Computer Mediated Communication:
Investigating Self-Disclosure

This project was conducted as part of a course on Computer-Mediated Communications (CMC), taken Spring 2017. Throughout the semester, my teammates Greg, Haewon, Olivia, Sloane and I sought to examine self-disclosure patterns on CMC platforms, defining self-disclosure as a communicative process in which a person shares potentially private information about themself to another person. Our study aimed to examine how people self-disclose information to strangers using different forms of CMC, includng an instant messaging and video conferencing platform.

When creating our study, we were interested in using our topic of study to further examine the following research question: Do people engage more in self-disclosure during video-based conversations or text-based conversations? At the outset, we developed our first hypothesis H1: People are more likely to self-disclose information about themselves over a video conferencing platform as opposed to an instant messaging platform. Video conferencing lacks affordances like revisability and reviewability; as such, participants may be more likely to quickly respond to questions and then expand upon their answers as part of the natural progression of their dialogue.

In seeking to understand the difference in levels of self-disclosure across these platforms, we conducted in-person experiments. Each experiment asked the participant personal questions relating to money or school. In order to gauge the conditions and questions most likely to affect self-disclosure, each question had differing levels of intensity.  While creating these experimental conditions, we developed our second hypothesis H2: In both video conferencing and instant messaging environments, participants are more likely to self-disclose. Because these questions are inherently less-risky, we hypothesized that participants would be more likely to elaborate in their responses.

In addition, we wanted to examine the sentiment of statements made by participants. This led us to develop our second research question: Do people make more positive or negative sentiments when disclosing information about themselves during video-based conversations or text-based conversations? Based on this, we came up with our third hypothesis H3: people are more likely to make fewer negative sentiments over a video conferencing platform than an instant messaging platform. Due to a lack of physical presence and visibility, people may be more comfortable sharing negative information over instant messaging.

Conducting literature reviews throughout this ideation phase also helped to inform our hypotheses. However, many of previously-conducted experiments were done as a comparison between CMC and FtF (Face-to-Face) interactions. In order to delve further into CMC specifically, we decided to look at how levels of self-disclosure may in fact differ across various online platforms, particularly instant messaging and video conferencing platforms, as each of these present a diversity of affordances. We believed that these two forms of communication were unique enough to reveal differences in levels of self-disclosure.

In order to test our hypotheses above, we conducted an interview-based experiment with ten students at Cornell University (nine undergraduate students, and one graduate student). Participants engaged in two different interviews with a confederate. We used this to see whether or not levels of self-disclosure differed as a result of different CMC mediums.

We created two questionnaires, one for each interview. The first questionnaire asked participants about their relationship with money and the second asked about how a participant was doing in school. We categorized the questions based on how invasive they were. Both questionnaires consisted of ten questions: three mild intensity questions, five medium intensity questions, and two high intensity questions. 

When participants arrived, they were asked to read and sign our consent form before they were eligible to begin the experiment. Participants were told they were taking part in a Video Conferencing and Instant Messaging experiment and that they would have two interviews with strangers, one on each medium. Participants were in a separate room from the confederate, using Google Hangouts for each interview. Each participant was randomly assigned one of the two topics for each type of interview (text messaging or video conferencing). The order of each type of interview was also randomized. Participants were told they could either answer a question or simply say ‘pass.’


After completing both interviews in a random order with a random topic, users were asked to complete a post-survey. Participants were then debriefed and the experiment was completed.

In order to assess levels of self-disclosure by participants, we utilized a word count method. Word count was used to determine differences in levels of self-disclosure between video conferencing and instant messaging. It was also used to look at the differences in self-disclosure for questions of varying intensity. We also conducted a sentiment analysis, and analyzed the results of the post-survey. Conducting a word count of our participants’ interviews first involved transcribing the audio of those interviews held over the video conferencing medium. After all interviews had been transcribed they were further examined, during which time interjections by participants, including words such as “uh,” or “um,” were removed. We also removed any response by a participant requesting that the question be passed, as this was a right given to participants upon consenting to partake in our study. Once each interview had been amended to meet these specifications, a count of the remaining words was conducted and averaged.

The average number of words disclosed by participants over video conferencing was 234.9. The average number of words disclosed by participants over instant messaging was 176.2 . This supports H1, that participants are more likely to self-disclose over video conferencing than they are over instant messaging.


In addition to measuring word count disclosed by participants over video conferencing and instant messaging, we also analyzed a smaller subset of questions in order to determine the presence of a relationship between question intensity and levels of self-disclosure. For each set of interview questions (money or school) we chose a question of mild, medium, and high intensity. We then went back through each interview, and recorded the word count of participants as they responded to each question type in both mediums. We then averaged these results across all participants in order to gauge how questions of differing intensity levels impacted levels of subsequent self-disclosure. The data collected supports H2: in each platform participants are more likely to self-disclose information in response to mild intensity questions over medium or high intensity questions. This information is also in line with H1, as for each level of intensity self-disclosure was always greater in the video conferencing platform than it was in the text messaging platform.

Just as important as the number of words a participant emits is the content of what they say . On the highest level, we wanted to measure how positive or negative a user’s responses were between instant messaging and video conferencing. Using sentiment analysis in order to analyze sentence construction and vocabulary, we were able to scale each participant’s responses to a +1.0 and -1.0 range. Each sentence was analyzed using the popular Text Blob library in Python to conduct the analysis. After conducting this analysis, we found that there were a total of 5 instances of negative sentiments in instant messaging, and 9 instances of negative sentiments in video conferencing. As a result, this information rejects H3, as people are more likely to make more negative sentiments over a video conferencing platform than they are over an instant messaging platform.

Despite the data showing that our participants disclosed more on video conferencing than on instant messaging, post-experiment Qualtrics survey results show that the majority of the participants felt more comfortable self-disclosing over instant messaging. The participants were asked to self-evaluate their behavior after the two conversations, and 60% of participants stated to have felt more comfortable disclosing information via instant messaging. Some stated reasons for this sentiment include that the participant could revise their answers, had more time to think of the right way to phrase thoughts and responses, felt safer without eye contact, and felt more anonymous and could therefore say anything they wanted.

Throughout this study, we sought to explore the behaviors surrounding self-disclosure on two different CMC platforms: instant messaging and video conferencing. By observing our filtered data in conjunction with the other statistical data mentioned above, we discovered some interesting points that either confirmed or rejected our hypotheses, as well as afforded new insights into the nature of self-disclosure. The data clearly shows that individuals are more likely to  self-disclose information over a video medium. Across all intensity levels, average word count was significantly higher over video than text. Upon further inspection, the most words disclosed across any interaction type and intensity level are mild questions over a video medium. Consequently, our results supported our first two hypotheses in that participants were more likely to self-disclose over video conferencing rather than instant messaging and that regardless of platform, participants were likely to disclose more for mild intensity questions over medium and high intensity ones. Nevertheless, our third hypothesis was unsupported, as participants had more negative conversations during the video conferencing interview than instant messaging interview.

Potential explanations of this observed result are that the pure nature of FtF interaction, even over video conferencing, reduces the amount of time to prepare a statement than text. The unconscious necessity to respond quickly drives individuals to immediately emit words, despite the fact that they typically are emitted unnecessarily or in a verbose manner.

As we conducted the sentiment analysis on each user’s sentences, the results surprised us. We found that there were only 5 instances of a negative response in instant messaging, and 9 instances of a negative response in video conferencing. We theorize that possible reasons for this increase in negativity is due to an increased sense of ambiguity and misunderstanding over video rather than chat. Given our data, it seems as though people feel less inclined to show their negative opinions over text with a complete stranger. This kind of insight can be observed in the real world in instances where people are soliciting others, such as a job application, or even a Tinder profile. Although it is deceitful, people would still rather hide their negativity until they are face to face.

While our study led to some interesting discoveries, there were some clear limitations that could have inhibited our experiment. For example, given the limitations of time for our experiment, we were only able to achieve 10 participants. Another limitation of our study was in the nature of what we deemed mild, medium and high intensity level questions. We created and categorized questions using our own subjective opinions. It is possible that by changing the questions, the results could change as well. For future studies, a survey could be sent out prior to the study asking participants (separate from the actual experiment) to rate several questions from mild to high intensity, thus quantifying an otherwise moralistic quality.

This study was important because it taught me to realize how critical a platform can be in impacting how a user interacts with others. Not only that, different forms of CMC can also cause people to feel more or less comfortable. In the case of this study, although the video-conferencing and instant-messaging platforms used both existed in an online environment, they still managed to elicit different responses from users, including not only how much users responded but also how they responded (sentiment). In addition, users perceptions and real-world actions did not always align, which marks the importance of using various methods of experimentation to better understand how users operate. Various forms of CMC can bring with them diverse sets of etiquette and norms, which may ultimately alter the user experience, and as such should be seriously considered in the design of new applications and programs. 

Throughout this study I helped conduct initial research, design the study, and write our final report. I participated in many of the studies, as both the research assistant and confederate. I also assisted in transcribing audio, and organizing/analyzing the results of the study, specifically gathering word counts per participant and per question intensity.