How to Use Facebook to Survey Engagement

Talent Management is focusing on HR technology in the current issue. Here is what “Psychology at Work” had to say about technology and HR in a provocative post earlier this year.

Engagement survey, meet the 21st century. With the explosive growth of social media and amazing new computational programs, massive data sets have become available to researchers in the social sciences. Soon we will have the capacity to measure in real-time the mood and emotions of any given set of people through their digital communication, other than cave-dwellers (and even cave dwellers, if their cave has Wi-Fi) through Facebook, Twitter, and other social media.

Now, what do we do with all of this?

At the University of Pennsylvania, home of cutting-edge research into computational assessments of mood and emotions, an effort is under way to track well-being in various countries by mining massive Facebook and Twitter data sets. In the ambitious World Well-Being Project, a team of researchers under the watchful eye of famed psychologist Martin Seligman, chief among them Andy Schwartz, Johannes Eichstaedt and Lyle Ungar, are using psychological text analysis and combining it with ever-increasing computational sophistication to measure well-being at any point in time across vast geographies.

I asked Eichstaedt, a young research scientist on the project and a friend, to explain it to our readers: “Using these programs, we can infer psychological states at any given time by having the computer count the occurrence of certain words as a fraction of the overall text volume. For example, when people use the word “joyful” in a status update on Facebook, researchers infer that either that user or a friend of that user has experienced that emotional state … we have gotten much better (through more sophisticated natural language processing) at assessing the mood of a given population from their Facebook posts.” In other words, serial Facebook poster, computers know what you are thinking – and feeling. All the time.

OK, I know what you are saying. This is getting a little bit scary. Where is George Orwell when you need him?  Although I might argue that we as a society crossed this Rubicon a long time ago – why do you think Facebook is worth billions of dollars? As an easier way to share baby pictures? – our purpose today is not to debate moral philosophy. The truth is, companies have the legal right to monitor your digital traffic – email, Twitter, Facebook, LinkedIn, etc. – while you are at work, and many have explicit policies to that effect. The question is: should companies ditch the annual, standardized questionnaire and embrace these new computational programs to find out whether their employees are really engaged – in real time?

I think so. It would not be very difficult to develop a program to measure engagement using words that reflect its presence, or absence. As a wild guess, I would think if words like “hate,” “bored,” and “suck” show up in employee emails or tweets a lot,  things aren’t great. Conversely, frequent appearance of “fired up,” “juiced,” and “love” would indicate something very different.

Let Johannes explain how their research works. “We have a few thousand Facebook users take a standard psychological survey. After they consent to our use of their data for research purposes, we ask a computer program to find those lexical features (words, but also punctuation, etc.) which best predict their survey results. In this way we can automatically generate targeted dictionaries for different psychological constructs measured in questionnaires or demographic properties (such as gender, age), which make no theoretical assumptions whatsoever and which are completely determined by this new kind of data.”

What are they finding? “We find that extroverts mention the word party a lot and  agreeable people don’t say no.” Really. “Men use the word gaming a lot and women say shopping. Placing many of these words together allow us to create dictionaries which with high statistical precision collect those words which are most predictive of people being happy, or in good relationships, and so forth.” And by extension, although Johannes won’t say so, predictive of employee engagement at work.

I think it is a matter of time before corporate America embraces this technology. It is becoming too affordable to measure the psychological well-being of  large numbers of people – with scientific precision – for us to ignore it. (The New York Times recently said we might as well get used to the idea of big data in more aspects of our lives). Sure, there are some issues to work through – do you “friend” some computer in HR? – and we have to guard against abuse, but the potential this has to improve engagement surveys is to0 great to ignore.

Now, time to go post this on Facebook.