Extracting Communication Networks - Orientation

[HyunkuKwon]

Post questions about the following orienting reading:

Goldberg, Amir, Sameer B. Srivastiva, V. Govind Manian, William Monroe and Christopher Potts. 2016. “Fitting In or Standing Out? The Tradeoffs of Structural and Cultural Embeddedness”. American Sociological Review 81(6): 1190-1222.

[lilygrier]

I found the dynamic between social embeddedness and cultural fit fascinating here, especially that people with constrained social networks are penalizing for fitting in too much, as they fail to stand out. People can get away with being creative only if they're accepted by the group, in which case they are seen as "edgy" or "avant-garde" rather than just as "weird." Adding the layer of social embeddedness helps distinguish between interesting and weird, especially in a culture that promotes individuality (but doesn't always reward it). While it made sense to use being fired and performance reviews in a workplace context, I wonder how this analysis could be extended to families and other social groups. Do the greater cultural fit pressures facilitated by the workplace mean this results might generalize differently to cases where people get to choose their networks (i.e., friends)? Does the fitting in/standing out tension change based on the extent to which cultural fit is valued across contexts (e.g., middle-schoolers vs. adults)?

[k-partha]

The intuitions behind the 2nd order effects between embeddedness and cultural fit were fascinating to read. However, I am a little concerned that the methods adopted (especially the LIWC based count) are a little 'heavy-handed'/blunt in their approach to assessing cultural fit. I have two concerns here:

1) I feel that work-emails exchanged amongst colleagues are a poor/highly noisy indicator of cultural fit - especially when we have not accounted for any specific work-role synergies in the email-based metrics. Wouldn't informal communications outside work better relate to cultural alignment?

2) Even if emails contain significant cultural information, the semantics of the emails might be most relevant to assessing the cultural fit - raw word counts may be a highly noisy indicator.

Do you see these issues as especially significant? What would a modern approach to processing these forms of communications to extract cultural similarities look like?

[chiayunc]

I find this article particularly fascinating in the sense that it starts off assuming almost diagonal dimensions that are fitness and structural constraint. Even seeing the results, I can't really grasp how such a network would mirror other environments.

My question is in another aspect though. As a non-native English speaker, I oftentimes find myself using English in a different style than others., and am constantly aware of the differences between myself and other people. I wonder if this aspect is taken into account, i.e. gradually non-native speaker is bound to become fitter and fitter, or is it important at all?

This also triggers another question. It is nice to see the network being illustrated, but I think if more mechanisms, such as the one I mentioned above, is presented or discovered, we could probably have more refined understanding of the two forces.

[yushiouwillylin]

This paper is very interesting in the sense that it lies heavily on social network theory, and tries to use computational method to provide quantitative evidence to help create new insights into existing theories. I have a question (not exactly a question but an argument) about the independent variable chosen in this paper.

The authors mentioned they are not aiming for time-dependence arguments, yet it is hard to believe that the language used in a month have links to an involuntary leave happening within that month. A more reasonable argument (in my sense) would be that the worker knew informally he/she was going to be fired with high probability, thus changes his/her communicating strategy, either giving up or wants to overturn the decision. It is thus a bit hard to theoretically argue that this is not actually reverse causality. Specifically speaking, it is not people showing certain embeddedness having higher attainment, but actually people guessing what the HR or group leaders want to see (Especially the main argument stems from a negative attainment instead of a wage raise or something objectively positive).

[william-wei-zhu]

When the authors use emails to measure network embeddedness, do they include incoming emails from outside of the company? A broker that connects multiple internal departments may have very different job performance from a broker that connects the company with external organizations (e.g. clients, partners, etc).

[RobertoBarrosoLuque]

This paper uses email exchanges from a technology company of around 600 employees to validate their hypothesis that there exists a trade-off between cultural embededness and structural embeddedness. Furthermore,  in their conclusions they discuss applying their methodology of measuring cultural fit over time using email data to infer cultural similarity among individuals. What are some constraints both computational and domain specific that should be addressed by the authors in order to see if their results are consistent across different domains?

[romanticmonkey]

I wonder if this study's result is cultural-specific to the U.S., given that the elitist culture is more prevalent in American corporate comparing to the companies in other countries.

[xxicheng]

I have a similar question as @lilygrier, how to generalize the methods to other social groups or relationships? Do you think there will be a noticeable difference between different cultures?

[jcvotava]

The authors write that they were convinced that the firm in question had "strong corporate culture" after speaking to the CEO. My question is: how do we know that the FIRM had "strong culture" and that people in the firm were not just imitating some cultural source outside of the firm, for instance a business guru or something?

If we don't lean on the assumption of a closed company culture, then it seems to me difficult to describe individuals as "fitting in" culturally when someone who appears to not fit in could either be a trendsetter, a cultural deviant, or a perpetual late-adopter.

[Raychanan]

I agree with the authors' definition of cultural fit, but I still have reservations.

First, I believe that the authors' so-called measure of "cultural fit" is simply a result of the company's email tradition. Specifically, companies provide training to their employees, including email etiquette. So when we send and receive internal emails, we take care to write in a way that makes it look professional and matches the company's training requirements as much as possible. When we look at it from the company's perspective, everyone complies with the professionalism and training requirements of the emails, and the so-called "cultural fit" is actually the result of the company's email tradition, and does not reflect the cultural fit of individuals.

Second, I think it would be more interesting to use the team chat messages as data for this study. One advantage is that The amount of data is almost likely to be much larger. Another advantage is that group chat is a relatively freer and more relaxed way to communicate, and people are more likely to express their true opinions in a more natural way, which I think we can capture better information about cultural fit.

[Rui-echo-Pan]

I have a similar concern with Ray on the measure of cultural fit. I also think some categories of passages with formal expressions, such as e-mail, cover letter, and most typically academic paper, have more formal form rather than personal feelings/styles. So my question would be, how is it possible to capture the personal preference/style in such formal corpus?

[MOTOKU666]

This is an interesting topic. I'm just curious about whether high-tech firm samples are able to generalize to other firms. I'm assuming that the way and culture of the high-tech firm may be different from the traditional companies in terms of their expression, frequency of emails, and employee background.

[jinfei1125]

This is a very interesting paper and I enjoy reading it. However, I am a little curious about the ability to get data, as they said:

To protect employee privacy and company confidentiality, we agreed to a number of restrictions in working with the email data. Our resulting data set included approximately 10.25 million messages. In addition to email data, we obtained human resource records that included employee age, gender, and tenure

In my opinion, emails are very private data so the researchers must get a strong affiliation with that company then they are able to get them. But I still find it hard to believe a big company will agree to share their employees' emails with researchers. Can we young researchers get similar data?

[zshibing1]

From Table 1, the descriptive statistics show that while the distributions of network centrality and network constraint are skewed to the left, the distribution of cultural fit is skewed to the right, albeit slightly, as in Figure 2. How would this affect the validity of the analysis and the generalization of the results?

[jacyanthis]

Do you know how the authors got access to this treasure trove of email data? How can we get access to emails?

[sabinahartnett]

Does work done with more informal means of communications (i.e. slack or other instant messaging systems employees can use) show similar results? I am a bit skeptical of the generalizability of email communication.

[ming-cui]

My question goes to the definition of culture fit. Indeed, the authors computed the similarity of language use between received and sent emails. This is a fit, definitely, but is it cultural? It looks more like a generalized fit. This concern leads to another question. If it is cultural fit as what the authors claimed, what about fits in other aspects? Personally, I would theorize a general fit to make conclusions from a holistic perspective. That being said, this article seemed to overlook some other aspects of fit.

[theoevans1]

In suggesting ways that a measure of cultural fit like the one they use could be applied, Goldberg et al. suggest this question: “In the interorganizational context, how culturally compatible are two firms that are contemplating merging or forming a joint venture or alliance?” (1211). I find this question particularly interesting because it compares compatibility between two communities, rather than between individuals and a community. Has there been research using a similar methodology to this paper that looks at cultural compatibility between organizations/communities in that way?

[hesongrun]

I am wondering if the measure constructed with network can account for heterogeneity among staffs in a company? Admins, and HR may send out many emails to many people in the company. Will this network over-weight opinions of these people? If that's the case, is there any methodology we can use to de-bias the estimates? Thanks!

[Bin-ary-Li]

While network constraint and linguistic entropy are well-established measures in their respective fields. I wonder how much does those relate to the author’s ideas of cultural and structural embeddedness. As a reader, I feel like I will benefit more if the author could give some concrete examples to those definitions. What exactly make a person “assimilated broker”? Is a foreign worker who love to connect with people in the office and actively participate in company meetings an “integrated nonconformist”?

[egemenpamukcu]

As someone who is not very familiar with this kind of research, I think I would benefit from a deeper description of the four quadrants as some have mentioned above.

[dtanoglidis]

This paper uses a sample of 10.2 million e-mails from 601 employees in a technological firm to examine the cultural embeddedness of these employees in the firm. Many of my colleagues above have mentioned the limitations of this method, e.g. focusing on only one company is hard to generalize etc.

I want to touch a related aspect. Having a large volume of e-mails but a relatively small number of people these e-mails are coming from may lead us to over-confident statistics (because of the number of e-mails), e.g. make the all-toο-common mistake of confusing precision with accuracy.

[mingtao-gao]

This paper provides a great insight in statistical relationship between cultural fit and network alignment. My question is can word choices capture the cultural perspectives embedding in it? I'm also curious about how cognitive, semantic and emotional categories resembling typical patterns are defined in the model. In this case, it's likely we can get different outcomes with respect to even a slightly different definition of those patterns.

[toecn]

How flexible is the cultural fit variable to accommodate different ways of expressing (e.g., people from different countries, regions, ...)?