Using data to learn about successful online interactions

CDS affiliated prof Paul DiMaggio & co. combine data science with sociology

What makes a successful online interaction? In a working paper that combines sociology with data science, affiliated faculty member at CDS and sociology professor Paul DiMaggio, along with Clark Bernier (Princeton), Charles Heckscher (Rutgers), and David Mimno (Cornell) explore the rules that govern successful online conversations.

Titled “Interaction Ritual Threads: Does IRC Theory Apply Online?” the paper is “the first empirical application of [sociologist] Randall Collins’s theory of interaction ritual chains (IRC) to internal corporate interactions.”

Introduced in 1981, IRC theory has been widely used to predict whether human interactions will succeed according to how much emotional energy those interactions produce.

IRC proposes two things.

First, that people gravitate towards situations where their emotional energy will be enhanced.

Second, that higher levels of emotional energy are associated with “greater buoyancy, confidence, attractiveness, and influence, and shared feelings of conviction and moral rectitude.”

When applying IRC theory to online interactions, however, some stumbling blocks arise. After all, doesn’t Collins state elsewhere that strong human interactions rely heavily on physical co-presence and temporal synchrony — the two ingredients that online conversations lack?

Online forum posting, for example, does not require physical presence or temporal synchrony (since one may post a comment that someone else will not read until ten minutes later).

Yet, as DiMaggio and his researchers argue by way of another theorist, Mikhail Bakhtin, online communication is so distinctive (e.g. acronyms, emojis, hashtags) that it can be considered a family of distinctive speech genres.

So online communication is not merely junk: it comprises many unique language forms.

Anyone who reads or writes on the internet, then, still enters into a meaningful conversation, even without physical presence or temporal synchrony.

With this in mind, DiMaggio and his team applied IRC theory to analyze online interactions in two internal IBM online forums: “The Values Jam” and “The World Jam.”

After extracting over 40,000 forum posts, the researchers cleaned the textual data and used topic modelling to categorize each post according to one of 30 topics. They also tracked the use of pronouns, the response time between posts, and average word length of posts.

The textual data revealed two significant conclusions about what predicted success in these online interactions.

1. Online posts that focus tightly on a specific topic, and sustain that focus over time, are more likely to elicit responses. “Common focus,” the researchers explained, “matters in the way that IRC predicts, if not through precisely the same mechanisms.”

2. The quicker the response time to a post, the more likely it is that someone else will respond as well. Again, temporal rhythm matters “in the way that IRC predicts, even without mechanisms that require co-presence.”

Although more research is needed to examine whether the same rules apply elsewhere, their fascinating study suggests that IRC is a useful theory for assessing online communications.

by Cherrie Kwok

Show your support

Clapping shows how much you appreciated NYU Center for Data Science’s story.