I’ve been answering difficult questions about my PhD thesis all day to prepare for my viva on Monday. I thought I’d post a few of them up here because they offer a handy FAQ and, I think, dispel of a few myths about what my research was about:
In one sentence what is my thesis?
I sought to identify why an attitude and a behaviour were adopted by participants in an online community in situations when the structure of the community might not have facilitated it or when the psychological features of influence did not appear to support it.
Why did I do this?
I was introduced to the relevance of online communities for social science inquiry in the early 2000s in my capacity as a technology journalist. I had been involved with the Web for a decade as a user, but it was the phenomenon of online games that really hit the concept home to me: people were developing online identity, meaningful virtual groups and ascribing value to objects in a space that had no physical reality. I realised at that point that this was an environment that shouldn’t have these social phenomena, but that because it did, it could be used to understand our elemental psychological and social selves.
For my Masters I examined this more closely, specifically looking to see if the online psychological experience might have an effect on the offline identity. The evidence there was that there was a positive transfer of self-efficacy, and that virtual world participants’ online experiences increased their self-esteem offline. This presented an interesting question about what else might transfer offline – Attitudes? Beliefs? Behaviours? – and what role online interpersonal interaction might play in influencing the personal attitude offline.
What were the main questions in the research?
There was one main question in the thesis: what does social network analysis offer, in addition to that which we know from social psychological theories, to the prediction of influence in an online community?
This was broken down into three specific questions, one for each study: 1) given the success of ‘network strength’ in predicting pathways of influence in network analytic studies, what does ‘network strength’ mean psychologically? What psychological processes are network analysts not recognising?; 2) How do influential network features (structure, position, strength) work with psychological features of influence (interpersonal and normative attributions, attitudes, experience) in predicting an attitude, and how do they work separately?; 3) how do network features work with psychological features in predicting the adoption of a behaviour, and how do they work separately – and does this change over time?
What were the main findings?
Each study found something different, depending on the question asked:
The first study’s answer was that, depending on the criteria for ‘strength’, network strength measures predict interpersonal and normative features of influence. Specifically in the context of this research, virtual world activity that had the potential to threaten the reputation of the online identity (identity fraud, exposure to offline cues) resulted in the greatest attributions of trust, credibility, social comparison and prototypicality.
Attributions of trust were related to behaviours that implicated social and personal risk. Attributions of credibility were mutually generated in public environments. Social comparison, the most personal attribution of the three, was associated with public performance and identity play. Prototypicality attributions were situated within the online environment, and this finding provided evidence for the presence of online-only norms.
The second study’s answer was that psychological features predict attitudes, network features predict the perception of others’ attitudes and exposure doesn’t predict any similarity between attitudes, despite the strength of the relationship or the interpersonal and normative attributions associated with a connection.
The third study’s answer was that psychological predictors were best at predicting if an innovation would be adopted and when during the innovation’s lifecycle it would occur. The network features supported continued diffusion, but were not predictors in themselves. Innovations progressed through a legitimatisation process that was described by network predictors and resulted in adoption because of positive attitudes towards the service.
With regards to the main research question, these results offered two answers:
First, network and psychological predictors are complimentary and related. Network strength measures predict social psychological attributions, but the psychological attributions and other psychological predictors are better explanations for influence than structural descriptors of social networks. However, network structure and position can be used to predict influential psychological features, and to describe when in a diffusion event un-seen psychological processes are occurring (legitimisation).
Second, features of the online environment under scrutiny were demonstrated to be relevant to the psychological theories that were expected to be compromised by the shortcomings of Computer-Mediated Communication. Specifically, negotiation of the online identity to reduce social risk played a role in attributions associated with offline influence, and online norms played a role in attitude and behaviour uptake. However, the duality between online and offline (virtual identity and person at the computer terminal) affected this: the perception that online others are more similar than they are offline that is projected into the virtual environment created a dependency on offline experience in projecting others’ attitudes and experiences. In other words, there was less accuracy about what others believed or did, and the inaccuracy was based on the offline experience or attitude.
Which topics overlap with your area? What other theories may explain your findings?
Two areas of social network analysis that are focussed on the structural properties of networks might have explained the outcomes. The first is structural equivalence. Structural equivalence would explain the diffusion of innovations and influence when the connections aren’t present by describing ‘spontaneous’ adoption around a network based on the equivalent position in the network, identified by the number and types of connections.
The second is the concept of threshold, outlined in Network Theory. This proposes that adoption of an innovation (specifically in diffusion processes) is as a result of ‘tipping’ an individual’s personal threshold of the number of people connected with him/her who have already adopted. This is a disputed process, as it is applied retrospectively and is related to exposure, which this research suggested didn’t have any significant effect on influence outcomes.
The results would also be explained by the Long Tail theory, an economic model devised by journalist Chris Anderson who uses it to explain why there is rapid uptake in online innovations, but that the majority of the returns come later. The evidence from this study, doesn’t contradict this however; instead it offers an explanation for why the long tail works.
Latane’s Dynamic Social Impact Theory (.pdf) and Social Impact Theory also overlap with the outcomes, and these were briefly mentioned in the thesis. This theory proposes that the number of connections, their strength an their immediacy with the target contribute to group-level influence outcomes. It describes how pockets of attitudes and behaviours form in larger groups, outlining three processes: consolidation, correlation and continuing diversity. The results in Study 2 may have been explained by these processes rather than by pluralistic ignorance and network phenomena, and this was proposed in that chapter. Specifically, the unexpected effects of embeddedness on the accuracy of the group’s average attitude was explained by the DSIT outcome of buffers around the more central person; people with different attitudes are kept out of the awareness of the central person.
What is original about my research?
This research was original theoretically and methodologically.
Theoretically it was original in that it sought to integrate two disciplines in order to understand what it is that makes each successful in predicting influence outcomes, and how they might enhance one another.
It was also original in that it provided support for the research that has maintained the existence of online identity and online norms, but for which there has been little empirical evidence. Further, it demonstrated an influence outcome based upon these constructs.
It applied psychological theories to an online environment that researchers have argued offers a representation of natural human behaviour, rather than experimentally simulated online groups. It was expected that the outcome would be more true to actual behaviour.
Finally, it was theoretically original because of the context in which it was conducted. The contemporary pervasive virtual world is a unique environment in which influence has not been assessed, but which researchers who have examined online influence have called for because of its unique design that encourages the development of strong and perpetuating communities.
Methodologically, it was original because network analysis that has focussed on the diffusion of innovations has not been conducted in a whole network with populations this size, and with such a complex dataset. Not only was this diffusion of actual behaviour, but it was actual behaviour collected over time combined with self-report information. It addressed concerns of previous network and psychological research.