As part of my studies I have been introduced to First Monday which is one of the first peer-reviewed open access journals on the internet, about the internet. In an attempt to get to grips with quantitative (measurables) and qualitative (opinioned observation) research methods, our group has been tasked with reading October’s papers and analysing the approaches utilised.
One study presented was Gaydar: Facebook friendships expose sexual orientation (Jernigan and Mistree 2009) especially as I had read coverage of the research and the privacy debates in the mainstream media in recent weeks.
The research utilised the quantitative techniques to profile 4080 Facebook users within the MIT Facebook network. And hypothesis testing was undertaken to establish “a method for accurately predicting the sexual orientation of Facebook users by analysing friendship associations” (Jernigan and Mistree, 2009).
Despite the success of the study (which did accurately predict Facebook user’s sexual orientation based on the information the information collated about their Facebook friends) I think it is interesting to consider some of the “problems” surrounding this methodology (many of which were identified by Jernigan and Mistree).
Equal contact status – This was part of the basis of the study (people having a predisposition to socialise with people “like” them). Although this may predominately be the case it ignores the fact that there are anomalies and not everyone behaves in this way.
Human sexuality – In Facebook users sexuality has to be defined by distinct categories (interested in men, women, both men and women or undefined). In reality sexuality is more complex than this… as the authors describe “the world can not be divided into sheep and goats” (so to speak!) Is it realistic to expect the identity chosen in Facebook to always match the “reality”. Also this means that identities such a transgender remain ignored.
“Friends” – What is a Facebook friend? Many FB users have 100s(+) of so called “friends” within their network. How representative are these connections in comparison to an individuals “sympathy group” which is the average number of people (12) whose death would leave you truly devastated. This piece of research can not get under the skin of this question.
The tools – The recruiting/profiling of each FB account was undertaken by an automated spider (Arachne). Although this helps remove human bias from the research process it also means the subtleties of human interpretation are missed. For example the study can not account for the differences that arose between lesbian and gay male FB relationships. Also the way in which the spider operated meant that it collated data over 18 days in total rather than taking a ‘snapshot’ at one point in time.
Trade offs – The nature of the logistical regression model used means that there is an ongoing balancing act between “specificity , a measure of the model’s susceptibility to false positives, and sensitivity, the probability the model correctly predicts a characteristic” (Signorovitch, 2007 in Jernigan and Mistree, 2009).
Threats to validity – Jernigan and Mistree also identify various threats to the validity of the results of the study, including the selection method (the validation dataset came from profiles known to the authors to be gay males by prior knowledge) and history – National Coming Out Day was held 13 days before the research began.
Ethics – The data for this research was obtained without the knowledge of the FB users, although it was in the public domain and the individuals have been anonimised. This raises many questions regarding privacy and the ethics of collecting information in this way.
In conclusion it can be said that the study can not reveal causation but only correlation. The methodology is useful to obtain an overall picture however human identity and relationships are complex and therefore maybe can not be fully understood utilising these techniques.