Research White Paper

Digital Bereavement: Bringing “back stage”

emotion onto an entirely new “front stage”

More than 400 million people use Facebook to stay connected with others, but when one of these people dies, what happens to their profile? In this digital age, we create an online presence that is shaped over the course of many years; Facebook, LinkedIn, MySpace, Twitter, etc. in a sense become digital blueprints of our lives. Not only do these profiles carry on the memory of the deceased, they also become a valuable anchor for a social support network for grieving friends and family. In October of 2009, Facebook officially announced the ability to “memorialize” Facebook accounts, a feature which protects the privacy of the deceased while allowing friends and family to continue posting on his or her wall. Memorialized Facebook pages provide a new way for mourners to cope with grief, reminisce, relay logistics of funeral services, and even communicate with the deceased. Our research will focus specifically on Facebook, but we intend to be able to apply our conclusions to social networking sites in general. Our goal is to inform social networking sites of ways to facilitate digital bereavement through, for example, improved social support networks and more effective website structure/setup.

RESEARCH QUESTION

When Facebook members post on deceased members’ profile walls/memorial groups, what does their language use tell us about their motivation for posting? Emotions and thoughts traditionally reserved for the privacy of diary pages are now being publicly (or somewhat publicly) posted for many to see, even though these posts are often directly addressed to the deceased. Facebook has thus made it possible for mourners to bring their “backstage” emotions of grief and loss to a digital “front stage,” (Goffman) but what are their motivations to do so?

EXPERIMENTAL DESIGN

Our experiment will include a survey and count and code approach. Survey questions will be based on past similar research and prior questionnaires (i.e. those designed to assess bereavement experience and process). Ideally, we want to target Facebook friends of deceased Facebook users, but decided it was better to survey (via Facebook or email) a wider range of Facebookers. To help increase participation, we will offer an incentive such as a chance to win a gift certificate.

For our other approach, we will be counting and coding based on different categories of wall/memorial group posts. We intend to define motivation categories such as: reminiscence/memories, conversational, longing, disbelief, and logistics. By mapping specific words to our categories, we will be using language to deduce motivation. Furthermore, we will distinguish and analyze potential differences between posts directed to the deceased versus those directed to the social support network at large. The frequency of posts in regards to time (right after the death versus after some time) and the timing of non-immediate posts (whether they tend to fall more on special dates such as birthdays or the anniversary of the death) will also be recorded.

We are contacting Facebook for more information about their decision to include the memorializing feature and for statistics regarding the number of memorialized profiles.

Our research topic is particularly relevant in light of the recent deaths among the Cornell community. As our lives and interactions increasingly become situated online, it is not surprising that online spaces have become a new form of social support.

QUESTIONS

• Do you know of any resources that can help us define categories of Facebook posts and associate words to categories?
• Advice on ways to measure emotions such as grief? We've come across questionnaires/scales measuring bereavement/grief from past research, but none that really 100% fits our experiment. Do you have advice on constructing a more effective survey?
• Have groups in the past had trouble getting enough survey responses? Is there a way that past groups have distributed surveys (especially to the Cornell community) that has been more successful than others? (i.e. email versus Facebook)

Sources

Karas, Tanya. (2010). The psychology of postmortem Facebook. http://www.dailynorthwestern.com/the-weekly/the-psychology-of-postmortem-facebook-1.2139775.

Kelly, Max. (2009). Memories of Friends Departed Endure on Facebook. http://blog.facebook.com/blog.php?post=163091042130.

Statistics: Company Figures. http://www.facebook.com/press/info.php?statistics.

Assignment 3- Online Dating Experiment

Online dating has grown to be a widely used resource for today’s singles. In 2003, 2 out of 5 singles were using an online dating service (Ellison, Heino, & Gibbs, 2006). One issue with online dating is the truthfulness of participants’ self-presentations. In 2001, one study showed that over 25% of users misrepresented some aspect of their identity (ibid). Considering the prevalence of both participation and misrepresentation in online dating sites, we believe there is much to be gained from exploring further the influences of misrepresentations in online dating profiles.

Past research has shown people tend to lie less in communication modes that encompass some form of archived record. The presence of some form of archived information is correlated with lower lying rates (Hancock, Thom-Santelli, and Ritchie, 2004). Expanding upon this idea, we propose an experiment that would explore how the availability of records, in this case a dating profile, might be influential on one's inclination to misrepresent oneself. We hypothesize that users would misrepresent themselves less if they believed others would be viewing their profile (archived information) frequently throughout their interactions.

For this experiment, we would create an online dating service with two different layouts to alter a medium variable. In the first condition, the limited access condition, people could browse others’ profiles freely at first. Then, in order to show interest in someone, they would click something which would bring them to a screen where they could create a message to send to the person of interest. From that point on, clicking on the person and/or profile of interest would bring the user directly to the message string and they would no longer be able to access the other's profile (some users may even suspect this to be an oversight in the functionality of the dating website). In the second condition, the imposed information condition, users again would be able to browse potential mates’ profiles freely and again would click something to show their interest which would begin a string of messages between the two users of the dating site. In this condition, whenever the user wrote messages in the message string, they would be forced to face the profile information as the profile would be above the text box in the message writing screen (much like the way facebook wall text boxes are nested among information about a person).

In order to operationalize the level of a participants' misrepresentation in his/her profile, we would survey those who met face to face with the participant as it would be very difficult to create an objective, generalizable coding scheme for gauging the accuracy of a users' profile. It also measures misrepresentation according to Clark's notion of "what the speaker is to be taken to mean," what is ultimately important in dating profiles. We would ask the survey-taker about how much the other misrepresented subjective qualities about him/herself such as his/her appreciation of humor or love of the outdoors, giving them a copy of their partner's profile to look at while they completed the survey. One factor we must control is how much increased exposure to another’s profile influenced perception. For example, if someone was continually reminded that her partner considered himself ambitious, she may consider him more or less ambitious than if she had only seen this information once. To control for this, we would look at all combinations of pairs, limited access users who dated other limited access users, limited access users who dated imposed information users, and imposed information users who dated other imposed information users. The participants would not know that others had a different layout than their own, so they would not factor this in to their profile creation. If there were enough data that only a subset was needed for significant results, we would use the CDMA data sampling technique of sampling by individual and surveying all those who interacted with individuals within the subset. We hypothesize that limited access users would be rated as misrepresenting themselves far more than the imposed information users.

*Note: We recognize it would be extremely difficult to actually create an online dating service for the purpose of an experiment, but you said our experiment didn’t have to be entirely feasible. A more realistic experiment might involve studying the level of misrepresentation between several preexisting dating sites which allow different levels of accessibility to user profiles.

Sources

Ellison, N., Heino, R., & Gibbs, J. (2006). Managing impressions online: Self-presentation processes in the online dating environment. Journal of Computer-Mediated Communication, 11(2), article 2. http://jcmc.indiana.edu/vol11/issue2/ellison.html

Hancock, J., Thom-Santelli, J., & Ritchie, T. (2004). Deception and design: The impact of communication technology on lying behavior. In E. Dykstra-Erickson & M. Tscheligi (Eds.), Proceedings of the 2004 Conference on Human Factors in Computing Systems (pp. 129-134). New York: ACM.

Status Analysis

In regards to our semantic analysis, our high status emails contained an overall feeling of excitement, topics about education, and an “expected” quality. This is most likely due to the fact that our high status emails were to individuals regarding job searches, and perhaps emails to strangers as well. The words that came up in our high status emails frequently, were: “-ing” words, or action verbs, “was” and “am”. These results are concordant with our semantic analysis results as well. For the most part, I believe when we were crafting our high status emails, we wanted to create a positive perception of ourselves when addressing these high status individuals. Enthusiasm, through the use of “-ing” verbs can help to create a feeling of excitement and interest in either the potential job, mentor, or advice the student is seeking from the high status individual. By using words such as “was” and “am” we are describing ourselves and activities to the high status individual. It also makes sense that these words would be seen less in the low status emails because most communications with low status people are usually to delegate tasks or to maintain a relationship. The student also might not have met the high status individual yet. This could shape the tone and content of the email to include background about themselves in the emails, which would help explain the “am” words as well as the “expected” tone and education topics that were mentioned.

On the other hand, our low status emails covered topics of personal belongings, army/warfare, and people: male and female. Words like “you” were used frequently, and single letters of the alphabet were used more frequently in low status emails. The students who crafted the low status emails might have served as an authority figure over the low status individual. As explained before, "you" words were probably used more to help with the delegations of tasks. This could explain the frequency of the “you” words, which are can be commanding and authoritative. Another take on the frequency of “you” could be the fact that individuals crafting low status emails are more interested in having a rapport and “you” is more conversational. Emails written to lower status individuals might be more playful, and less serious, therefore topics about belongings and gossip (people: male/female) might be mentioned more than they would be in a more serious, high status email. It is interesting to note that weapons came up in our semantic analysis, and when I looked through our emails, I couldn’t find anything that matched up, until I realized Christine is a Film major and used the word “shoot” and “shooting” frequently. This was probably directed towards underclassmen and has nothing to do with weapons, but it is a semantic misunderstanding of her use of “shooting”.

Euphemism Coding

Euphemism Definition:

As defined by Wikipedia, a euphemism is a substitution of an agreeable or less offensive expression in place of one that may offend or suggest something unpleasant to the receiver, or to make it less troublesome for the speaker, as in the case of doublespeak. Typically, these are made of words or expressions that are taboo in normal social settings.

Coding Instructions:

1. Disregard salutation.

2. Unitize emails into separate ideas: Each sentence is a separate idea. Commas separate ideas when they separate independent clauses. Parentheses are also a separate idea when they contain an independent clause.

3. Disregard closings that are not indepedent clauses (i.e. “regards”, “best,” etc.)

4. Code each unit for euphemisms (1 = contains at least one euphemism, 0 = no euphemism)

Reliability:

The results were 99.5% reliable.

Euphemism Examples:

1. I don't think we'll have to worry about it being too spring break-like.

2. KD is not covering dinner.

Confusion Matrix:

	No Euphemism(0)	Euphemism(1)
No Euphmism(0)	212	0
Euphemism(1)	1	3

Testing Testing

Is this thing on? Just checking.