eavesdropping, that their information isn’t very confidential. So it could be that what we’re getting is that everyone is lying, under every condition. And it’s just not clear if it’s uniformly bad or uniformly good reporting in our studies. But it’s uniform; it wasn’t affected by the interface.

OK, thank you very much.

INTERFACE OF SURVEY METHODS WITH GEOGRAPHIC INFORMATION SYSTEMS

Sarah Nusser

CORK: The next emerging technology that we’re going to discuss is the interface and uses of global positioning satellites (GPS) and geographic information systems (GIS) and their potential uses in survey research. To talk about those, we have Sarah Nusser from the Department of Statistics at Iowa State.

NUSSER: I’m going to talk today about digital geospatial data and survey data collection. We’ve actually been using spatial data in the survey process for a long time; it’s just been in primarily paper forms. And yesterday we talked a lot about the process of developing software; I’m talking more about the planning, navigation, and data collection side of the survey cycle. And I just wanted to give you a little context for why I’m working on this. I work on the National Resource Survey, sponsored by USDA, that uses a lot of geospatial information that has historically been in analog or paper form. And they do both photo interpretation and field studies, and over the last five years, we’ve been using—and, thank you, Marty [Meyer] and Jay [Levinsohn]—Apple Newtons to develop the data collection software.49 And there’s a client-server setup to send out information to these people, have them collect it and send it back in.

But we had not been working with spatial information. Our first foray was back in 1999, when we sent people out into the field with GPS units. And we couldn’t get precise positioning signals back then, so we had these great big military plugger machines. And you can see the interface on the GPS machine is 4 lines with about 20-some characters. And we found people made gross mistakes in how they set up the equipment and then how they used it to capture spatial information. So what we did was connect the two together and develop an interface that allows them to see what’s going on in the guts of this machine in a way

49  

Apple Computer’s Newton device was the first entrant into the handheld computer market, and is a major focus of Jay Levinsohn and Martin Meyer’s presentation, which follows Nusser’s presentation.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 198
eavesdropping, that their information isn’t very confidential. So it could be that what we’re getting is that everyone is lying, under every condition. And it’s just not clear if it’s uniformly bad or uniformly good reporting in our studies. But it’s uniform; it wasn’t affected by the interface. OK, thank you very much. INTERFACE OF SURVEY METHODS WITH GEOGRAPHIC INFORMATION SYSTEMS Sarah Nusser CORK: The next emerging technology that we’re going to discuss is the interface and uses of global positioning satellites (GPS) and geographic information systems (GIS) and their potential uses in survey research. To talk about those, we have Sarah Nusser from the Department of Statistics at Iowa State. NUSSER: I’m going to talk today about digital geospatial data and survey data collection. We’ve actually been using spatial data in the survey process for a long time; it’s just been in primarily paper forms. And yesterday we talked a lot about the process of developing software; I’m talking more about the planning, navigation, and data collection side of the survey cycle. And I just wanted to give you a little context for why I’m working on this. I work on the National Resource Survey, sponsored by USDA, that uses a lot of geospatial information that has historically been in analog or paper form. And they do both photo interpretation and field studies, and over the last five years, we’ve been using—and, thank you, Marty [Meyer] and Jay [Levinsohn]—Apple Newtons to develop the data collection software.49 And there’s a client-server setup to send out information to these people, have them collect it and send it back in. But we had not been working with spatial information. Our first foray was back in 1999, when we sent people out into the field with GPS units. And we couldn’t get precise positioning signals back then, so we had these great big military plugger machines. And you can see the interface on the GPS machine is 4 lines with about 20-some characters. And we found people made gross mistakes in how they set up the equipment and then how they used it to capture spatial information. So what we did was connect the two together and develop an interface that allows them to see what’s going on in the guts of this machine in a way 49   Apple Computer’s Newton device was the first entrant into the handheld computer market, and is a major focus of Jay Levinsohn and Martin Meyer’s presentation, which follows Nusser’s presentation.

OCR for page 198
that was more appropriate for the task that they were pursuing. And we really did this seat-of-the-pants; now that I’ve been working in this area a little bit longer, our developer did a pretty good job of thinking about how people work with this information. But I want to take this from a more systematic perspective. So I’m going to start out talking about: what are geospatial data? They come in a lot of different formats, and they have a lot of issues about them that really affect how we set up our survey data collection systems. And I’ll probably only have time to talk a little about the cognitive aspects; there’s a whole other set of problems that have to do with computing infrastructure, because these data are very voluminous and require some adaptability in how you work with them. If you want to know something more about this research, go to this Web site. This is also funded by NSF, the Digital Government project. And my collaborators are both at Iowa State and [the University of California at Santa Barbara (UCSB)]. So, we’re used to dealing with this sort of thing; we know a lot about how to phrase questions, we have these nice clean coded texts and— even if we’re doing a scientific study—we like to develop protocols to construct precise and definable measurements. When we move into geospatial data, it’s a different ball of wax. There are two general types of spatial information: one is called vector and the other called raster. Vector data are just points; lines are connections between points, and polygons are just a bunch of segments put together to form a polygon. The basic form is what you get out of your GPS unit, which is a single point, or a sequence of points to make a line. This is a road map that is a set of lines or polygons; if you tap on these lines it’ll give you back the street name. So, even though you have this spatial information, there’s sometimes attribute information linked behind it. Raster data we can basically think of as two-dimensional array, where the cells are basically pixels and—in this case—the value is a color that’s provided on the screen. So all the information is visual; there’s no extraction of numeric information or identification of features, so it’s up to the human to figure out what’s going on here. Something like this, which is a soils map, you might be able to—it’s a raster map—you might be able to tap on it, and there is some attribute information connected to it, what kind of soil is there and what are the properties associated with the soil. More often than not, in our world, we’re going to want to combine a couple of different sources. This is a topo map with a sample unit boundary. And we might even have dynamic data, even in the form of video, which I’m not going to talk today. Or it might be in the form of—I don’t know if you guys in the back can see this, but—getting GPS readings as you’re driving along or walking along a route.

OCR for page 198
So let me talk a little bit about how we might use this type of information in three different phases of the survey process. This is just sort of a backdrop on what the issues are. When we’re doing “planning” here, I’m thinking about a field representative who has been given a set of sample sites which they have to go to, so they’re working with multiple destinations. They need to know what order they should visit these sites in and what should the route be? So the basic spatial information here is just street maps—annotated vectors, essentially—and the determination of routes along the vectors in that street map. And, mainly, we look at things at a large scale; we might want to zoom in on some smaller areas as part of the preparation process, but generally we’re in the larger scale when we’re doing this kind of thing. When we’re actually navigating to a site, we’re generally focused on one destination at a time, a sequence of single destinations; we might start off with an overview of where we need to go, but eventually we need to focus in to find out what the specific instructions are, in terms of how to get to a place. Interface design is going to depend somewhat on your mode; if you’re in a car, you don’t have hands or eyes particularly available for looking at a map. If you’re on foot you have more availability. How quickly you go determines how fast that map might be going by, and that interacts with screen size. So there [are] some issues here we need to deal with. Also, because we’re mobile, we’re starting to come into this dynamic component where we might have a GPS moving along in a map, and we get to the first data collection we want to do, which might be capturing a route to get back out of a convoluted neighborhood or for the next field rep to be using. Finally, in data collection—this is really tough, and there’s a lot more to be done on this side—it’s usually a very focused activity at a specific location. And we go beyond our GPS and maps to using other kinds of information. If I’m doing a Census Bureau-type study, I might be having to go out and do some listing. I might want to be looking at plat maps, the map address file. I might be interested in topographic maps if I’m in a new neighborhood and really trying to figure out, “Where am I in this neighborhood?” And we’re using this material as reference material, to find out what is going on around us in our surroundings or as the base for collecting data. So, if I’m collecting a GPS location I might want to see it on the backdrop of a photograph to make sure that I’m actually collecting the right thing. We might also want to make hand annotations on an image; I was actually just talking to Marty last night about doing just that sort of thing. So what are the issues for survey data collection? Well, the first thing is that spatial data—and raster data in particular—can be quite large in size, which raises computing issues that come into play in terms of

OCR for page 198
transferring data out and where you want to do your computing, whether you want to do it on the server side or out in the client. The other big thing is that we are very focused on error in the survey world, and all locations—all geospatial locations—have some error, and it may be acceptable and it may not. And part of it comes from the fact that we’re trying to take the Earth’s surface—an irregular sphere—and put it down onto a screen or a piece of paper. And that involves some distortion, and there are many different ways in which we can do that, so it’s also possible to get geospatial data sets that have two different formats, put them together, and it’s not a good idea. It’s very easy to make mistakes due to the complexities involved in spatial data formats. There are also issues of resolution and the positional accuracy of nodes in your vector, and so on. But, once again, it’s always going to be present; it’s just a matter of whether it’s going to be too much or whether you can live with that amount of error. This gets back to some of the things Roger was talking about. Historically, the primary mode of interaction with geospatial data was visual. We have very little feature extraction associated with … particularly image data, to help us as humans to extract information from some of the more complex formats. I really think that’s down the road; it’s not something that we can’t do, it’s just something we haven’t focused on. In addition, we have a whole new cognitive process we need to think about, and we need to find out whether there’s a framework we can work from in presenting this information to the interviewers and perhaps someday to the respondents. So there’s a literature on spatial cognition theory, and one of the very first questions we asked was, “does this literature apply in our world?” I’m going to talk a little bit about this, and one thing to keep in mind is that there’s a lot of variability in how people perceive spatial information, so we want to bring that back into creating settings and interfaces and tools that allow us to get back to our principle of repeatability in data collection. So the basic goals for using digital geospatial data in field data collection [are] to develop designs, tools, and systems that will minimize measurement error, in part by accommodating variations in the users, and to be able to provide appropriate information for a given field environment. I probably won’t talk very much about this, but … on the computing side, we’re trying to develop systems that will allow us to have different kinds of computers and different types of abilities out in the field. And sort of make the infrastructure much more flexible than we’re used to doing. I’m going to talk a little bit about spatial cognition theory and interface design. And then talk about a user study I did with Jean Fox at [the Bureau of Labor Statistics (BLS)] that helped us get the first step under

OCR for page 198
our belt in this area. I’d like to talk about developing spatial knowledge, and I think about this as though you’re moving to a new area, a new city. There are basically three levels of knowledge that are conceptualized in this theory. The first is that you have “landmark knowledge.” You have isolated points on the ground that you are aware of, but maybe you haven’t connected them. So it’s the very first stage, and you pretty quickly get to the stage where you can connect lines between points. And this kind of knowledge is called “route-based” or “procedural knowledge;” you tend to be moving along in sort of a linear sequence in your mind, you tend to be orienting yourself on the ground, looking forward with left-right type reference frames. As you become more familiar with the area, you develop much more detailed knowledge. Your network begins to fill up so that you have a pretty detailed network from which to work. Your reference frame tends to switch to a “birds-eye” view where you tend to start viewing the system from above, with cardinal directions. And people who work in this framework usually work well with cognitive maps. How far along you get along in this development, and how quickly, is mediated somewhat by your spatial ability. The dimensions of spatial ability that are important for this problem are the ability to rotate information, so in your mind be able to rotate a map and see where you’re going even if the map is north-up; to orient yourself or find yourself on a map; and then confidence in or anxiety over performing spatially-oriented tasks. And what appears to be known is that as we get older our brains don’t work as well as they used to, our ability to rotate information is not as good, but we know more. We have more experience, and so are able to orient ourselves on maps more readily. An interesting aspect, it turns out, is that our anxiety over performing spatial tasks varies very widely across people and sometimes within people. So if you’ve been lost, and you’ve gotten sort of befuddled, and you have trouble placing yourself on the map, this is kind of your mental state decreasing your spatial ability, and that’s important when you get lost out in the field and are trying to find your way to your sample site. So, you can think about different strategies that people use in dealing with spatial information as being a continuum, from route-based thinkers on one side to configurational or map-based thinkers on the other. Route-based thinkers tend to think of things in linear order, an egocentric reference frame. They tend to prefer written directions, lists of instructions with landmarks noted on them. They tend to want to stay on those major thoroughfares because they haven’t developed that network. Whereas people who use the map-based view tend to think from the bird’s eye perspective; they’re very comfortable with maps, often they have just a map in their head and don’t really need a paper map

OCR for page 198
or a digital map, and they’re willing to get off the road to get to where they want to go. OK, so let me shift gears a little bit and talk about how we work with maps and the navigation process. What you’re doing is aligning three different reference frames: one is yourself, another is the reality around you, and the third is the representation of that reality, the map. If these things are not aligned, you have more mental rotation to go through. Most of the maps we’re used to looking at are north-up maps; the nice thing about these is that, for example, if you’re a pilot and you have to communicate with air traffic control you have a very clear communication mode with this kind of a system. But you have to do much more mental rotation, and that raises the cognitive load involved in using this sort of presentation. Research has been done to compare with “track-up,” or “head-up,” or “forward-up” type of orientation where the map is always turned in the direction that you are going. This eliminates the mental rotation step. That’s good for our setting in the sense that, usually, interviewers are working by themselves and not needing to be in communication with someone else. When you add driving in, you’ve got two really demanding tasks, cognitively. So when we think about using spatial information in this setting, it’s very important to try to minimize the mental effort needed to interact with the spatial materials. When we think about map interfaces, we need to think about our task; I’m talking about navigation right now, but it might be a different task in the survey process. How someone prefers to get and think about their spatial information; which way is it easiest for them; and some issues related to presentation. You only want to have the stuff you really need on that map; it needs to be very visible, so if there are any cues to assist you in finding yourself on the map the interface design will be better. So, that’s a quick primer. Any questions at this point, at all? So what I’d like to do now is talk about a small user study that Jean Fox and I did. We basically used field staff that are associated with the Commodity and Services Survey; that’s an establishment survey where they’re going out to collect prices on goods and services that people have given them through another survey. They get new businesses that they need to go to through the refreshing of the sample or by being assigned to new areas, and that happens quite frequently. We were focusing not on the Wal-Marts and the McDonalds, but on the things that were harder to find—the service industries, gardening services, upholstery, medical services, and so on. And also looking at areas that they were not familiar with. And the questions that we were interested in answering dealt with these modes of dealing with spatial information, and how they impact survey work. And, then, is there evidence that if you give somebody a digital map versus a paper map that there’s any bang-for-the-buck?

OCR for page 198
Third, if we then add a GPS to that environment, will that help them in any way or is that just gilding the lily. And we got pretty lucky; we worked with four staff, and they laid out very nicely along this continuum between very route-based and very configurational. There was one person who hadn’t been very long in the city she was working in and had fairly low spatial ability; really preferred to work without a map at all but with directions listed down. At the other extreme was someone who had lived in the city for a very long time; she was willing to use maps but she had it all in her head, so you could tell her something and it would log in her head and she would be able to go. She didn’t really need to look at a map much at all because she had it all in her head. What we did was—thank you, Marty, again—used one of these. We also looked at using an IPaq which is more this size, but there were software issues there so that we didn’t get much out of that.50 We had a very cheap software, Streets and Trips, made by Microsoft—lots and lots of features, really great for this experiment. After training them, we gave them 6 to 8 outlets and they had to plan their route using the software. They had to navigate to half of them by only using the digital map and to the other half adding in the GPS. The way the software works is that you put your businesses—you can use Excel for this—or other addresses into the Route Planner; you can ask it to optimize the order, given the beginning and end of the route; and produce directions in both written and visual form. The staff did that; they generally accepted the order and they always accepted the route. So you’re left with dealing with a map and route listing interface; you can adjust the size of these, you can adjust the font size, you can keep or hide features like hotels and restaurants. There are a lot of things they could do to alter the interface. What did we find out from this? The best thing we found from this was that they mapped very nicely into the spatial cognition framework. The behaviors that they exhibited followed closely what would be expected from the literature. People who are route-based tend to make that part of the interface much bigger than the actual map; vice versa for those who are accustomed to looking at maps. One woman who had pretty good spatial ability but was new to the area would use her map as a set of instructions; she would zoom in and kind of scroll through, so she was soaking up the step-by-step instructions but in a visual form. The preferences for interfaces, and this is a fairly loose part of the study—they basically didn’t do certain things. They did increase the font to make it larger, so that they could see more readily on the screen. We told them about the landmarks they could add to the screen but they never wanted to do that. They were most interested in cross streets 50   The IPaq is Compaq’s handheld computing device.

OCR for page 198
and the street number of the business they were headed to, and they had all sorts of suggestions to try to make that more visible. They really preferred the simpler interface and did not try to clutter it up with anything that seemed like marginal info. GPS was really quite effective at reducing the cognitive load associated with finding yourself on the map. So, what people could do is that they could pick up the tablet and see whether their icon—a little green icon—was on the line or not. Just a very quick assessment, so they could take a quick look and keep driving. If they got lost or on a different route, they could pull over and immediately see that, well, I’m here, and this is where I’m supposed to be and they could choose a new route. The software could actually help them with that, but we did not teach them that feature. The people who used this feature were really much more comfortable in doing this assignment because they got this quick and easy feedback; the route-based person, of course, didn’t use this because she didn’t look at the map much. If you’re not using a map, it doesn’t do you much good to do GPS. So our conclusions from this study is that it’s probably possible to identify a couple of different approaches to developing interface design, or to make the interface so that it’s suitable to a spatial strategy. It’s possible for the setting of personalized training, to take someone who is route-based and try to move them towards being more map-based. We thought that this would be very desirable to get people up to the level of survey knowledge [before getting into] the navigation phase. There are a bunch of unanswered questions. The first ones on our docket—we’re just in the middle of planning a second study—include: what would happen if we left these tablets with them for more than a couple of days, which is basically all that we worked with them on this time. Would they change how they use the map resources? Would they be more flexible with them? Would having a GPS help those route-based staff use maps more readily? Would it provide those additional cues that would get them over the hump in using maps? And then there’s a whole number of things that GPS software can do—not the one we were using— in terms of recording routes, capturing coordinates, and using voice cues in navigation. So there’s a whole other set of interface questions looming on the horizon, if we use another piece of software. What we haven’t dealt with are a couple of very big issues. One is raster data, these more complex images. Spatial cognition theory is really centered on way-finding and maps; what do we do with things like this? Is there a way in which we can augment it so that laypeople can be using this kind of interface, for example, if we were going to be having them doing decennial census work? The other thing that we really haven’t broached at all is: what principles should we be thinking about in collecting spatial data? We’ve got a little bit of information on

OCR for page 198
that on the point level from our other studies with USDA, but there are a lot of questions there. The second big area—and an area that’s of interest to me, and that Marty’s going to spend some time talking about—is what can we do with small interfaces. So, we’re very interested in using something this size [holding up a Palm Pilot/IPaq-sized device] because the tablet—particularly for environmental surveys—is just a little bit too big for the things that people need to do. There’s also—part of the project is looking at other emerging technologies. This is a set of glasses where the screen is being projected into a little inset in the glasses, and it would actually overlay information on your view. So you might have a see-through view. And there are other questions related to other modalities coming down the pike. So, this is really the “What are we going to be doing in five to ten years?” type of research. So let me just give you a little bit of a heads-up of what we’re looking at in this project. This is actually a different kind of interface where someone has a screen clipped to their glasses, so instead of being an overlay like this it’s really more like a separate screen off in your peripheral vision. People at UCSB are looking at how you might design that interface and how you would interact with that interface given that it’s up here instead of in front of you in a tablet. They’re also hoping to explore—although this is probably next generation stuff— how to use a see-through screen; in other words, if you overlay screen information on the environment, can you label things like houses and say, “this is the one that you want to get to”? And how much dissonance is there when you have this in your vision field and you’re trying to do other things? Where am I on time? CORK: You have 10–12 minutes. NUSSER: OK. I can pause here for questions. Yes? ROBINSON: You said your route-based person didn’t use maps … NUSSER: Right … ROBINSON: So what did that person do when she got lost? NUSSER: Well, she had to use the maps to recover. But what she really did was—the map software would through this particular route. She would look at the map to begin with and pick the interstate—to go 15 miles out of her way to do something that was very comfortable. So she was less likely to get lost but she was spending a lot more time getting where she needed to go. So that was an overstatement on my part. GROVES: A similar sort of question: How did you sort them into these headings? How did you measure their spatial ability at day one?

OCR for page 198
NUSSER: We were not permitted to do testing on these employees, so we did this through observation—looked at how they were using the maps, setting up their screens, and so on. So that was a post hoc classification of these individuals. GROVES: And how did you know how they performed on the task? NUSSER: We were with them in the car. And actually one of the questions we have for next time is how can we get that assessment a priori. We have some—I think people are pretty good at rating their spatial ability and their preferences, and so we need to come up with a set of questions that will allow us to do that ahead of time. CORK: Any other questions for Sarah? NUSSER: Do you want me to talk about computing stuff at all? CORK: You have time, so go for it. There’s one more question … PARTICIPANT: You said that many people, I believe, were interested in the address, and had suggested changes, and I wonder if you could explain a little. NUSSER: Yes. Their job is to navigate to a housing unit or a business, and the key identifier there—once you get to the right street—is the street number. So they’re driving, and they know they’re on the right street; then they need to know that they’re on the right block, and then what the sequence of the street numbers is. And they wanted that street number to be about this big on the screen because we were in larger cities, and the street numbers could be four or five digits long, and you can’t remember that. PARTICIPANT: [inaudible, but centered on differences in spatial ability between right-handed and left-handed people] NUSSER: I’m trying to think … yeah, that’s right, and I think we had all right-handers. PARTICIPANT: Because they think very differently … NUSSER: I think we did have all right handers … PARTICIPANT: [inaudible] NUSSER: I’m married to one of those. [laughter] But I think we had all right-handers, all female right-handers. There’s also some literature that says that females think differently from males spatially; there’s also in the literature a convergence in how they think, so I think that my personal theory is just a little bit like the Web: as time goes on, and socialization becomes different, then people will tend to be thinking in the same way. MARKOSIAN: I didn’t quite understand how you could differentiate preferences for north-up versus head-up maps … NUSSER: We could not, because this software only does north-up. You would need a compass to do that, and we didn’t do that. The GPS we had can’t do that.

OCR for page 198
MARKOSIAN: The newer GPS handhelds can provide that. NUSSER: Right. Jay? LEVINSOHN: Do you have a sense of the trainability on these, on how possible it is to train people to modify their ability to use the more advanced capabilities? … How likely are these devices to move people forward? NUSSER: Yes, I—I don’t think we have a … The question was, do we have a feel for whether it would be possible to move these route-based people forward—well, not forward—into configurational points-of-view. I don’t have a very good feel for that from the experiment. I do have a theory that if we add cues to that environment, that they can hang on to—“you are here”—and use to orient themselves—“you want to go here, and the path is this,” that would be much better than just getting a map. Their prior experience is just to get a paper map that has no annotation that’s relevant to them. So that’s one thing we’d like to test in the next study. LEVINSOHN: Give them personal icons … NUSSER: That’s a very good idea, actually. Do we have time, still? CORK: If you want to take a few minutes, sure. NUSSER: I’ll just go over this quickly; I threw this in because we had this comp sci community, and in the survey world we tend to have a very rigid structure in the way we set up our data collection systems. So I’ll just throw these ideas out here. We tend to work in a system where we’ve got people out in the field, and we’ve got a repository that we’ve prepared, and a client-server interaction where we’ve prescribed everything that’s going on between these actors. And part of the research project is looking at a different kind of model, now that there’s so much available on the Web—and that’s particularly true of spatial information—where a field user could not only get to the stuff that they need from a methodological point of view but other information that might help them when they’re in a situation that’s not covered by the repository prepared for them. And we’d like to do that in a way where the user can be completely naive; it’s mediated by the infrastructure. And so the goal is to seamlessly deliver spatial or other kinds of information in formats appropriate to the field environment. “Appropriate” is a very loaded word; here, we’re thinking about variation in user characteristics, variation in what kind of field computing environment we have—whether we have any number of the devices you’re going to see in the next talk, and variation over time within a survey or a task on how these characteristics change. So, just to give you a better feel for that—again, we’re thinking more broadly than just demographic surveys in this research project. People come in with different levels of spatial knowledge into the project—you

OCR for page 198
might have a social scientist who works with maps all the time, who will have a higher level of spatial ability than somebody like an interviewer who uses maps for navigation but their primary duty is to interview somebody. And then, working with the decennial 2010 group that’s thinking about mobile computing devices, they’re basically assuming that their workforce is citizens—we make no assumptions about baseline ability in terms of spatial strategies. We have the spatial strategy component that I talked about earlier, and then physical aspects having to do with how quickly you’re travelling, whether you’re travelling by car or foot, what you’re doing—do you have your hands and eyes free, and then disabilities that really are no different from these physical settings, just an additional cause for having limitations. In the field computing environment, we’re sort of thinking about the traditional things: what is the screen size, what is its resolution, is it color, do you have a little device that doesn’t have much storage or processing ability, and how much does storage decline over time as you collect information, communications aspects and interface modalities. So the notion behind this project is to borrow from the notion of a “data wrapper,” which basically provides an interface between the computing infrastructure and the data. So you have a request coming into a front-end server that is shunted to the database; well, this wrapper will translate from—if you’ll allow me the colloquialisms—the language of the infrastructure to get a query into the database that the database understands, and then generate data back that the infrastructure can understand. So, in this project, we’re developing the notion of a “field wrapper” which is basically the same sort of thing. You have different conditions associated with the device, your communications, and the user. That’s metadata about the field environment; that’s shunted into the infrastructure through an interface, if you will, that translates into the language of the infrastructure. So the idea would be to send information on this field environment forward to what’s called a mediator that generates a sequence of actions that are appropriate for this interface. So in this case there may not be too much processing capability, not too much storage space, so the sequence of actions would be: get me a big chunk of data from here, put it in the computation space, and get ready to send out little chunks, several little chunks at a time, as you move around in the space. The upper setting, that’s basically a PC. So you might ask it to go ahead and get a chunk of data here, and it might be wired a different way via a land-line or something like that. And you can just go ahead and transfer the entire file to that setting. So the idea is to get us away from this notion that we always have to buy 500 of the same box, and move into more of an incremental change strategy as was talked about yesterday, of

OCR for page 198
having some like this and then—as new products come on—be able to add in different kinds of devices out in the field. So I’m probably by now way over my time, but let me just say: there are a lot of problems here. And why are we doing this? Well, almost everything is tied to a point on the ground. This is becoming an increasingly common mode of communicating: we have our MapStats up for the federal statistical community. If you take the “geo” out of what we’ve been talking about today, you basically get the flowcharts that were up yesterday. So we’re getting very used to interacting with visual information. And geospatial data are really rich in information content—much more so than a list of addresses—and I think that there’s a lot to take advantage of there. And the quality of the data is increasing, so we can use it in a way that will feed into our measurement error structures. So, that’s about it. CORK: We can take one or two questions for Sarah while we switch computers up here. PARTICIPANT: How large are the files? You said that they are big … NUSSER: They range from just a few kilobytes—which is what a vector file would be—into megabytes. Generally you’re not dealing in gigabytes, but it’s not hard to get there fast with a few images. So this presentation which had a number of clips in it is probably 10 megabytes. When we think about the data for the National Resource Survey which we do—which has about 300,000 sampling units—we’re definitely thinking terabytes for that data. PARTICIPANT: So that is something you really need to think about. NUSSER: You really do, and I guess in part that’s why we’re not thinking so much about data structures but rather how can you work with—for example—storing this in different places and being able to pull it into a computation server to break it up and send it into the field, rather than having things set up in a prescribed way before you go out in the field. PARTICIPANT: [inaudible] So would you need some kind of dial-up connection? NUSSER: It can be slow, and so if you’re doing dial-up, this is one of the settings we’re working with in the project, where you have pretty poor communication. So you want to take just a subset of the data and send it out, something that’s small, that you can send out via a wireless or dial-up line, or whatever. And we have a long ways to go. Yes, Mike? COHEN: Handheld devices are being trumpeted for use in data collection for the 2010 census. Do you see the need for different types of devices inside the blue line versus outside the blue line? NUSSER: Sorry, could you give me the jargon?