overload on consumers’ subjective state towards buying decision in the internet shopping environment, Electron. Commerce Res. Appl. 8 (1) (2009) 48–58. [18] M. Cheung, C. Luo, C. Sia, H. Chen, Credibility of electronic word-of-mouth: informational and normative determinants of on-line consumer recommendations, Int. J. Electron. Commerce 13 (4) (2009) 9–38, doi:http:// dx.doi.org/10.2753/jec1086-4415130402. [19] C.H. Chou, Y.-S. Wang, T.I. Tang, Exploring the determinants of knowledge adoption in virtual communities: a social influence perspective, Int. J. Inf. Manage. 35 (3) (2015) 364–376. [20] D. Constant, L. Sproull, S. Kiesler, The kindness of strangers: the usefulness of electronic weak ties for technical advice, Org. Sci. 7 (2) (1996) 119–135. [21] R.L. Daft, R.H. Lengel, Organizational information requirements, media richness and structural design, Manage. Sci. 32 (5) (1986) 554–571. [22] J.J.A. Denissen, R. Geenen, M. Selfhout, M.A.G. van Aken, Single-item big five ratings in a social network design, Eur. J. Personal. 22 (1) (2008) 37–54. [23] A. Dennis, Information exchange in group decision making: you can lead a group to information, but you can’t make it think, MIS Q. 20 (4) (1996) 433– 457. [24] A.R. Dennis, S.T. Kinney, Testing media richness theory in the new media: the effects of cues, feedback, and task equivocality, Inf. Syst. Res. 9 (3) (1998) 256– 274. [25] A. Diamantopoulos, M. Sarstedt, C. Fuchs, P. Wilczynski, S. Kaiser, Guidelines for choosing between multi-item and single-item scales for construct measurement: a predictive validity perspective, J. Acad. Market. Sci. 40 (3) (2012) 434–449. [26] J.S. Donath, D. Boyd, Public displays of connection, BT Technol. J. 22 (4) (2004) 71–82. [27] N. Epley, T. Gilovich, Anchoring unbound, J. Consum. Psychol. 20 (1) (2010) 20– 24, doi:http://dx.doi.org/10.1016/j.jcps.2009.12.005. [28] M.J. Eppler, J. Mengis, The concept of information overload: a review of literature from organization science, accounting, marketing, MIS, and related

[29]

[30] [31]

[32] [33]

[34] [35] [36] [37] [38]

[39] [40] [41] [42]

[43]

[44]

[45]

[46]

[47] [48] [49] [50]

[51]

[52]

[53] [54]

[55]

[56] [57]

[58]

[59]

[60]

disciplines, Inf. Soc. 20 (5) (2004) 325–344, doi:http://dx.doi.org/10.1080/ 01972240490507974. Facebook, How Many Friends can You Have? (2010) . Retreived on February 16th, 2016 from https://m.facebook.com/notes/facebook/how-many-friendscan-you-have/72975227130/. Facebook, Facebook Query Language, (2012) . Retreived February 16th, 2016 from https://developers.facebook.com/docs/reference/fql/. L. Feng, Y. Hu, B. Li, H.E. Stanley, S. Havlin, L.A. Braunstein, Competing for attention in social media under information overload conditions, PLoS One 10 (7) (2015). S.T. Fiske, A.J.C. Cuddy, P. Glick, Universal dimensions of social cognition: warmth and competence, Trends Cogn. Sci. 11 (2) (2007) 77–83. L.P. Forbes, E.M. Vespoli, Does social media influence consumer buying behavior? An investigation of recommendations and purchases, J. Bus. Econ. Res. (JBER) 11 (2) (2013) 107–112. H.H. Friedman, T. Amoo, Rating the rating scales, J. Market. Manage. 9 (3) (1999) 114–123. J.J. Gibson, The Ecological Approach to Visual Perception, Lawrence Erlbaum Associates, Hillsdale, NJ, USA, 1986. E. Gilbert, K. Karahalios, Predicting tie strength with social media, Proceedings of Conference on Human Factors in Computing Systems, ACM Press, 2009. S.D. Gosling, P.J. Rentfrow, W.B. Swann Jr., A very brief measure of the Big-Five personality domains, J. Res. Personal. 37 (6) (2003) 504–528. M. Granovetter, The strength of weak ties: a network theory revisited, Sociol. Theory 1 (1983) 201–233. (Retrieved from) http://www.jstor.org/stable/ 202051. M.T. Hansen, Knowledge networks: explaining effective knowledge sharing in multiunit companies, Org. Sci. 13 (3) (2002) 232–248. F. Heider, The Psychology of Interpersonal Relations, Wiley, New York, 1958. R.A. Hill, R.I.M. Dunbar, Social network size in humans, Hum. Nat. 14 (1) (2003) 53–72. A.E. Holton, H.I. Chyi, News and the overloaded consumer: factors influencing information overload among news consumers, Cyberpsychol. Behav. Soc. Network. 15 (11) (2012) 619–624, doi:http://dx.doi.org/10.1089/ cyber.2011.0610. N.P. Hummon, P. Doreian, Some dynamics of social balance processes: bringing Heider back into balance theory, Soc. Netw. 25 (1) (2003) 17–49, doi:http://dx. doi.org/10.1016/s0378-8733(02)00019-9. A.N. Joinson, Looking at, looking up or keeping up with people? Motives and uses of Facebook, Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ACM Press, 2008. D.H. Jonassen, K. Beissner, M. Yacci, Structural Knowledge: Techniques for Representing, Conveying and Acquiring Structural Knowledge, L. Erlbaum Associates Inc., 1993. Q. Jones, G. Ravid, S. Rafaeli, Information overload and the message dynamics of online interaction spaces: a theoretical model and empirical exploration, Inf. Syst. Res. 15 (2) (2004) 194–210, doi:http://dx.doi.org/10.1287/ isre.1040.0023. G. Kane, M. Alavi, G. Labianca, S. Borgatti, What’s different about social media networks? A framework and research agenda, MIS Q. 38 (1) (2014) 275–304. G.C. Kane, R.G. Fichman, The shoemaker’s children: using Wikis for IS research, teaching, and publication, MIS Q. 33 (1) (2009) 1–22. E. Katz, P. Lazarsfeld, Personal Influence: The Part Played by People in the Flow of Mass Communications, Free Press, New York, 1955. B. Kim, Understanding antecedents of continuance intention in socialnetworking services, Cyberpsychol. Behav. Soc. Netw. 14 (4) (2010) 199– 205, doi:http://dx.doi.org/10.1089/cyber.2010.0009. K. Koroleva, H. Krasnova, O. Günther, STOP SPAMMING ME!’ exploring information overload on facebook, Proceedings of the 16th Americas Conference on Information Systems, AIS, Lima, Peru, 2010. K.K.Y. Kuan, Y. Zhong, P.Y.K. Chau, Informational and normative social influence in group-Buying: evidence from self-Reported and EEG data, J. Manage. Inf. Syst. 30 (4) (2014) 151–178, doi:http://dx.doi.org/10.2753/ mis0742-1222300406. N.N. Kutner, Applied Linear Regression Models, 4th editio, McGraw-Hill, Irwin, 2004. C. Lampe, N.B. Ellison, C. Steinfield, A Face(book) in the crowd: social searching vs. social browsing, Proceedings of the 20th Anniversary Conference on Computer Supported Cooperative Work, ACM, 2006. P. Lapides, A. Chokshi, S. Carpendale, S. Greenberg, News feed: what’s in it for me? Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, ACM, New York, NY, USA, 2015, pp. 163–172, doi:http:// dx.doi.org/10.1145/2702123.2702554. A.M. Ledbetter, Media multiplexity theory, Engaging Theories in Interpersonal Communication: Multiple Perspectives, (2014) p. 363. P. Leonardi, When flexible routines meet flexible technologies: affordance, constraint, and the imbrication of human and material agencies, MIS Q. 35 (1) (2011) 147–167. Q. Liu, M. Zhou, X. Zhao, Understanding News 2.0: A framework for explaining the number of comments from readers on online news, Inf. Manage. 52 (7) (2015) 764–776, doi:http://dx.doi.org/10.1016/j.im.2015.01.002. Z. Liu, L. Liu, H. Li, Determinants of information retweeting in microblogging, Internet Res. 22 (4) (2012) 443–466, doi:http://dx.doi.org/10.1108/ 10662241211250980. J.F. Long, Regression Models for Categorical Dependent Variables Using Stata, Stata Press, 2014.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007

G Model INFMAN 2954 No. of Pages 13

K. Koroleva, G.C. Kane / Information & Management xxx (2016) xxx–xxx [61] A. Majchrzak, S. Faraj, G.C. Kane, B. Azad, The contradictory influence of social media affordances on online communal knowledge sharing, J. Comput. Mediated Commun. 19 (1) (2013) 38–55, doi:http://dx.doi.org/10.1111/ jcc4.12030. [62] C. Maier, S. Laumer, A. Eckhardt, T. Weitzel, When social networking turns to social overload: explaining the stress, emotional exhaustion, and quitting behavior from social network sites users, ECIS 2012 Proceedings (2012). Paper 71 Retrieved from http://aisel.aisnet.org/ecis2012/71. [63] D. McFadden, Conditional logit analysis of qualitative choices, Frontiers, (1974) . [64] J. McPherson, L. Smith-Lovin, J.M. Cook, Birds of a feather: homophily in social networks, Ann. Rev. Sociol. 27 (2001) 415–444. [65] S.M. Miranda, C.S. Saunders, The social construction of meaning: an alternative perspective on information sharing, Inf. Syst. Res. 14 (1) (2003) 87–106. Retrieved from http://isr.journal.informs.org/cgi/content/abstract/14/1/87. [66] M.S. Mizruchi, L. Brewster-Stearns, Getting deals done: the use of social networks in bank decision-making, Am. Sociol. Rev. 66 (2001) 647–671. [67] M.R. Morris, J. Teevan, K. Panovich, What do people ask their social networks and why? A survey study of Status Message Q&A Behavior, Proceedings of the Conference on Human Factors in Computing Systems, Atlanda, GA, USA, 2010. [68] J. Mowen, S. Brown, On explaining and predicting the effectiveness of celebrity endorsers, NA – Advances in Consumer Research, (1981) . [69] B.A. Nardi, V.L. O’Day, Information Ecologies: Using Technology with Heart, MIT Press, Cambridge, MA, 1999. [70] R. Petty, J. Cacioppo, D. Schumann, Central and peripheral routes to advertising effectiveness: the moderating role of involvement, J. Consum. Res. 10 (2) (1983) 135–146. [71] G. Pfeffer, G.R. Salancik, The External Control of Organizations: A Resource Dependence Perspective, New York, Harper and Row, 1978. [72] J.L. Pickering, J.L. King, Hardwiring weak ties: interorganizational computermediated communication, occupational communities, and organizational change, Org. Sci. 6 (4) (1995) 479–486. [73] S.C. Schneider, Information overload: causes and consequences, Hum. Syst. Manage. 7 (2) (1987) 143–153. [74] V. Schöndienst, L. Dang-Xuan, A friend indeed? An empirical analysis of interactions in online social networks, Proceedings of the European Conference on Information Systems, AIS., Barcelona, Spain, 2012. [75] U. Schultz, B. Vandenbosch, Information overload in a groupware environment: now you see it, now you don't, J. Org. Comput. Electron. Commerce 8 (2) (1998) 127–148, doi:http://dx.doi.org/10.1207/ s15327744joce0802_3. [76] E. Shearer, 5 Key Takeaways About Twitter, Facebook and News Use, (2014) . Retrieved from http://www.pewresearch.org/fact-tank/2015/07/14/5-keytakeaways-about-twitter-facebook-and-news-use/. [77] H.A. Simon, Satisficing, The New Palgrave: A Dictionary of Economics, Macmillian, London, 1987, pp. 243–245. [78] S. Sundar, S. Knoblock-Westerwick, M. Hastall, News cues: information scent and cognitive heuristics, J. Am. Soc. Inf. Sci. Technol. 58 (3) (2007) 366–378. [79] S. Sundar, C. Nass, Conceptualizing sources in online news, J. Commun. 51 (1) (2001) 52–72. [80] D. Te’eni, Review: a cognitive-Affective model of organizational communication for designing IT, MIS Q. 25 (2) (2001) 251–312. Retrieved from http://www.jstor.org/stable/3250931.

13

[81] J.W. Treem, P.M. Leonardi, Social media use in organizations: exploring the affordances of visibility, editability, persistence, and association, Ann. Int. Commun. Assoc. 36 (1) (2013) 143–189, doi:http://dx.doi.org/10.1080/ 23808985.2013.11679130. [82] L.K. Trevino, J. Webster, Flow in computer-mediated communication, Commun. Res. 19 (5) (1992) 539–573. [83] J. Turcotte, C. York, J. Irving, R.M. Scholl, R.J. Pingree, News recommendations from social media opinion leaders: effects on media trust and information seeking, J. Comput. Mediated Commun. 20 (5) (2015) 520–535, doi:http://dx. doi.org/10.1111/jcc4.12127. [84] J.D.M. Underwood, L. Kerlin, L. Farrington-Flint, The lies we tell and what they say about us: using behavioural characteristics to explain Facebook activity, Comput. Hum. Behav. 27 (5) (2011) 1621–1626, doi:http://dx.doi.org/10.1016/j. chb.2011.01.012. [85] B. Uzzi, Social structure and competition in inter-firm networks: the paradox of embeddedness, Adm. Sci. Q. 42 (1997) 35–67. [86] H. van der Heijden, User acceptance of hedonic information systems, MIS Q. 28 (4) (2004) 695–704. (Retrieved from) http://www.jstor.org/stable/25148660. [87] K. Voss, B. Grohmann, E. Spangenberg, Measuring the hedonic and utilitarian dimensions of consumer attitude, J. Market. Res. (JMR) 40 (3) (2003) 310–320. [88] D. Williams, On and off the ’Net: scales for social capital in an online era, J. Comput. Mediated Commun. 11 (2) (2006) 593–628, doi:http://dx.doi.org/ 10.1111/j. 1083–6101.2006.00029.x. [89] S.A. Woods, S.E. Hampson, Measuring the Big Five with single items using a bipolar response scale, Eur. J. Personal. 19 (5) (2005) 373–390, doi:http://dx. doi.org/10.1002/per.542. [90] C. York, Overloaded by the news: effects of news exposure and enjoyment on reporting information overload, Commun. Res. Rep. 30 (4) (2013) 282–292, doi:http://dx.doi.org/10.1080/08824096.2013.836628. [91] C. York, J. Turcotte, Vacationing from facebook: adoption, temporary discontinuance, and readoption of an innovation, Commun. Res. Rep. 32 (1) (2015) 54–62, doi:http://dx.doi.org/10.1080/08824096.2014.989975. [92] Y.-Q. Zhu, H.-G. Chen, Social media and human need satisfaction: implications for social media marketing, Bus. Horiz. 58 (3) (2015) 335–345, doi:http://dx. doi.org/10.1016/j.bushor.2015.01.006. Ksenia Koroleva is an Assistant Professor at the Department of Technology and Operations Management at Rotterdam School of Management. Her research interests include social network analysis, group communication dynamics, bounded rationality in decision-making, as well as sustainable behaviour of prosumers on the energy market.

Gerald C. (Jerry) Kane is an Associate Professor of Information Systems at the Boston College’s Carroll School of Management. His research interests include exploring the role of information systems in social networks, organizational applications and implications of social media, and the use of IT in health-care organizations.

Please cite this article in press as: K. Koroleva, G.C. Kane, Relational affordances of information processing on Facebook, Inf. Manage. (2016), http://dx.doi.org/10.1016/j.im.2016.11.007