The Spiral of Silence Online: A Conclusion

After nearly two months working on my project, I am finally ready to draw some conclusions, and reflect on my research process. I encountered some unanticipated challenges throughout — such as the thorny survey logic required to design a 2×2 factorial experiment on Qualtrics — to more anticipated ones, such as learning to speak R. In this post, I would like to present some highlights, and outline some of my possible future steps.

After a couple days of tinkering around in R, I figured out how to make comprehensible data visualizations. Here are a few group means bar graphs of my results:

 

willingness to comment bar

Willingness to comment on a post expressing an opinion about the Ferguson grand jury decision, grouped by opinion.

willingness to like

 

Willingness to “like” a post expressing an opinion about the Ferguson grand jury decision, grouped by opinion.

 

After running some statistical tests, I found no significant difference between the endorsement group conditions and willingness to interact with the post online across the two opinion groups. I was ready to reject my hypotheses. I recognized a constant pattern across all of my dependent variables, however — the social endorsement group (the group with 15 likes on the viewed post) was consistently more likely than the no social endorsement group to report a willingness to interact with the post, or reveal an opinion on the topic, online. Although my results were not significant, I did not think they were likely to be purely due to chance. After conducting a power analysis — I discovered that my experiment was underpowered — that is, my sample size of just under 200 people who answered the necessary questions was too small to detect statistically significant results if there was a true effect.

 

If I could go back in time and do this experiment again, then, I would capitalize on the convenience and speed that the Internet allows for data collection. Instead of stopping at 200 respondents on Mechanical Turk, I would have let the experiment run for a couple more hours — with double the respondents, I would be sufficiently statistically powered to detect results.

 

Thus, I see the future steps for this study as follows:

– Replicate with a larger sample size, able to detect statistically significant results

– Develop operational definitions for Facebook behaviors, and identify their real-life counterparts (“liking” a post means endorsing/supporting a post, but what does sharing it mean?)

– Fully assess the relationship between individual personality traits (willingness to self-censor, for example) and online social behavior

 

I learned a lot while conducting this little survey experiment, and had a lot of fun along the way. Thanks to the Charles Center for providing me the opportunity, and thanks for following along as I explored the research process.

 

Comments

  1. rjdirisio says:

    This is an interesting survey. I thank you for doing this research, as it is fascinating to look at how group dynamics, psychology, and the internet intersect. With social media comes uncharted territory; your work is attempting to bridge the rapid expansion of technology and the quickly expanding field of scientific research. Ad well as that, you are using one of the most powerful and life-changing tools invented by man: the internet. I am happy to see preliminary results to this, and I am proud to see a fellow student pushing the envelope of technology.