If Facebook Can Predict Breakups, Why Not Your Next Referral Hire?

The big news in the last week seems to be that Facebook can tell you if your relationship is going to work. No, they're not clairvoyant, it's based on research from Jon Kleinberg of Cornell University surrounding somet... Jon worked with Lars Backstrom of Facebook and researchers to organize analysis around this foundational question: given all the connections among a person's friends, can you recognize his or her romantic partner from the network structure alone?

Turns out, you can. To find out, they took data from a large sample of Facebook users - 1.3 million - and developed a new metric of connection strength: `dispersion' --- the extent to which two people's mutual friends are not themselves well-connected. The results offer methods for identifying types of structurally significant people within an individual's life... at least online.

Drawing on the theory of social foci, the researchers argue that dispersion is a structural means of capturing the idea that a friend spans many contexts in one’s social life - either because they were present through multiple life stages, or because they have been intentionally injected into multiple social circles (the way you would with a romantic relationship and/or spouse during the "meet the friends" stage).

Breaking it Down

For awhile now, we've been calling social networks "communities." You have your Twitter community, your Facebook community, etc. Following that logic, each individual's connections (the set of people to whom they are linked) form "network neighborhoods," which have been shown to have important consequences in things like social support based on network make-up and professional opportunities through created a competitive advantage that exists within connection gaps. As people increasingly use their online social networks to manage and even excel within varying aspects of their lives, the structures of their network neighborhoods reflect this.. and the associated complexity of it.

Dissecting the network neighborhood of the average online user typically shows a rather diverse set of relationships comprised of family and various types of friends - from the very close friends from childhood to the "Holiday Card" acquaintance set. Pepper in connections with current and former coworkers, members of religious communities, activities, and potentially romantic partners or the individual's spouse. When we use the available features in the data of all of these connections, we can identify the variation in the types of relationships held within the network neighborhood.

This opens up a wide application for analysis of the interface between an individual and the rest of their neighborhood: in the way they manage their own identities, how they group and identify their connections, and the information they take in from each of the groups in their networks.

Tie Strength and a Rich Tapestry of Connections

What we're looking at is something called "tie strength," which forms an important dimension allowing us to characterize a person’s links to their network neighbors. Sociologists have been actively studying this for year, and have found that the strongest ties are those "embedded" within the network neighborhood through a large number of mutual friends and extensive interaction. 'Weak' ties, by contrast, have fewer mutual friends, but that doesn't mitigate their significance. In fact, one could say it's the crux to successful relationships - because they serve as "bridges" that diversify the network neighborhood. This gives the individual access to broader perspectives, new ways of thinking and novel information - the interesting, anecdotal stuff people like to talk about.

Going back to how Facebook can predict breakups? What this research revealed is that the more well connected a couple's mutual friends were, the more likely they were to break up because it lacked dispersion: those diversified bridges. By basically sharing the same social network neighborhood (high levels of shared "embededness" and low dispersion), they essentially have less to offer because it keeps them from acting as intermediaries (the bridge) between different parts of their networks for their partner. This gives each individual in the relationship value through "if not but for" causation: it opens up new doors and experiences for their partners that they wouldn't have if not but for the relationship. But, the more shared connections a couple have, the lower the dispersement level - the less likely they are to be able to provide that value and the more likely that the relationship will ultimately implode.

Bottom line, people work best when they have their own "lives" and friends, but shared interests and relationships that respect that tend to work out the best.

Applications for Employment

It's a fascinating concept, really, and I can't help but wonder: can this be similarly applied to determine the likelihood of success of referred candidates in employment? I tend to think it can, especially for companies who are still focused on longevity of tenure vs "project-oriented" employment. As a society, we tease each other about having "work spouses" and remark that our teams can become like family because we spend so much time with them. Makes sense, when you think about the fact that we spend approximately 2/3rds of our life at work. We know that referrals often make the best hires... because the relationship with the referrer brings with it loyalty and often an increased level of performance (at least initially). The referrer doesn't want to look bad for recommending the candidate for a role and so has an investment in the relationship. The referred hire has a connection and relationship in the company which eases transition, and doesn't want to let the referrer down, so perhaps tries a bit harder in their work.

We've long since established there's a benefit to this... but when do the scales tip? Is it possible that having too many shared connections between 2 individuals at work could lead to a similar implosion in their working relationship, causing turnover? If so, could we use the dispersion theory pre-hire to analyze if there's too many shared connections to have a sustainable working relationship?

Should we be looking at the connection density and identifying useful "bridge connections" of our employees, who exist at healthy dispersion levels, as part of sourcing to produce better cultural fits and hires for our organizations? To some extent, we already are, but I think this dispersion metric and the methods to analyze it, has real application for those in Talent Attraction & Management. As making sense of "big data" continues to trend for both HR Technologists and organizations alike, being able to maximize recursive dispersion for better referral hiring certainly seems like something that bears a closer look.

As always, I would love to hear your thoughts.

Photo Sources: Top Left: Reuters/Dado Ruvic, Middle Right: 123rf.com

Comment by Keith D. Halperin on November 5, 2013 at 12:46pm: Thanks, Crystal. I think this sort of analysis has great potential. At the same time, putting a great deal of time and effort to determine how long someone(s) will stay seems to be less effective then setting up incentives for the people you want to stay to remain, and assuming that if you DON'T have something like that for them, there's a reasonable chance they'll go sooner or later... In other words: tie them down with a golden chain if you don't want them to wander off. The exact composition of each "goden chain" is something well worth researching....

Cheers,

Keith

Comment by Recruiting Animal on November 5, 2013 at 4:01pm: I thought that couples with common interests have the most chance for success. If they aren't interested in the same things they are boring to eachother.

But you're saying that couples who simply duplicate eachother's interests are boring - and that partners who can share information from outside each person's area of expertise are likely to have the strongest ties to eachother.

I don't know if that's true. Baseball fans want to hang around with other baseball fans - not with people who like badminton.

Likewise, you're saying that work teams that have diverse interests are the strongest because there is always new, interesting information coming in. That sounds reasonable but only if the main interest is the same and these diverse interests simply enrich it.

So, overall, this seems wrong to me. Unless diversity is the minor key and shared interests the major basis of the relationships.

Comment by Crystal Miller on November 5, 2013 at 5:12pm: Animal,

What I'm saying goes beyond "boring" - it's sustainability that's attained through a combination of shared interests but the existence of variance when it comes to each individual's connections and experiences. In relationship terms, think of it like this:

If your significant other has all of the same friends you do, is connected to you online, AND also has all of the same online connections; so, there's really very little for you to discuss in terms of social/people interaction (which is a major topic of conversation for most people). You each had equal opportunity to see it. You probably don't need to ask how one another's day is... you saw it in status updates/tweets/etc. I touch on a this in a blog post I wrote last year, as well - before dispersion had been published: http://www.theonecrystal.com/2012/03/14/my-relationship-status-its-...

Baseball fans want to hang out with other baseball fans - but they don't necessarily all root for the same team. Some of what makes that fun is the competitive angle that exists in the "connection gap" of liking different teams. Same general interests, diverse in fandom?

In relation to work, what I'm saying is that if the dispersion model holds true - and it makes sense to me - then the same thing that could potentially cause a romantic relationship to implode could happen in the workplace - could too many shared connections could ultimately create a suffocating feeling rather than that "warm fuzzy" feeling we want employees to have with referral relationships?

Comment by Crystal Miller on November 5, 2013 at 5:29pm: Keith,

I think the bulk of the time spent invested needs to be associated with sourcing/recruiting as to Talent Management - perhaps talent management needs to look at referral attrition, causation information (where available), and shared connections to create "dispersion/profiles" that can then be used by the sourcing team when amassing potential candidate information. This could then potentially be something that's looked at - much like assessments or any other data point when used in making a decision.

The point of the effort (or the bulk of it, anyway) isn't necessarily to "screen out" as much as it is to understand the workforce you have, what it brings to the table, and what you are potentially adding to it. :)

Comment by Keith D. Halperin on November 5, 2013 at 7:32pm: Thanks, Crystal. I'm not sure I follow you. Are you saying dispersion/profiles to see who to go after? If that's what you're suggesting, I've not been in a sourcing situation where my colleagues and I could afford to get that "granular" without any guarantee of (potential) candidate interest. Similar to my previous comments about spending a great deal of time doing candidate research, ISTM unfeasible in most (quickly-moving) recruiting scenarios. Instead of using a rifle, it's best to use a shotgun- "Source 'em *all, let God (or their own wishes) sort 'em out."

In re: to employees not getting a "'warm fuzzy’ feeling we want employees to have with referral relationships." Personally, I don't care if the "referrer" likes or even knows the "referee" as long as they think the person is qualified and hirable. I'll give an example: back in the Diesel-Punk Era (or was it the Bronze Age?) I was recruiting for a client that had so much employee buy-in for the ERP that the employees would pay for newspaper classified ads (told you it was a long time ago) out of their own pockets to get people to refer. I'd like to hear about the 2013 equivalent of THAT again....

Cheers,

Keith "Love Don't Pay the Rent" Halperin

*Using the best parameters you can include to get a decent number of potential candidates.