05/07: State

[shevajia]

Comment below with questions or thoughts about the reading for this week's workshop.

Please make your comments by Wednesday 11:59 PM, and upvote at least five of your peers' comments on Thursday prior to the workshop. You need to use 'thumbs-up' for your reactions to count towards 'top comments,' but you can use other emojis on top of the thumbs up.

[weijiexu-charlie]

Thanks for your presentation. I'm curious about your thoughts on how alternative social media such as Instagram, LinkedIn, etc would affect the formation of social networks.

[Panyw97]

Thank you for your presentation! It's pretty interesting, which invokes my memory of research that also explores the college networks through Facebook network data. This is a developing research area since our modern society is filled with various social media, which can provide us with abundant data and information. This topic has much potential to deeply explore!

[JuneZzj]

Thank you for presenting. I am wondering if students from private schools who have a more concentrated social network are facing some potential problems in reaching out of a wider range of society if that is the case. Thanks.

[TianxinZheng]

Thanks for your presentation!

[mingtao-gao]

Thank you for your presentation in advance! I found this paper very interesting to me as I am also learning social network analysis and trying to conduct research in the related field. It is very inspiring to me this research combined different data sources and matched online social platform data with the college scorecard dataset. Besides, creating and using ego-graphs to compare networks in different sizes is insightful to my own research. My question is among all the features of ego-graphs you constructed, why is the relationship between social network features and school characteristics only focused on degree, clustering, and homophily features, especially when eigenvector centralization and degree assortivity appear to be the most distinguishable features?

[cytwill]

Thank you for presenting this paper and I look forward to seeing more explanations and applications of the dataset mentioned in this research. I am wondering if you would like to make it an open-source for other interested researchers. Moreover, as you mentioned that the network built in this research would be helpful to understand people's social network size and patterns for a lifetime, so I am curious about how would you bridge the gap between current findings to tendencies that might happen in the future? That is to say, how to verify the temporal generality of the findings?

[minminfly68]

Thanks for coming to the workshop and sharing your work with us. My question is a little bit similar with prior students. How could students get access to Facebook? I remember that King proposed an excellent framework for cooperation between private and public sector. I am willing to know if you could share about the accessibility to current students. Thx!

[wu-yt]

Thanks for presenting such an interesting topic! For facebook accounts, how do you deal with people who have multiple accounts or people who don't use facebook or even fake accounts?

[ShanglunLi]

Thank you for such an interesting paper to read. I want to ask that how organizations other than fraternity and sorority affect the networks?

[ChivLiu]

Thank you for the presentation! I wonder whether, on mobile phones, the apps would share the information of users such as searching data and chat messages. I had the experience that I emailed my friends some issues, and then youtube and Facebook suddenly popped up ads about the things from my email. Would that information temporarily be stored or it would remain in my user profile as a recorded activity?

[ellenhsieh]

Thank you for this interesting presentation. In the paper, you mentioned that the data of school and class is self-reported. Thus, I am wondering that maybe not everyone would put on their information on their Facebook account so it is possible that we are missing numerous college students in this study. Then, how can you ensure that the result of the analysis can be representative? or can you elaborate more on how do you resolve this issue?

[hihowme]

Thanks a lot for your presentation in advance! Using Facebook data to study the network analysis among college students is truly insightful. I am wondering how would other platform usages would affect the network on Facebook?

[hesongrun]

Thanks a lot for the presentation! It is indeed a very cool topic. The dataset used in your research is quite novel and it is very interesting to see how the network one forms may have some profound impact on study and life. Have you identified some link between the feature of the social network and some social-economic aggregate variables?

[sunying2018]

Thanks for your presentation! I have a question about the similarity measure of networks described in this paper. It trains a random forest classifier on each pair of schools and interprets this pairwise classification accuracy as distance. But why this measure of similarity, instead of other commonly used measures of similarity of networks such as SimRank?

[yalingtsui]

Thank you for your presentation I wonder how the analysis of the college could extend to broader population, what are potential problems if we target at different subgroups.

[romanticmonkey]

Thank you for your presentation! From the result of your research, what do you think social media platform should all strive for? Do you think your result should be educated to the public? If so, how do you think you are going to educate the public?

[Anqi-Zhou]

Thanks for your interesting presentation in advance! As we know, Facebook has the algorithms of referring five friends. Will this kind of mechanism affect our outcome to some degrees?

[anuraag94]

Thanks for sharing this very interesting work with us, and the rich dataset that you've constructed.

Data construction and analysis

• Why did you choose n = 250 as the number of ego graphs for each entry-level cohort
• I notice that you use the modularity of the modularity-maximizing partition as a feature for your ego graphs (Clauset, Newman, and Moore 2004). Did you consider any other modularity-based methods for community detection, like the Louvain method, and why?

The relationship between perception and your measured similarities I think it would be very interesting to consider the perceived similarity between schools by employers, and the similarity as you've measured it using ego graph sampling. How much of the variation in perceived similarity can be explained by your network-based features, versus other school covariates?

Moreover, I know that this is not a huge longitudinal dataset, but if we could measure temporal shifts in both similarity and employer perception, it would be interesting to see what the perception lag looks like as schools shift in network structure over time.

Extending this RF method to temporal/dynamic networks Very broad question, but I'm wondering whether you think there's any intuitive extension of your AUC inverse similarity method to finding similarities between the network dynamics of pairs of temporal graphs.

Discovery of new classes of institutions based on structural similarity You have done a lot of confirmation of separation along the lines of existing school covariates. Do you have any thoughts about how you would reverse the process and discover new classes based on the network similarities that you have found?

[bazirou]

Thank you for your presentation! It's pretty interesting, which invokes my memory of research that also explores the college networks through Facebook network data. This is a developing research area since our modern society is filled with various social media, which can provide us with abundant data and information. This topic has much potential to deeply explore!

[AllisonXiong]

Thanks for your presentation! Personally, i have been curious about user behavior research going on inside FB and it is great to have a chance to take a closer look. My question is, (as mentioned by several people before), the dominance of FB seems to diminish in recent years and even many people continues to build weak connected networks on FB, they may strengthen these links on other platform. How would this trend affect the pattern of social network found in FB?