Microsoft's Workplace Social Network Becomes Emotionally Aware
Mining workers’ messages for emotional trends could help managers monitor morale, but the technology struggles with humor.
Bosses who want help gauging employees’ morale can now turn to Microsoft’s workplace social network, Yammer. A new feature offers managers a kind of emotional surveillance system, showing which feelings workers are expressing in messages posted to a company’s Yammer network, which has similarities to both Facebook and Twitter.
The feature, called Crane, was developed by startup Kanjoya, which makes software that does the emotion recognition and logging, with close collaboration with Yammer. Once the feature is switched on for a company’s Yammer network, it offers managers a view of the “trending emotions” within a company, using a line graph to show the level of excitement, confusion, and other feelings over time. The topics or words most often associated with those feelings are also shown. The software is able to identify 80 distinct emotions, but it condenses those into 15 for display and shows only the most prevalent ones to reduce the complexity of the interface.
While business-focused social networks are gaining in popularity, data scientists are also developing new ways to extract useful insights from the data that flows through general-purpose social networks (see “What Facebook Knows”).
Yammer’s new feature can track emotions from a particular group within a company—for example, the sales team—but it does not assign emotional readings to individuals. The tool does, however, highlight “influential” users who seem to be swaying the sentiment that others express. It is intended to help managers stay abreast of morale and monitor the reaction to important changes such as a corporate restructuring or a product launch.
Yammer had a few of its existing customers try out the new emotion-recognition feature for three months earlier this year to provide feedback that was instrumental in fine-tuning the feature before its launch. One of those companies used it to monitor reaction to a new e-mail system. “They were able to see not only that people were unhappy but that it was causing annoyance and that people were being made to feel stupid by this new system,” says An Le, VP of business development at Yammer. This allowed the company’s IT department to address the problem more quickly than it otherwise would have, says Le. Yammer is offered to companies on a subscription basis, and the emotion-tracking feature will cost extra.
Kanjoya trained its software to recognize words and phrasing suggestive of emotions by drawing on contributions to its Experience Project, a site where Web users have anonymously shared more 19 million personal experiences and confessions. “We’ve got millions of people teaching us how emotion is expressed online,” says Armen Berjikly, founder and CEO of Kanjoya. “We used that to create Crane, an engine that tracks the range of emotions, their intensity, and their change over time. Our vision is to enhance technology with emotional awareness.” A company administrator can train Crane to be aware of company-specific jargon or phrasing that is being misclassified by the system, to make its assessments more accurate, says Berjikly.
Microsoft recently acquired Yammer for $1.2 billion and plans to integrate it with the Microsoft Office suite. Le says that the next version of the emotion-tracking system will be able to send an alert if expression of a particular emotion suddenly increases or decreases, making the system less burdensome for managers to use.
Emotion-detection software has its blind spots, however. Similar technology is already used by many companies to track feelings about their products expressed on Twitter (see “A Social Media Decoder”). But Shel Holtz, a consultant who advises companies on internal-communications technology, says such technology is known to struggle with humorous messages. “After the Toyota recalls a couple of years ago, sentiment analysis suggested that [consumer] feelings were pretty even—but they weren’t,” he says. Sarcastic tweets such as “My Toyota didn’t accelerate too far, a wall stopped it” were logged as positive.
“If you don’t look at the comments behind the emotion score, you could make some flawed assumptions,” says Holtz. All the same, he believes using the graph should enable managers to be more responsive to their employees, as long as they don’t trust it blindly.
Managers interested in attempting emotional surveillance will need to make sure employees know that their messages are being analyzed that way, and some may find the notion unsettling. This could be an additional challenge for companies interested in Yammer. Workers can already be apathetic about using a social network for work because they may find it hard to see how it could be useful to them personally in the way that Facebook and similar networks are.