Paper 1: A 61-million-person experiment in social influence and political mobilization
Summary:
The authors study the effects of social influence for effecting behavior change. The results of their randomized controlled trial of political mobilization messages shows that these messages influence users’ political self-expression and real world voting behavior. The results also indicate the importance of social ties in spreading real and online behavior in humans.
Reflection:
The users were unevenly divided into the three groups with number of people in ‘social message’ being much larger than the ones in ‘informational’ and control group. But otherwise the experimental design and the way real life behavior was observed (by checking the public voting records) is amazing!! Also the scale of their experiment is huge!
The authors don’t mention if they were aware of the political leaning of the users or whether the dataset was balanced with equal no. of democrats and republican supporters. I read online that FB received backlash for this research since the dataset is assumed to be skewed towards democrats and it was claimed that the study influenced the voting results since more democrats came to vote.
The authors “arbitrarily” define close friends as the ones lying in the 8th percentile or higher of frequency of interaction among all friendships. I’m not clear about this assumption. Was it just done so that in their dataset 98% of people end up having atleast one close friend (the avg being 10).
One of the premise in the paper is that social ties are crucial in spreading real and online behavior in humans. If that’s the case, can it be exploited to make people come out of filter bubbles. If we are able to identify “close friends” with opposing political ideologies and rank their political posts higher in the feed, will it make people click on the shared content?
Both papers are great examples of how social media can influence human behavior and thus the importance of social computing research.
Paper 2: Experimental evidence of massive-scale emotional contagion through social networks
Summary:
In this paper, the facebook researchers study emotional contagion outside of in-person interactions between individuals by modifying the amount of emotional content in News Feed of about 0.68 million people. The results of the study showed that displaying fewer positive updates in people’s feeds causes them to post fewer positive and more negative messages of their own. People who were exposed to fewer emotional posts in their News Feed were less expressive overall on the following days.
Reflection:
Design of Experiment: Not much information is given about the 0.68 million people selected for the experiment. Gender, age group and demographics can control the results of such an experiment. So knowing these details would have thrown more light on the results.
The researchers determined whether a post is negative or positive depending upon if it contained atleast one positive or negative word. I’m not convinced about this strategy of classifying facebook posts.
People who posted at least one status update during the experimental period were selected for the study. The threshold of one seems too less to claim the emotional state of a user.
Evaluation of results: Have the author’s considered scenarios where some event could have lead people to post more positive/negative posts [e.g home country winning a sport event etc], rather than the effect of emotional contagion.
The way people express their emotions on FB is complex [use of sarcasm, double negatives, images along with text etc.] So I’m not convinced that counting total number of positive and negative words accurately depicts the emotional state of a user. It seems very naïve. [Paper “Does counting emotion words on online social networks provide a window into people’s subjective experience of emotion? A case study on Facebook.” challenges the same belief]
Ethical consideration: The paper says ‘it was consistent with Facebook’s Data Use Policy, to which all users agree prior to creating an account on Facebook, constituting informed consent for this research.’ I wonder if the participants whose feed were affected were personally informed and if they explicitly agreed to be a part of it. If a study is expected to negatively affect someone’s mood, then I believe explicit consent is required.
Overall I’m still skeptical about the main claim of the paper “results suggest that the emotions expressed by friends, via online social networks, influence our own moods”. I believe other’s emotions might affect what we express on social media, but in no way it is a representative of the user’s actual emotional state.
