Charli’s Penn World ….Failure!
http://www.mysocialnetwork.net/blog/mt-tb.cgi/240
I anticipated this project’s outcome for so long and on that dreary Thursday before Thanksgiving it was quite anticlimactic to find that I was let down by my network! I was completely confident that my folder was one of the two which had reached the target. Briefly I will discuss whether my hypotheses were supported in the results, I’ll offer some theories for why my seemingly perfect path deviated (or died rather), and then the class results will be examined and compared to the Stevenson et al. results.
I originally hypothesized that in selecting a weaker interracial tie versus a closer more homophilous tie, that social distance would be more efficiently bridged. This hypothesis was inspired by the Korte and Milgram study on "Acquaintance Networks Between Racial Groups" as well as the Granovetter study exploring the "Strength of Weak Ties." I also hypothesized that since I had more multiplexity with this tie as well as higher frequency and duration of contact with this tie per week that I would have more control over the process, and she would be more motivated to direct the folder to her next link. This is supported by the McPherson et. al study entitled, "Homophily in Social Networks." In having this added multiplexity in the tie I thought I would also have the increased ability to inquire about the whereabouts of the folder and a better ability to ensure that it reached the target. My tertiary hypothesis was that the chain would only cross affiliations one time and that it would be at the time of delivery to the target. Though my folder did not reach the target, it is still interesting to examine the characteristics of each link. This hypothesis was mostly based in the Korte and Milgram study of “Acquaintance Networks Between Racial Groups.”
H1: A Weaker tie can span more social distance and network distance therefore it is the idea first link.
After reading the McPherson et al piece discussing Homophily in Social Networks as well as the Granovetter piece, I was convinced that in order to bridge the social distance that exists between me and the target that a weaker tie would be ideal. McPherson’s argument that homophily limits people’s social worlds was evidenced when I took a step back and examined the make up of my close ties. Each time I asked a close tie for help either they went blank (“What’s Wistar?) or they had no clue who their next link would be. Jim, my other choice for a pass off person had connection with the “science world” here at Penn; however had contact with no one from Wistar.
H2: Multiplexity will increase likelihood of successful path.
My second hypothesis hinged on the idea of multiplexity. The McPherson et. al piece lead me to believe that due to the high multiplexity characteristic of my tie with Erica that she would be motivated to reciprocate my actions. Since we are roughly structurally equivalent in our Track network, the same gender, participate in the same event (throws), and have value homophily I thought that these characteristics made Erica a clear choice for hand-off. I was also relying on the fact that aspects like duration of tie, and frequency of contact would influence the path of my folder.
H3: The folder will only cross affiliations one time from Student to Staff.
Though the Korte and Milgram Study particularly examined the travel of folders to targets of a different race, I think in our small world, rather than race being the “obstacle” per say, bridging affiliations would be the “obstacle” in this case. As observed in the Korte and Milgram study, the most successful chains only crossed races at the target and similarly I thought my chain would only cross affiliations at the target.
Theories for My Chain Attrition:
While I thought I was covering all bases by selecting a tie where there exists multiplexity in the relationship as well as enough social distance for our networks to largely not overlap (weak tie), I thought that there was no way for failure. Unfortunately my chain died after the third link. This puzzled me at first since I remember Erica saying that the woman she handed it off to (Priya) knew someone working in the same lab as Antonio! I mean since we aren’t spanning the country or anything like the Milgram study I thought that 3-4 ties would suffice. If six links can theoretically connect me to anyone in the world, then shouldn’t 3-4 links connect me to anyone on PENN’s 7 block campus? One would think that in such a small geographic area social distance can be easily traveled, however this is not true. As Milgram suggests in his “Small World Problem” piece, communication is dictated less by physical distance and more by social distance. In my case the concentric circles separating my network from Antonio Polley’s proved too difficult to navigate. According to the information I received from Erica (my first link) my chain was one link from completion nearly two weeks before the folders were to be returned. My chain was to travel this path:
Charli-Erica-Priya-“Student Working in Lab”-Antonio
Unfortunately it died after the third link. Perhaps I overestimated the motivation of busy students to pass a folder with no urgent message on to a target. I also perhaps overestimated my links knowledge of her social network. Being that Erica didn’t have a grade depending on the completion of this chain, I’m quite sure that she didn’t give her network as much painstaking thought as I did. As for Priya, the third link, evidenced in the fact that it took her 11 days to ever pass on the folder, as well as the fact that she sent back postcard Roster #10 instead of #3 and dropped it in the inter-collegiate mail system; I’m quite sure that there was little attention paid to instructions or her network structure. Not much careful network visualization was done by this point in the chain and from Priya’s actions it seemed as though this was an inconvenient task.
It seems that the careful thought given to my link was not a proxy for chain success. According to the Killworth study exploring path accuracy and next choice accuracy, nearly 80% of paths deviate from the conceptual most efficient path (this is a high rate of error) and nearly half (48%) of links have next choice accuracy with respect to one’s network. This could be the basis for errors in my chain, and I’m sure that the less thought and time given to the task the higher the rate of error for next choice accuracy. Rather than conceptualizing my chain as a simple 5 person link, I should’ve viewed it with more depth. The five links after all represent 5 discrete social structures.
Class Results:
In examining the class results aggregated, I believe that there are a few trends that were supported in the readings as well as a few trends which are a bit more difficult to conceptualize. It appears that one group; the Susan Yoon group had much more success in relation to the Antonio Polley group. This suggests that the GSE is much closer in social distance to undergraduates than the SM. The mean number of intermediate links was longer for the Yoon group at 4.5 compared to Polley’s 3.25. The percentage of group success was higher at 80% (Yoon) compared to 25% in the Polley group. These startling differences in success rate speak more to the network accessibility of the GSE in comparison to that of the SM. The difference between 4.5 social structures between the participant and target and 3.25 social structures of difference are considerable. When reviewing the literature on past small world studies, a 25% and 80% success rate are particularly impressive. Perhaps this speaks to PENN student networks being systematically better at traversing social distance (when confined to a campus network)?
There is also an interaction with gender which is well documented in both group’s results. The Yoon group had 87.5% of completed chains’ final transfers sharing the gender, while 100% shared gender with their final link (competed chains) in the Polley group. Milgram suggests that certain types of communication are conditioned by sex roles. His Kansas study similarly found that participants were 3 times as likely to pass folders onto those of the same sex. Perhaps this tendency is ingrained in us from childhood. The Killworth study suggests that children are more likely to delete a cross-sex friendship than add one. Therefore there is evidence of conditioning to be gender segregated thus less likely to solve intransitivity across genders. Whether these effects are attributed to nature or nurture is inconclusive, but as children mature, it is also observed that social communication activity is relegated to same sex cliques in adolescents. Even in adulthood (when kin are separated from the mix) many tend to have networks mostly homophilous in terms of sex. This could explain the prevalence of same sex links in our class results.
There was also a marked time difference in terms of mean time between transfers. Since 80% of the Yoon groups’ folders arrived to target I thought it would be meaningful to explore any differences here. At first I thought that the time it takes to navigated social distance is proportional to the network structures separating the link and target. Since the mean number of intermediary links was 3.25 for the Yoon group perhaps this suggests that more time would be needed to traverse and additional network (since Polley’s mean intermediate tie is 4.25). I couldn’t have been more wrong! It is completely puzzling that the Yoon group has a success rate of 80% but it took on average longer to travel, while the Polley group has a success rate of 25% and it took only 4.45 days on average for completed chains to get to the target. The Yoon group has an abnormally high success rate since most literature predicts (or has shown) success rates between 13-33%.
Possible causes for failure in the Polley group could be attributed to the percentage of transfers to people in the same school. Completed chains only passed of the folder to those in the same school 55.5% of the time while incomplete chains passed the folder to those in the same school 71.43% of the time. McPherson would contend that by passing off the folder to a homophilous tie, that the structure of that tie’s network would resemble one’s own. Therefore the reach of that links network wouldn’t be enough to close the social gap between the participant and the target. Since weak ties have access to more diverse information and have different social circles it is much more likely that they could reach the target and bridge the gap. It is interesting to note that for the two successful Polley chains, the chains stayed in the SAS until crossing over to a different school at the target. This provides more evidence that schools at PENN are similar to the race obstacle in the Korte study.
While passing the folder on to a homophilous tie in the Polley group was a detriment, in the Yoon group, 75% of final transfers were within the same school and nearly 40% shared affiliation. Perhaps these findings suggest that the further the social distance between a tie and target, the more one should utilize their bridging social capital through weak ties. When there is less social distance traveled mobilizing homophilous networks works remarkably well.
In comparing our class SW study to the Stevenson study I find that there are several major differences which likely contribute to the differing success rates between results. The first major difference would be the fact that the sampling was randomly dispersed among all four classes of undergraduates. Due to the nature of our class there are only junior and seniors enrolled so our results reflect the trend that students tend to be more engaged with campus networks as they age. Secondly we also were given more difficult to reach targets as opposed an obvious target. The Stevenson study selected the dean of management due to the fact that he was located in a physically accessible and high traffic area. The exact opposite was Dr. Hampton’s goal in selecting esoteric targets that seemed at first like a needle in a haystack here at PENN. The sample size of the Stevenson study is also more than 3 times the size of ours, which has an effect on the interpretation of our statistics. For example, in the Yoon group only two people did not complete the chain therefore the statistics describing incomplete chains are skewed due to low failure rate. The overall success rate for our study between both groups is 56% compared to the 27% success rate of the Stevenson study. There are also idiosyncratic differences in the results like the fact that no student ever passed a folder to a student in a lower year. This was not at all true in our study as at times up to 40% of participants passed the folder to those in a lower year. The mean intermediate links in the Stevenson study is 1.25 and this is in stark comparison to our means of 4.5 and 3.25. Since the target was high profile and easily accessible this difference in chain lengths between groups is understandable. I was however surprised at the low completion rate for these same reasons. I wonder if theses students were graded on this project? If not, problems with motivation could explain some of these results as well.