Jaime Teevan, 40
Microsoft principal researcher
I am a principal researcher at Microsoft, in a division that’s known as Microsoft Research. I’ve been there for ten years.
Microsoft Research worldwide probably has about 1,000 Ph.D.’s, and then probably another similar number of people doing development and other things. There’s several hundred people in Redmond proper, and then we have labs in Beijing and Bangalore and Cambridge and other places.
It’s a unique, weird place. I’m a trained computer scientist, but I work with ethnographers and sociologists. My manager is a cognitive psychologist. She got her degree studying babies’ reactions to the world, how fast they sucked on their pacifier, and so on. There are people from all sorts of spaces.
The researchers are given a lot of flexibility. There’s no real structure. It’s just a bunch of people that the company hires with an interest in ensuring the long-term viability of the company, but we don’t have anything we’re supposed to do. It doesn’t really exist anywhere else. The only place you could compare it to would be like the old Bell Labs or Xerox PARC, labs focused on basic research that tackles hard problems that the company might need solutions for in the future, and might not even know they need right now.
Some people I work with are building hardware for computers. Some people are building algorithms to do things faster, or better. I work at the intersection of building algorithms to help people find information or interact with information, and human-computer interaction (HCI). A lot of people who do HCI work are social psychologists or cognitive psychologists, and study how people understand things. So I study how people interact with information, how they understand the information, and then try to build algorithms and tools that help them do that better.
A lot of the research that we do comes out of what your personal interests are. I did a whole couple years of research on social-network question-asking. You know, how do you ask questions of your friends, as a complement to search? Because I was out with our nanny one day, and we were trying to remember some Keanu Reeves movie. Instead of searching for it, she texted all her friends, and she got the answer. I just thought that was really interesting. So I spent a couple of years studying how people ask questions of their friends through technology.
As a large arch, I am interested in how people interact with information in general. I’m particularly interested in the fact that people are very consistent, and we have patterns of behavior. You know, you go to work the same way every day, or you do the same thing every time you start up your computer. How do we take advantage of those patterns of behavior?
In my work, what we call “fast search” is really just “normal search” as far as most people are concerned. Search engines work a lot of compromises to get people search results as quickly as possible — because a delay of even 100 milliseconds leads people to think the search results are worse. In contrast, “slow search” is where search engines intentionally use additional time to provide people with a higher quality search experience — both by helping people engage with the search results and learn as they go, and by using more complex algorithms to automatically identify and synthesize relevant content.
I was thinking a lot about how to help support personalized search, and then that broadened out to think about how people do these larger tasks in their information environments. Slow search is really thinking about tasks of work for search. How do you help people synthesize the information they’re finding, and find useful, shareable search artifacts?
You’re searching for the Facebook homepage — that’s a really fast search. If you’re searching for planning travel, or you are writing an article and you need to be doing background research for that, or you’re making a purchase, or you’re researching a medical diagnosis, those are very involved, and they involve a lot of learning. A lot of the steps involved in those searches are important, not just the endpoint.
Search engines are highly optimized to get you information as fast as they possibly can. One thing we want to do is to understand how important it is to get search results really fast, because that will tell us how we want to optimize the algorithms.
So when you issue a query on Bing, or Google, normally what we do, the first second we get your query, as fast as we can, we turn around and send you back results. But we might choose to hold onto your query for an extra 100 milliseconds before we send you results. A 100-millisecond delay is not something that people actually notice, but if we hold onto your search results for 100 or 200 milliseconds and send it back to you, you’ll actually think those search results are lower quality and you interact with them less.
Because of that, search engines have a choice: We can invest a little more time in getting you better results, or we can give them to you 100 milliseconds faster and you’ll think they’re better. So it becomes a trade-off. We make all these trade-offs to get you the results as fast as possible, and sometimes that affects the quality of the results, because you’re going to think they’re better.
One of the things that we see a lot is that about a third or even more of the searches that you run, you’re going to come back to. You’re going to revisit. If you’re searching today on a topic, and tomorrow you’re going to pick it up again — we have 24 hours when we could be doing smart things to gather good content for you. So we’ve been thinking about: What should we do during those 24 hours?
This interest I have in tasks has expanded beyond search to think about, how do we support complex tasks? I’m particularly interested in these complex tasks that are really hard to do.
One of the things I spend a lot of time doing is writing academic papers for publication. And sitting down and starting to write a paper is hard. How can we make that easier?
So one of the things we’re thinking about is how our time is so fragmented. A couple of people in my lab found that it takes a person 25 minutes to reach full productivity when you’re working — but we’re actually interrupted every 11 minutes. So people are never working at full productivity, because you have all this ramp-up time. A lot of research has been done to think about how to help prevent interruptions. But it’s also interesting to ask, Can we time their interruptions to be at a better time?
For example, I’ll block off an hour on my calendar, and be like, This is my work time. I’ll turn off Facebook or take email vacations. But rather than trying to change our environment to create long blocks, what about trying to fragment the work that we’re doing so that it fits in our fragmented work style?
So what would it look like to write a paper in 30-second bursts from my phone, so that I could get it done in these small little bursts? I’ve been working a lot with people who do attention management, thinking about, What kind of a context do you need to get a small task done? What is it like to be able to take a large, complex task and break it into these small, little pieces?
One of the fun things that’s really different about my job is my colleagues, a lot of them are external to my organization. It seems be really different from most of my friends who are in traditional roles. I go to conferences and I have long-standing 10- or 20-year relationships with researcher-collaborators at other universities or other places all over the world. Like, I know their spouses and I know their children. I wrote tenure letters for them.
It’s very academic. One of my main outlets is the papers I publish. I track my h-index. I often say my job is very much like what a professor does, without the teaching and the grant-writing. Certainly Microsoft has much more interest in us connecting with products and having our research have real outlets. But the people who work here find it interesting and exciting to have the ability to have real-world impact.
So it’s a very idealistic-type role. I’m trying to change the world [laughs] — for the better. Initially, with search, and I was starting really from the late ’90s, I was thinking about how to make people smarter with access to information. Like, the same way calculators changed our ability to do math because we have access to computation. How can having access to information change the way we can think?
You can see that happening now in the world. Our mobile phones are little augmented brains. There is so much that we don’t have to know because it’s right there.
I’m interested in the future of work, the future of productivity, and thinking about tasks, and task-completion. Once we start breaking down tasks into these small, little micro-tasks, and surfacing them in a way, we can start automating a bunch of it, getting human intelligence in a really thoughtful way, allowing people to focus on the creative and interesting aspects of their job more efficiently.
I could probably pretty easily find an extra hour in my day at work, just in these little micro-moments of time when I’m not being productive. If I could do an efficient four-hour day, where I make use of the time, then I could spend the rest of the time cooking dinner with my kids and going to the park, and that would be awesome. I want a productivity revolution. I want to change the way every single person works for the better, so we are happier and have more leisure time.