Linguistic Characteristics of English and Albanian Prepositions
3.3 Lexicon and Prepositions in English and Albanian
In 2007, Edwards, Edwards, Qing and Wahl conducted an experiment on the impact of user evalua-tions on the social rating platform ratemyprofessor.com. After watching a video of a lecture, users were asked to rate the respective lecturer’s performance in terms of credibility and attractiveness, the appeal of the lecture and their own learning motivation. Before the video was presented, participants were shown the ratemyprofessor.com profile page of the respective lecturer, which displayed either positive or negative user evaluations (statistical ratings as well as compatible comments). In addition, a control group was set up which had not seen any user ratings. The results showed the user evaluations seen on the ratemyprofessor.com pages affected all dependent variables.
Walther, DeAndrea, Kim and Anthony (2010) manipulated the valence of the user comments that accompanied anti-marijuana PSAs on Youtube. The participants’ attitude towards the PSA varied depending on whether the comments were positive or negative. However, their attitude towards the subject of the PSA - the risk of drug use - was not exclusively affected by the comments. Taking into account the level of identification with the group of commenters as a moderator, a significant effect could be observed; indicating that the attitude towards drugs was in line with the valence of the comments when participants identified with the group of commenters.
A more recent study that dealt with a mechanism similar to the like-feature on Facebook was a large-scale randomized experiment conducted and presented by Muchnik et al. (2013). On a social news aggregation website, where users submit and discuss online news articles, user comments can be evaluated via an up-vote or down-vote, resulting in an aggregated score for each comment (sum of up-votes minus the sum of down-votes) that is displayed along with the respective comments’ content.
Muchnik and colleagues randomly assigned about 100,000 newly (user-)generated comments to one of two treatment groups: up-treatment (comment received 1 up-vote right after creation) or down-treatment (comment received 1 down-vote right after creation). The control group did not receive any manipulated votes. The manipulated comments were then rated 310,000 times by other platform users. Results showed that an up-treatment increased the probability of a subsequent positive rating by 32% (in comparison to the control group). With regard to the down-treatment, Muchnik and colleagues observed that it not only increased the probability of subsequent down-votes
but also that of subsequent up-votes, the latter effect being even stronger. This way, the tendency towards conformity (elicited by the first down-vote) was overridden by a tendency to compensate for negative votes. At the end of the study, the final ratings of the down-treated comments did not differ from those of the control group. In addition to the main effect of treatment, Muchnik and colleagues made use of the fact that on the news aggregation website, users could indicate a positive or negative relationship towards other users (by naming them friends or enemies). Incorporating this into the analysis showed that friends of the commenters whose comments had been manipulated were more likely to show the behavior described above (positive herding when a comment was up-voted, compensating down-votes), while enemies were far less likely to be affected by the experimental manipulation.
There is one study which was concerned with social influence on Facebook and the respective psychological mechanisms and which did not focus on the like-feature in particular, but used a single video post as a stimulus to elicit conformity. However, it is included in this review because of its focus on psychological mechanisms potentially accounting for conformity within the Facebook environment.
Knoll (2013) confronted participants with a mock-up Facebook profile, on which the respective owner had posted a video about a new and largely unknown product. He randomly varied the information given about the profile owner, aiming to persuade participants that they either did or did not have a collective connection to that person or that they had or did not have an interpersonal connection to that person (a 2x2 between subject design plus control condition). Using an existing interpersonal connection, the profile owner was described as a young professor whom participants (students) knew, while the no interpersonal connection condition showed the profile of another professor who was unknown to participants. To manipulate the collective connection, the profiles showed either the logo of the university (with which participants were also associated) or that of an organization unfamiliar to the participants. Overall, the collective connection was found to be effective in terms of social influence, affecting participants’ purchase intention and intention to share the video, while the manipulation of an interpersonal bond did not show any effects.
A similar setting was used in a field experimental approach by Hagen and Hofmann (2013). Colleagues of the researcher posted eWOM (for different battery brands) on Facebook, assessing perceivers’
recall of the brand and respective attitudes by comparing those who had seen the stimulus with those who had not. While they did not find any effect in regard to product attitude, they did observe that the frequency of using the Facebook News Feed, tie strength as well as interaction frequency between the respective users (target and influencer) increased the effect of the stimulus on brand recall.
The first of the studies presented here that explicitly focuses on the effects of Facebook like-displays is a laboratory experiment presented by Bak and Keßler (2012). Confronting participants with neutral mock-up Facebook photo posts, they experimentally varied the number of likes the post had already received. Results showed conformity effects for those participants who used Facebook frequently, as they evaluated the picture more favorably.
Also in 2012, Bond and colleagues published the results of a large-scale randomized controlled field experiment conducted on Facebook in 2010 (during the congress elections in the U.S.), which employed a social plugin (see chapter 2.4) regarding a feature very similar to the like-button. The 61 million participants of the experiment were presented with a message on Facebook that invited them to vote. The message contained the call to action along with a link to a website that provided information on where to vote and a button that said I voted, which users could click. If they did so, the I voted-button was displayed in their own profile. This message was presented in two different versions; the information message was presented as described above while the social message also contained a display that showed how many users had already clicked the I voted-button along with profile photos of friends (Facebook connections) who had done so. A control group did not receive any message. Bond and colleagues assessed three different dependent variables: clicking the I voted-button (as an indicator of political self-expression, only available to the experimental groups), clicking the link (as an indicator for the need to gain more information about the topic, only available to the experimental groups) and the actual voting behavior (obtained by matching Facebook information with voter information). Results indicated that the social message was more persuasive than the informational message regarding all three dependent variables, suggesting that the people shown in the I voted-display exerted social influence. Furthermore, strong ties (defined by interaction frequency on Facebook) were more influential than weak ties. The latter were found to affect political self-expression (clicking the I voted-button), but not actual voting behavior.
The final study to be presented here is another field experiment, conducted in Sweden (Egebark &
Ekström, 2011). It is concerned with the influence of like-displays on Facebook and is thus very similar to the current setting. The researchers manipulated 44 neutral, non-controversial status updates of five associates on Facebook, assigning each post randomly to one of three treatment groups: The status update received 1 like from an unknown user, 3 likes from unknown users or 1 like from a peer (a close tie among one’s own connections, based on degree centrality). Status updates in the control group were not manipulated. Results showed that a single like made by an unknown person did not affect participants’ own like-behavior compared to the control condition. By contrast, both the three likes from unknown users and the single like from a strong tie affected participants’ willingness to like
the post themselves. Regarding the impact of the peer’s like, the target’s level of activity on Facebook was found to increase the effect.