One of the major use cases for Big Data is input from sensors. For the most part, this has meant sensor data on windmills, power lines and other inanimate objects.
But now companies are putting sensors on people and using Big Data tools to figure out what our daily movements mean.
The Wall Street Journal recently shared Bank of America’s experiment with 90 workers wearing sensor-enabled badges around the office for a few weeks. Not only did it record their movements, but it also captured the “tone” of their conversations. I don’t know what that means, exactly, but it makes my hair stand on end.
This little project demonstrated that “most productive workers belonged to close-knit teams and spoke frequently with their colleagues.”
As a result, the bank started scheduling employees for group breaks, rather than solo ones.
Yep. There go your plans to run to the pharmacy and pick up your asthma inhaler.
But it worked. Bank of America’s HR executive claims at least a 10 percent increase in productivity.
But during lunch time, that social activity dropped while employees went to desks to check emails, rather than talking with one another(!). So they redid their lunchroom, complete with better food and lighting so employees would opt to lunch together.
Then they cut back on superfluous water and coffee stations, so everybody had to use the same station.
Finally, the Journal adds, it set a 3 p.m. coffee break, “both to prop up sagging energy levels and to boost social interactions.”
All of the examples seem pretty positive for employees — better lunch room, more time for chatting — but the article also looks at some of the problems it could create — such as employers wanting individual data.
But from where I stand, that’s just a drop in the barrel of what’s bad about this use of sensors. Since you’re not here to read my personal views about how being an employee is not the same thing as being an indentured slave, I’ll stick with the data-related problems: Both these studies raise serious questions about how this kind of research is done and how the data is interpreted.
Data interpretation is always tricky business, but with large data sets, as many experts have pointed out, it can be particularly challenging.
My first question: How can they be sure these sensors were worn during productive “private time.” An inanimate object — which is how sensors have traditionally been used — will seldom take its sensor off and leave it on the car passenger’s seat. If you’re part of a study, of course you wear your lanyard in public – but I personally use to remove my work ID automatically when I sat at my desk. How can they know the sensor data was always on, or even in the same room as the participant, for that matter? Was that data pulled?
What about this issue of volunteers? Could this create a self-selection bias?
For instance, maybe only very productive employees volunteered. Or worse, maybe only extroverts volunteered, causing the productivity results to skew in the favor of extrovert-friendly work activities. Introverts can be just as productive (I would argue more), but they tend to value privacy, so it’s probable that they would not volunteer for this survey.
So was this a study that included all different types of workers? Or is this just a survey of more and less productive extroverts, who may or may not be effective at their job?
You could also cite differences among other groups — women versus men, young versus old.
Then let’s look at those numbers: 90 for one study and 30 for another. What percentage of the total work group does that represent?
Then there’s a question of long-term consequences. For instance, both companies took steps to promote or even force social interactions at times employees normally reserve for more solitary or even personal concerns.
Are more employees taking an afternoon off or sneaking out early because they didn’t run errands during lunch?
Are important emails with valuable clients going unanswered because employees are stuck at the coffee station engaging in a group discussion about ideas that would be just awesome … if everybody could just find the time?
Now, it may be that the companies who did these particular sensor studies may have addressed these issues and created a rigorous data sample, with statistically legitimate results.
If you’re Ben Waber, then it’s probably fine. The story briefly mentions him; he’s the chief executive of one such company, Sociometric Solutions, and he’s no statistical slouch: He did his doctoral research in media, arts and sciences at MIT, received his masters at Harvard, where he was a research fellow at the Institute for Quantitative Social Sciences for a year.
Waber doesn’t seem to be connected with the two specific studies mentioned. And it’s worth noting that in his company’s work, about 90 percent of workers participated in the sensor studies.
But this type of story raises some real red flags, not just for individuals concerned with privacy, but with companies concerned about not wasting money. As Big Data reaches wider adoption, we could see serious misuse of sensors and the resulting data in the workplace.
Then there’s this little gem: I’ll call it the Data Point of One.
Very little stands in the way of using data on individual workers, according to one source interviewed in the article.
"Not many service providers are going to refuse to give information to an employer that's paying the bill," Lewis Maltby, the president of the National Workrights Institute, told WSJ. “It would be very surprising if some provider doesn't start giving employers data about individual employees when they ask for it. That's not illegal. But do you really want your employers following around what you are doing? It's a creepy way to work.”
No kidding. The question is: Is it also a bad idea?