>> Mary Czerwinski: Okay, everybody, I think we can get started.
It is my pleasure today to introduce Kay Connely from Indiana University, my alma mater, so it's a real pleasure for me.
I don't know a whole lot about casework. I'm really looking forward to hearing about it.
She works in the area of Pervasive Health in the Computer Science Department within the School of Informatics. And I'm involved with the Advisory Board for the School of Informatics, so somehow our lives got crossed, and we were lucky enough to invite here over for a talk.
So enjoy, take it away, Kay
>> Dr. Kay Connelly: Thank you. So what I'm going to talk about today is an emerging area that's really come about in the last five years or so.
I think if you look, like, seven years ago people started having these workshops at CAI and UB
Comp all about technology in the health care arena. And two years ago we had the first
International Pervasive Technology for Health Care Conference and as you would expect for a first time conference, you know, the success rate wasn't something we were really proud of in publishing. But we had it again this last year and the success rate went up to 37 percent, we had so many submissions.
I think it's really kind of a field that's emerging in it's own right.
I've been working in it for about five years now.
One of the things that I've noticed, when we see a lot of presentations at the conferences in this area, is a lot of people are very much technical people. They're really interested in sensor networks and putting in systems in the hospitals and things like that. They don't have a very strong design background and they, a lot of times, they will design applications without ever talking to potential users.
So the message I'm kind of bringing out today is something that we're doing at Indiana University, and that's having a very human-centered or user-centered approach to designing these technologies.
And some of you, if you're in the HCI field, this is going to be very familiar to you, but I just want to really get that out to everyone who's involved in the pervasive technologies for the health community. So that's going to be the focus.
I actually started out in pervasive computing seven or eight years ago. I was looking at a lot of different applications, and I've slowly kind of gotten more and more involved in health care because of the potential impact that it has.
I think that people, you know, they have a hard time enjoying the rest of their lives if their health is at risk or is suffering. And so I really enjoyed working with people to help them maintain and improve their own health.
So just to give you a little bit of context into why this is an emerging area. I think a lot of our current health infrastructure in the United States, as well as other first world countries, are really becoming overloaded because we have a couple of significant trends that are going on.
The first one is that our medical care is really good; and so diseases that used to be terminal are now becoming chronic diseases. People are living, 20, 30 or more years with diseases such as cancer and AIDS, and they require significant medical attention. So we're having to spend a lot of our resources managing these chronic diseases.
The other thing is a demographic change, and it has to do with the Baby Boomers basically coming up to retirement age and starting to experience all the issues that you have when you're aging. People are just living a lot longer, so they're overloading those specific resources that are targeting elders as well.
Another thing though that is happening because, I think, a good part of it is because people don't have the kind of time to spend with their doctors that they used to back in my grandmother's day is people are becoming a lot more proactive about their health care. They can't rely on being able to spend 30 minutes with a doctor and have him or her lay everything out about their disease.
They're going on-line looking for resources, asking other people and trying to proactive in managing their own health. And this is changing the roles that doctors and patients are playing.
This has really opened up, I think, a lot of opportunities in terms of giving technology to people to help them manage their own health as opposed to just putting technology in the medical facilities.
So, we have a lot of opportunities here to support these health related behaviors and that's what
I'm particularly interested in.
So, what I'm going to do is talk about kinds of approaches we take at Indiana University this human centered approach.
I'm going to illustrate some of the big themes with some projects that are going on right now.
At the very end, just a couple of slides, I'm going to kind of announce a new initiative that's going on at I.U.
So the first big thing is having a very user-centered iterative designed process. Obviously, when you're looking at a new problem domain you want to do is consult the literature. You want to consult the experts, they have a wealth of knowledge you need to use. You shouldn't just rely on the experts. You need to really verify a lot of the ingrained assumptions and actually interact with your target population.
We also use a very iterative design process in that you don't want to come up with this great design spend a lot or resources building it and say oh, deploy it and discover some of your assumptions were wrong or something wasn't working right. So you want to obtain feedback from your users very early and often and incorporate that feedback into the design.
So, I'm going to illustrate kind of this process with an application that we are looking at -- that we're developing to help a very, very sick population manage an incredibly strict diet.
So I chose this quote because it really illustrates how hard it is for this particular population to live with their disease. They -- their whole life is centered around their disease and trying to manage their diet, and they're very exhausted by it. It's something that you have to think about when you're designing for that population.
So the population is hemo-dialysis patients. I'll call them dialysis patients from now on. Basically, the idea is these patients have lost all use of their kidneys. They don't function anymore and without a transplant their lifespan is probably 3 to 5 years because of all the complications that arise when you don't have use of your kidneys.
So, essentially what happens is, once every 48 hours, they have to go into a dialysis unit and they undergo hemo-dialysis where they're hooked up to this machine you see on the right-hand side of the picture and it acts as their kidneys. The blood is slowly removed from their body it filters out excess fluids and excess nutrients and then put back into their body. And this process takes about four to six hours to do. And it is very uncomfortable. By the end of it they are cramping and they are really pretty miserable. They are dehydrated, and their entire lives revolve around this. Every two days they have to go in and do this. Often they can't have a job. They have to really think about what they're consuming, and it just really exhausts them. So you can imagine if you only have use of your kidneys once every 48-hour what kind of dietary restrictions you're going to have. Imagine you can only urinate once every 48 hours. That's essentially what you're talking about.
So you have very strict limits on the amount of water you're allowed to drink as well as a whole host of other nutrients including sodium, potassium, phosphorous. It's really complicated to try to manage this.
It turns out 80% of the patients simply can't adhere to this diet even if they are motivated to adhere to it because it's such a complicated diet. One study has shown that up to a third of the patients cannot perform the calculations to go from the nutrition label to what the doctors can tell them they can eat. They don't have the mathematical capabilities or skills to do that. And it makes it harder.
Our particular target population that we're looking at at I.U. is an urban population that has low literacy rates. Some of these patients can't even read.
So traditionally what -- the kind of care that they're given is -- they meet with dieticians, and they try to review what they're eating and help them, you know, avoid certain foods and to do that they're told to keep paper diaries. You know, write down everything that you're eating.
The compliance rate for these diaries are as low as 11 percent.
It's very difficult to remember to write down every single thing you put in your mouth. This is even for motivated people.
And our particular population, since some of them can't read, when they're told to keep a diary of what they're eating they try to draw sketches of the food they're eating. You can imagine how complicated and time intensive that is.
In another population, the breast cancer population, 94 percent of the patients were actually able to comply when they had electronic diaries.
We had nurse researchers that worked with the population come to us, they were looking for a technical solution. We brainstormed with them and we came up this idea, which we called the
Dietary Intake Monitoring Application or DIMA. The idea was we wanted to give them an electronic diary. It could do the conversions that were necessary. It could give them real-time feedback instead of waiting till they met with a dietician. And maybe it could have significant impact on their dietary habits.
We wanted the device to be portable, because you don't always eat at home. We wanted something you could have at home, but you could also have at a restaurant or if you go to a friend's house. We needed to figure out a very easy way to input what they're eating. Obviously, if they can't read, they're not going to sit there and type in they're having Cheetos or something like that.
We chose two input mechanisms. The first one was a barcode scanner because most packaged foods have barcodes. We had access to a database that linked the barcodes to the nutritional information.
And the second one, if something doesn't have a barcode, perhaps you're eating fresh fruit or eating something you've prepared at home, we'd have some sort of icon interface. We didn't know what it would look like at the time, but that was the idea.
And then we would always provide feedback. They could always look and say, Where am I?
And, How much am I allowed to drink? The amount of sodium, phosphorous or whatever it is they're consuming.
We hoped by having real-time feedback they'd be able to make choices in real-time instead of waiting until they review it with a dietician and then make plans to alter their diet.
So to do this, because we're working with a pretty vulnerable population, one that often isn't designed for low literacy population. We had a very intense, iterative design process. We started this project back in 2005 before we had funding from the NIH. It's picked up speed quite a bit since we've got funding.
But one of the very first studies we did is, can they even use PDAs. This particular population, their physical health is much older than they are in reality. They compare more to elders. They have vision problems. They have dexterity problems and coordination problems that are a result of their particular disease.
We needed to see if they could actually use a PDAs. Could they see the icons. Could they press the buttons, could they use a barcode scanner, could they do voice recording, things like that. We found they could actually use these devices, but things like for the size of the icons, we couldn't use even the largest default icon size that came in most of the programming packages. We had to customize them so it would be large enough for them to see.
We can -- the icons that we're using and the device we're using, we can fit three icons in a row, and that's as much as we can fit.
Instead of a lot of traditional PDA design let's fit as much as we can in the screen because we had these young power users, we had to take a step back and think how do we design for this particular population.
Something else that we found. We tried a couple of different barcode scanners. And one that we thought was going to be really popular it was a little pin you wipe over the barcode.
It turns out it wasn't usable at all for a couple of reasons. One being that you were trying to hold the food item, the pen and the PDA somewhere. And it was just, you know, without three hands that was difficult to do. And another reason is the version that we used back then, the scanner didn't have any kind of visual or auditory indication that you were actually successful in scanning.
That was a big drawback. Instead we used the barcode scanner you plug into the top of PDA that is made by Socket.
So, once we knew that they could actually use the PDA, something that we found out in the study was that -- we actually performed that study with healthy young people, healthy older people and this particular population, and this was the only population that said, I don't need anything with barcodes.
And they were -- we'd ask them what they ate, and sure enough they did eat things with barcodes but they didn't recognize them. It wasn't part of their world, in their perception. So we wanted to know if we actually gave them a device and had them take it home, would they be able to find and successfully scan these barcodes.
That was the second study that we did where we gave them the device for three weeks and had them scan that everything they were eating, and if they couldn't find a barcode they would voice record what they were eating so they could come back later and we could train them where to look for the barcode.
.
We had some interesting results. One of the things the nurses were really afraid of was this population would not return the devices because they thought they might sell them to get money because it's a very impoverished population.
It turned out they were more careful with the devices than any group of people I've ever worked with. They were very careful to bring them back every single time. It was a real pleasure working with them.
But, what we discovered from this was, yes, they actually could with very minimal training, they could find the barcodes. They could scan them, no problem.
I think at the two-week mark we had our maximum, how much they were scanning every day.
Then all of a sudden there was a big drop. And what we found was they decided it was a little bit easier to voice record what they were eating than scanning, so they quit scanning.
But when you listen to the audio recordings of what they were recording it wasn't something that you could easily parse. It was, you know, someone who couldn't read when they ate Lucky
Charms cereal was saying, well, I had a bowl of cereal. The box on it has a Leprechaun on it.
It's not something we can come up with a voice recognition any time soon that can actually determine what that is so they can have real-time feedback.
But they obviously really did like voice, so we decided for the next study to actually pursue was this barcode scanning the way to go or should we think about doing voice recording or voice recognition as the input mechanism because they tended to like that.
We had avoided that at first because the nurses said that they were very sensitive about their disease, and they wanted to avoid any technology that had the stigma of disease associated with it.
So voice recording, if they were around someone else, it would be very obvious what they were doing.
So we had avoided that. We thought we would pursue this for one study.
We did this voice vs scanning study and so for half the time they had a cell phone, and they called into a number, and they went through this menu, and they recorded everything that they ate. The other half of the time they had this scanning application, but they didn't have a voice option in the scanning application this time.
And we were comparing how much and what kind of things they were actually recording with both mechanisms. What we found here was very mixed results.
At this point we really could of gone another way with the design.
They actually said they preferred the voice recording, but as it turns out they were actually less successful at voice recording and the kind of detail that we needed to give them real-time feedback.
What -- the reason it turns out they really like the voice recording is because they like the form factor of the phone. They liked having a cell phone on them, where the PDA was a bit larger and bulkier.
With the scanner they were recording more things, more detail about what they were eating and far less detail and even success in getting the voice recognition software to work with the cell phone. So we decided to continue with the scanning and the PDA solution because it also allowed them to have that real-time feedback every time they turned on the PDA. Where as with the cell phone we would give them the feedback over the phone, but then they'd have to call up if they forgot what their values were.
So we decided to stick with the PDA although we might have been able to go in either direction there.
So the last study is we've been talking a lot about the scanning, well, not everything has a barcode. And this is probably the more complicated of the design process.
Why do we want to actually do if they need to find icons? How do we structure it in such a way they can find the icon? That they're willing to do it because they have very limited amount of icons that fit on the screen.
So we did two studies. The first one we were looking at high level navigation. Do they like to organize things in terms of time of day or food groups or some other way?
And we also looked at the feedback icons, you know, we gave them a number of different kinds of feedback icons that they could interpret correctly in terms of where they were in their daily
consumption.
And the most interesting thing that we found out of this study -- it did help us with the kind of high level navigation, but with the feedback icons, we found that they actually preferred icons that they did not interpret correctly.
And so we couldn't use those icons. Obviously, it wouldn't be useful if they couldn't understand where they were in their daily limits, but if we had just kind of stopped at "what is your preference" we would of picked something that was really very bad for this particular population.
So we ended up using an icon they could interpret correctly even though it was a preference.
The last study we did was more how do you do the navigation so it's not too complex and you don't get lost. We tried a variety of things. Everything from maybe you have a tab for each food group.
All the way to you have back and forward errors, so if you accidentally tap on the icon you can go to the previous page and then follow it to very linear where you had to just keep go down the step and if you got lost, press home and start at the beginning.
It turns out that last one is the way that we went with because the other two were really complicated, and they got lost and confused about where they were. This was the one that people got lost the least.
>>: Why did you have to go back to home why not just the back button?
>> Dr. Kay Connelly: Because they got very confused. Part of it is you can pick fruits, vegetables, we then went with color after that. Is it a green fruit? A red fruit? And if they -- they got confused.
Am I in the fruits? Am any the vegetables? It really was difficult for them to complete the task when they had the back button. Not all of them, but a lot of them.
Question?
>>: Could you tailor the items that they could pick for a particular demographic?
>> Dr. Kay Connelly: Yes, we did. Th
So as it stands now, for DEMA, you basically build a meal. Say you want to scan a barcode or select an icon and if you're in the icon interface you work your way down till you find your food.
So you might pick, okay, grocery store items or fast food, it's a fruit, it's a red fruit. Now here's right -- if you accidentally hit the wrong thing you have to hit home and start from the beginning.
>>: Where were the items that you put that you made available that they could select for their particular demographics.
>> Dr. Kay Connelly: What we did -- the question was: Did we tailor the items for this particular demographic?
We have soul foods for example, because we're dealing with a large African American population.
We had done several studies where we were recording everything they ate over an extended period of time. Every food that was ever mentioned is available in this icon interface, and we have an easy ability to add.
I'm about to talk about a study we're about to do. If a new food pops up, we can have that rolled in by the next time that we see them.
.
Just some emergent things that came from one or more of these studies.
The first one is when you're talking about giving technology to everyday people, it's really important that you integrate into their daily routines. As an example of this, we originally thought we were going to have one device, it was portable so they could use it at home, they could use it when they're out of the house. It turns out patients either use it for one or the other.
They don't tend to take it off the kitchen counter, go out to eat and then come back and remember to put it back on the kitchen counter. Not all, but a lot of them had issues with this.
So even though our current NIH grant doesn't provide funding for us to have two separate devices. That's kind of the future iterations. We're going to have something in the kitchen that stays on the counter and then the portable device. It's something we didn't anticipate and it came out through these kinds of studies.
As I mentioned before, the nurses were really worried about the stigma of disease, and they didn't want us giving anything that said, "I'm sick."
As it turns out this particular population is so sick, that they're entire life is defined by this disease.
Everyone they know knows they're sick. And they're completely overwhelmed by it.
And giving them a piece of technology was actually the one bright shining moment that they had associated with this disease. And they really enjoy showing it off to other people.
In fact, so much we have to be really careful about data analysis because when we're going along and looking at the barcodes they're scanning, then all of a sudden they have 30 barcodes that includes paper towels and bleach and everything else, you have to wonder, maybe they're not eating this.
So, actually, we might want to utilize the fact that in some populations and some situations technology as a status symbol is a good thing and it might promote adoption.
The domain experts did not always have it right. You know, the technology as a status symbol is one example. But what the nurses think the patients are eating is really pretty wrong, including the dieticians. What the dieticians think the patients are eating, it's not what they're recording and it's not what they're telling us.
One of the major things when you have an illness that is so restrictive, you want to cheat. You don't want to tell your caregiver because you don't want to be nagged about it. You need to build
these technologies so the patient can reflect on what their actual behavior is without necessarily telling someone else what that is.
So the nurses had wanted, in the beginning, wanted the next iterations of this application to be networked, so they could get real-time notices if someone was drinking too much or eating too much potassium, and they could intervene right then and give them a call.
I think if we did that the patient population would lie to the technology then, and they wouldn't be honest with themselves. They wouldn't be able to reflect and see that, Oh, I am cheating more often than I thought I was. They don't want the nurses to know that.
And I think that that's really important when you're talking about health care, building in this ability to cheat on whatever your regiment is.
Okay, so that is kind of the wrap-up of what I think that the user-centered iterative design is really important. We saw a lot of things that we wouldn't have gotten from the nurses if we hadn't really go in there and worked with this particular population.
Oh, and we have the prototype now.
What we're able to do is we have an NIH funded pilot study. We have six weeks, there's going to be 20 patients actually using our application and another 20 in the control group where they're getting the technology, but it has something very minor about looking at your daily activities. No feedback or anything like that. We want the make sure that giving them technology is what motivates them to change their behavior, but it's actually doing the tracking.
If this is promising, then we're going to go forward with the clinical trial.
And for me, if you guys are HCI researchers in the audience, this is really exciting, having a study of over 40 people over 6 weeks with technology is a really big deal in this domain and the clinical trial we're talking about hundreds of patients. So for me this is just wonderful for me to be able to team up with a group of researchers who this is just normal. This is how you have to do research to have these kind of large studies.
All right, so the next kind of idea that I want to get behind for, you know, things that we do when we're looking at these kind of health interventions is looking at the existing behavior theories that are out there. And Mary's going to know all of these since she's in psychology.
But, there's a lot of existing literature out there about why people behave certain ways. And how they change behaviors. And I think that it's really useful for technologists to understand what those theories are, so they can think about their interventions in terms of those theories and maybe target them more specifically for what they're trying to do. This is a really famous one in the health information area it's Bandura's Social Cognitive theory. There are many versions of this. This is our interpretation.
And some of the kind of key constructs that come out are things like self-efficacy. People have to feel confident that they can do a certain behavior in order for them to do it; otherwise, they won't even try.
You have behavior capability, so this is having skills to actually achieve whatever behavior you're
wanting. In the case of DEMA we were really affecting their behavior capability by giving them a tool that allowed them to adhere to their diet. Whereas, before they had no idea what they were doing with respect to their diet. That in turn can help their self-efficacy.
Then you have social persuasion. The social context really has an impact on what you're doing, and in fact there's kind of a recent study out there where they were looking at overweight and obesity and they were looking at overweight and obesity, and they were looking at the social networks. And if you have a lot of friends that are obese, you're more likely to be obese. So social networking -- your social networks have a very significant impact on your health behaviors.
Also, of course, reinforcement.
If you see changes or if you're congratulated for achieving a behavior, you're more likely to continue. So this is one of them.
Another one that I really like is the Transthoracically Model. It looks at the stages of change, and you go all the way from pre-contemplation where it's not even on your radar, you're not thinking about change, to, okay, now you're thinking about change and contemplation. You're thinking maybe you want to do it, preparation, you've decided you want to change your behavior. You start planning it.
Action, you're actually changing your behavior, and then maintenance and termination.
The change model is that -- you move back and forth between stages all the time. And what you want to think about with the technology intervention perhaps is maybe you want to target one or two stages, maybe the contemplation, preparation and see if you can move people to the next stage. That would be a success.
So it's not necessarily saying "anyone" who is in this population, I want to give them this intervention, and I want to see that they actually maintain this positive behavior change.
Instead what the user studies, perhaps you want to say, I want to look at people who are thinking about making this change and see if I can get them to the action stage, and that would be a success.
I think a lot of times researchers just open it up to anyone that's of a certain age, so the results are very mixed because you're going to have people who are at different stages here.
So, there are other models as well. Those are just two that I particularly enjoy. And so I'm going to kind of illustrate this with this second project, which looks at physical activity.
JFK gave this quote many year's ago, and talked about how we're under-exercised as a nation, and some of the older people in the room probably remember he had this president's physical fitness thing in the elementary schools. I remember receiving my certificate. Really trying to encourage us to be more active in our youth.
Well, the situation is a lot worse than when JFK was around. In the past 20 years the rate of obesity in adolescence has tripled.
So we're worse off than we were back then. And it is due, most likely, to reasons that you could
guess.
Inactivity.
So we're not being physically active. We're sitting on our bums a lot and driving in our cars, and poor dietary habits. We're eating a lot of junk foods and fast foods and things like that.
This particular project is focusing on teenage girls. It was actually part of the CAI Student Design project a couple years ago, and I was one of the faculty advisors for it. And the team chose the teenage girls for these three reasons: Girls tend to become more inactive in adolescence. Boys are still encouraged to do sports, whereas girls tend not to be. They want to be girly. They also turn to more unhealthy weight control methods such as anorexia and bulimia so this has significant health impact both now and in the future. But as it turns out some studies show that they are more receptive to interventions at this stage for changing their behavior. So it is really an ideal population. They have the worst possible implications, but they are receptive to changing, so they decided to look at this.
So they came up with this project called "Chick Clique". And they got the name from the actual user population. They didn't come up with it themselves. And they won the competition that year and we then proceeded to implement it and do a user study that was in Pervasive Health earlier this year. And they did a couple things that are grounded in Bandura's Model of Behavior
Change that I had in that first slide.
The first thing is they wanted to model positive behavior, so they wanted girls who maybe weren't being as physically active to see other girls that were being more physically active so they could see well, if if my friend could do it, then maybe I could do it too.
They also wanted to include a lot of social support. And so they had this clique of girls, so they would have up to four girls doing this application together.
So it was utilizing their existing social networks and trying to have this social support.
And then they also had this verbal persuasion, it was actually text messaging, so not verbal, but they had a way for girls to give each other feedback and really encourage each other.
The idea behind it is that the girls would have two devices with them. They would wear a podometer all the time, and they would have a cell phone. And the cell was running the "Chick
Clique" application.
Periodically, throughout the day the girls would enter the step count they had and it would automatically be sent to their friend's phones, and you could see how much the other girls had entered that day.
And you could look at your current group progress of here DeeDee entered a thousand steps.
Maybe early in the morning she went out and took a morning walk. And yesterday she hadn't walked much at all. She had gotten beaten by her friends, so you can see her progress over time.
And so this is how were utilizing that social network.
In addition, you have -- we built in text messaging into the application, so you could easily when after you looked at group progress, maybe you would go and send a message to the entire group or subset of the group encouraging them in some way.
So for our user study that we reported on earlier this year, we actually recorded all of the text messages that the girls sent to each other. Not just within this application. We wanted to see how the texting about physical activity might differ than their other texting as well.
So, I don't have time to go into a lot of the results, but here are just a few things that we discovered.
And the first is the girls really didn't know how to talk about physical activity in a positive way.
They didn't say anything negative, but they didn't know what to say and they felt kind of silly about it. We needed to incorporate some sort of scaffolding that wasn't there to help them formulate here's a good message you might want to send. So perhaps some templates.
All of the girls had a very positive experience, but they could see how this could become negative in some situations. Especially, if it was done over a long time and people got super competitive.
They felt that that could actually have a negative impact on self-esteem in the future.
So you need to be careful about that.
So if this is part of, like, a program you would have in your gym classes or something, you would want to think about having short, targeted interventions that were somehow supervised by the instructor, so that you could make sure that it remained a positive experience for everyone.
Another thing that we might be able to do, instead of having direct comparisons, you know,
Ritchie had six thousand steps today. I had ten thousand. I beat you, ha-ha. We could show comparisons in terms of your own personal goal.
So I am three quarters of the way to my goal, so is Ritchie. We're equal here. So we can be positive in encouraging each other to reach our goals, even if those goals aren't the same.
Reciprocity is really necessary when you're relying on social support.
We had one group of girls, there were three girls, one of them essentially didn't participate, and the other two eventually dropped out too. They didn't feel they were getting the same thing from their participant.
You either need to build in a way for this kind of a game to prompt the users who aren't participating or think about how can you keep the other people participating even if one of the other people drops out.
And finally, it was really critical to have a self-selected small group of friends.
When we talked about, would you want to do this with the entire class? No way would they want to do this with the entire class. Would you want to do it if you just had some random people assigned? No.
This is way too sensitive. So they wanted to chose other, you know, their girlfriends who they felt would be supportive. In fact, when we were looking at the text messaging, in general, from the girls that weren't related to this application, they were incredibly supportive. They were talking to each other about their boyfriends, their parents and all of these other things. They didn't necessarily know how to talk about physical activity, but they were really supportive of each other and their other life situations. So I think that was an important part of the application, to let them chose who was in the group with them.
So that's the second kind of big thing.
We talked about iterative design process. We talked about grounded and behavior theories. The last big thing that I think is really critical and most UBcomp applications, but especially in health care, is having these insinuative evaluations. If you just show people something and get their reaction, that's telling you something about their attitudes about maybe whether or not they go to the store and buy it. But it's not telling you anything about their experience and how they're going to experience it.
And I really advocate strongly for this.
Another thing is, a lot of times people are very excited about a technology, but as soon as that wow factor drops off, they stop using it. So you want to see if you can actually integrate it into the normal life process, so they would actually keep it up.
Maybe, if they don't, perhaps like DEMA, perhaps it's just too hard to actually record all the time.
We had a lot of participants telling us so far that this would have been perfect when they were first diagnosed to help them stabilize their diet, or maybe they would want to use it for two weeks to reflect and stabilize. It might not be something that's appropriate to have every day, 24/7.
So this last project that I'm going to talk about is actually looking at elders. And technologies in the homes of elders to help them maintain their independence -- how we doing on time?
And one of the things that a lot of people look at with elders is look at physical safety and physical health.
I think this quote kind of illustrates our philosophy at Indiana University where your cognitive wellbeing and social wellbeing are just as important because they have dramatic impacts on your physical health as well. We want the look at the whole picture not just the physical health.
And the project is called Ethos, it stands for Ethical Technologies in the Homes of Seniors. It's really motivated by this demographic change that I talked about in the beginning with the
Boomers retiring, where basically having a much larger percentage of our population is going to be elders by the year 2030. In fact, if you look at people 85 or over that's the fastest growing demographic in the country.
So, and if you -- I just heard a recent statistics, though I don't have a citation for it yet, I'm trying to get it.
If you reach the age of 80, you're more likely to live ten more years than you were in the previous decade -- 80 -- if you reach that age, you're pretty healthy and you're going to be around for a while.
We're about to overload the infrastructure for caring for elders.
The technology holds a great promise for helping us deal with this problem.
And a lot of people are starting to look at how technology can keep people in the home longer as well as help in assisted living facilities and things like that.
A recent number I heard was that it cost on the order of $10,000 a week if you take someone and put them in assisted living facility, as compared to if they stay in their home.
There's a huge financial incentive to start looking at this area.
One of the things we've seen, especially in the last five years, is when technologists are looking at this issue, they completely punt on the issue of privacy.
You know, I've seen quoted in papers where they try to address privacy issues. They'll say things like, Well, either the older person is going to give up all their privacy because they're moving to assisted living, or they can give some up to technology, which would you chose?
That was in a -- I'll go and look -- it was in a "Spectrum" article. So they're basically saying, you know, they don't have a choice.
I think this is really bad way to frame the issue for a number of reasons. One is many of these technologies are being designed, specifically to help loved ones decide when an elder needs to go into an assisted living facility, helping to make that choice. So they're looking for trends over long period of time.
So by definition we're asking elders to take these technologies into their homes before they're being forced to that decision. So, we have to start thinking about the ethics involved in these technologies or, you know, they're still independently living, they can refuse to have them completely.
So we shouldn't punt on the issue.
Another thing that's a little more selfish is, if you look at any technology like this that starts with one target population, it bleeds out into the rest of the population. So we need to think of these issues from the beginning.
So, as it turns out, if you look in the academic literature, there's a whole lot of ways to think about privacy, and here's just four of them. I don't even have my head around all of them. I'm not one of the privacy experts that's on this particular project. But you have everything from seclusion, which is the right to be left alone, other people can just go away and leave you alone. To autonomy, the right to do what you want.
Autonomy is a real big one with elders in general because they don't want to give up control over their own lives, to property, this is the legal reality that we live under. Whoever collects the data owns it and can use it pretty much any way that want. It's different model than you have, say, in
Europe, to spatial, things, and this is really pertinent if you're talking about things in the home.
You might be perfectly comfortable with a certain technology in your living room, but you're very uncomfortable having it in your bedroom or bathroom, different spaces might have different privacy expectations.
The problem, when you kind of look at all of these different paradigms is that neither designers who are designing these technologies or elders who are using them, the consumers of them, really understand them all that well. You know, a few might, but in general they don't because it's not something that we think about every day.
And so if you sit down a designer and elder, and say let's design something that is sensitive to your privacy needs, it's really a challenge because we don't really have a good way to communicate with each other about these, and we're not all well versed in what the various implications are.
So, what we're doing in the ETHOS project, the main goal here is to develop this tool kit, which is in year two. We just finished year one, we're starting year two. It's basically a way to enhance this communication between the designer and the elders and their loved ones. So we have a participant evaluation tool where the goal is to take them through a series of questions, finding out what do they care about, what are they are sensitive to. Who would they be willing to share this information with. But you just can't come out and say it that way.
If you, you know, paint a scary scenario and say, Do you care about privacy? Everyone's going to say, Yeah, I do. If you actually try to walk them through, well, do you mind having this in the bedroom? That might be very different than -- you might get very different answers. So we have to construct this using the language that elders would use in trying to think about how they actually think about privacy and what terms they use.
And then we'll also have this designer tool, which the designer will use as they're actually putting in different devices into the home, it can actually flag them if something violates a concern that the participant actually indicated in the previous tool. So, it can assist designers in designing something that is sensitive to that particular person's needs because everyone is different, let me tell you.
And finally, we have this library which makes it real easy to implement. Basically, it can obscure timestamps, aggravate data and things like that to make sure you're not just tracking every single possible thing. You're gathering the information that's important for your design, and so it's not going to be reused perhaps in a way that you didn't expect it to at a later point.
And so what we've done this first year is we set up a living lab. And that is a picture of it. And it's basically just a one bedroom apartment.
The idea is not to bring elders in to live there, but it's to contextualize the prototypes we want to talk to them about, so they can try to get a better sense for how these technologies might be embed in the their own homes. And it's just an intermediate step.
And, then we do a lot -- we did a lot of focus groups where we're trying to understand how elders talk about privacy without actually bringing up the term privacy ourselves.
So, you know, they came in for these focus groups, and we would ask them things about usability, and need, and where would you put this, but we didn't mention privacy till the very end
to make sure that we actually covered everything that we needed.
So we've done that and again, it's more to understand how they think about privacy, not to come up with a one size fits all privacy solution, and then that's going to build into this tool kit. And then year three, we're going to be doing this longitudinal study.
It's going to be at least six weeks. Hopefully, if we can afford more, it will be longer, but we're going to have four families. They live in a retired community in Bloomington. We're working with a retirement community. We're going to have four groups of student designers designing for these families. Two of these groups will use this tool kit and two of them will not.
We'll then try to do an evaluation on how were the privacy needs or concerns actually addressed in the designs based on using this tool kit. Was it useful or not.
So, I'll talk just briefly about a couple of the prototypes that we're looking at.
In the focus groups, we actually used things that were off the shelf too. We had this big honking medicine reminder that was really ugly.
We used that. We also developed our own prototypes, because we wanted to get away from just medicine adherence and physical safety. We wanted to look at some other aspects because there are different kinds of data and different uses of the data if you start looking at say, social applications. So we designed some of these. And Ritchie Hazel who was in the audience here, he has a heavy hand in a lot of what I'm talking about right now.
So the first one is the ambient plant pot. You see there are two pots in these pictures, one on the left side of each of them. The idea is they simply have a motion detector on them. If you walk by one of the pots in one home, the other one is going to slowly flash this kind of deep blue color, so you can see that someone is by their pot at the other end, basically.
And the idea is just to give people a feel for kind of the daily routines that people are undergoing.
So in this case we actually deployed it, and there was this sun room here and the older couple kept to the sun room and they spent a lot of time. Their daughter kept it in her home office.
One thing we found in the focus groups, almost universally, this concept was disliked. Nobody could get their head around why you would want to have this.
And, in fact, the participants who participated in the this study, they thought the same thing at the beginning of the study. They looked at this, you know, we're really uncertain about it. But by the end of the study they didn't want to give it back to us.
They had to experience it before they could actually understand how it could be useful, because it wasn't filling an immediate need that they saw. But, they would be sitting in their sun room and the plant pot would go off and they'd be like, Oh, Lisa's working at home today. That's so nice.
And they jus didn't realize how nice of a connection that would be to kind of see the daily rhythms of their loved one's household, so it was -- the attitudes really changed after experiencing that.
And I think that you see this a let in UB comp in general.
The second one, I don't have a good picture of because it's actually implement in the a mirror, you know how difficult it is to take a picture of the mirror without seeing your own reflection.
I put up one of the initial sketches, which isn't exactly what it looks like now. It's mirror motif. It's basically having some sort of reminder system embed in the an everyday object. A lot of people have mirror in their home, and they chose the mirror to do this.
So the final one doesn't look like this. Basically, when you come up to the mirror, a motion sensor can detect that you're there, a proximity sensor. Then it might ask you, Have you measured your blood sugar today?
You just kind of waive, gesture interaction, yes or no. Or it might ask you if you've taken your medications. It might ask you if your daughter could submit an invitation. It could say, Johnny has a soccer game tomorrow, do you want to go with them? Those sorts of things.
It has a little bit of monitoring, it can tell your responses to any medication adherence, but it's mostly just a reminder system.
And what we found interesting in the focus groups, we haven't actually deployed this one, people just inherently trust this system. They didn't really worry where the data was going or what they were doing with it. Privacy wasn't even on their radar when they talked about this particular system. When they talked about the ambient plant, privacy did come up a fair amount.
So this last one is the portal monitor.
And there are actually a few different uses of this. It's basically having cameras around the door.
And the first one that you see here is any time someone rings the doorbell three quick snapshots are taken, and it's sent via cell phone to whoever you designate. So it could be your daughter, your son.
And the idea here is a lot of elders fall prey to people who are trying to scam them. And your son or daughter could just see, Oh, I recognize this person, that's fine, you know. It's my dad's friend,
Joe. He came to visit.
Or you could see some construction-like looking guy who's trying to sell them a new roof. You can give them a call or call a neighbor up and help protect against that kind of predatory behavior.
Another way that this can be used -- we actually have a camera pointing from the outside and from the inside, and every time the door opened, the snapshots are taken and sent to the cell phone.
And this can help prevent wandering, if your parent is starting to loose it a little bit cognitively.
And then all of a sudden you see your mom is walking out in the middle of winter without her coat.
Maybe that would give you cause or concern to call someone that's nearby.
And people, the people in the focus groups really, really liked this particular idea because it focused on the physical security.
And it was a need that they could immediately relate to. They had all these horror stories of people being taken advantage of or people going out and wandering.
And so they could really relate to this one, and because of that they were really comfortable with the fact that there were pictures or even video being taken at the door. They were fine with that.
But things like the ambient plant, not all of them, but several of them were very concerned. Well, that's kind of like Big Brother, somebody knowing that I'm walking in my kitchen now. Even though it's motion, it's a very low signal, they were uncomfortable with that. Whereas, a complete picture was perfectly fine for them.
I think that just kind of illustrates how you have to contextualize the kind of data and what it's being used for, for people to really be able to react appropriately. You can't just say video. You can't just say, pictures. You have to understand what the use is for.
Okay.
So, just some initial results, because we're just now transcribing the focus groups, we haven't done the data analysis. But unlike the "Chick Clique" application where reciprocity was really important, here it actually made elders uncomfortable.
They didn't mind, necessarily, having the plant pot in their house. They might not understand why it was used, but they're like, okay, it can be there. But they felt like it was an intrusion of their daughter's privacy to ask her to put it in her house so they could see.
We had thought that reciprocity would be really important to encourage them to adopt, so they could see their kids, and they were not just the ones being monitored that that would be more acceptable, but that was not necessarily the case. They felt like they were intruding.
Data as property, even though it's the legal system under which they live, it's kind of a foreign concept. They really didn't talk about it in terms of property. They talked about privacy and other terms.
So that's just interesting in and of itself because it's, you know, the legal reality is very different than what they think about. And that the data granularity is not the deciding factor, how much information is being given is not the deciding factor here, it's how it's being used.
And the utility that's given in response. You know, physical security, falling down, these are concerns that they really have, and they're willing to give up a lot for that. Actually, they might not necessarily be willing to give up a lot, but their friends would. That's something else you find a lot with elders, they do not visualize themselves as being older and needing these kinds of technologies, but they all have friends who could use something like this.
And then finally, just longitudinal, insinuative studies give you very different results than if you just bring them in and ask them a bunch of questions.
I just wanted to put up, real quick, the other faculty that were on this just to show you the kinds of collaborations you need to have to be able to do a project like this. You know, we have Jean
Camp who is our expert on privacy. Lisa Huber, she's a specialist with elders. She has been working with them for a long time. And we have a social scientist who's looking at our methods as well.
So those are the kinds of the human-centered approaches that we're taking at I.U.
We have this iterative human-centered design process. We ground it in human behavior theories as much as we can. We do a lot of longitudinal testing, which is very expensive, but well worth it.
And the ETHOS in particular, we're really looking at the methods at how to elicit the privacy concerns because you really can't.
So the thing I want to announce is that is really exciting to me is we have this really large security group at Indiana University, and they're all listed here. Everyone from a lawyer here, to Jean, the privacy specialist, cryptographer, trusted computing, etc..
We've recently all gotten on board to look at pervasive health applications. We've gotten funding from the Lilly Endowment to have this new institute. In fact, Dennis was integral in getting this funding and then he left us. We're going to focus our security and privacy work on this particular application domain. I think that's really exciting, because you don't see a lot of people out there doing that.
And in, hopefully the spring but perhaps next fall, we're going to host a workshop, kind of like a grand research challenges workshop, where we're going to bring leading researchers around the country, both on the technology side,on the privacy and security side and the health care side.
We're going to bring them all together, try to identify what are the major challenges that we need to be looking at over the next ten years if we're hoping these technologies are really going to be pervasive in helping us with our health care. I hope that will help define kind of some of the research agenda in the future.
Look for that next year. If you're interested in this, definitely get in contact with me. We'll be having the solicitation for basically a short proposal on why you would want to attend the conference and what you would have to add to it, because we want to issue this report that will have a big impact.
So, that's my talk.
[applause]
Any questions? Questions? Uh-hum.
>>: [inaudible] -- sensors that make a big impact where people actually fall --
>> Dr. Kay Connelly: We have not done that. I think it's a very interesting area. The focus in the
ETHOS project was really to try to come up with a variety of types of applications that address different, social, physical, different needs, just to facilitate the conversation.
The end really was not in the prototypes themselves, but it was in what is it that makes this data gathering acceptable to you where something else might not. So we ended up talking a lot about fall detection. That was a need that a lot of people saw. And so even though we didn't have a prototype for it, there's a lot of conversation in there about fall detection.
In fact, the plant pot people kept trying to turn it into something that would detect if you had fallen.
Well, if I didn't walk into the kitchen then my daughter would see, and then she would call me and
discover that I had fallen, and I won't be left there for 24 hours. Even though it's not designed for that, and it wouldn't be very good for that, they turned it into that a lot.
>>: [Inaudible] had it for many, many years. I also find your -- on the cell phone concept quite powerful.
>> Dr. Kay Connelly: What?
>>: Sending photos taken at the door very powerful. Are you familiar with the Sensescan?
>> Dr. Kay Connelly: Yes.
>>: We have -- that's also the main idea that you're taking photos --
>> Dr. Kay Connelly: Right.
>>: At interesting times, and sending the data out --
>> Dr. Kay Connelly: The reason this is actually acceptable to the people we showed it to was that it was such a targeted location that they felt it was okay. They really did not want picture and video in the rest of the house. They didn't want to feel like they had to do their hair every time they walked through the living room. That kind of thing.
But at the door they could see why it would be useful, and they didn't go to the door unless they were going to see someone else any way. So that was okay. So the Sensecam, it's a fascinating project. I don't know how comfortable elders would be with that just because they want to know and they want to have it targeted at a location where they don't feel like they're having their picture taken when they're not expecting it.
>>: [Inaudible] --problems.
>> Dr. Kay Connelly: Oh, yeah, I can understand that.
>>: One last observation. Given the experiment of putting cameras in the door. I'm sure you're aware that security companies like ADT -- [inaudible] -- health care monitoring services.
>> Dr. Kay Connelly: I think that's going to be really poplar in the future. Even though a person might not want their daughters seeing something, almost universally they had no problem with an
ADT having that information. If there was video monitoring and it was an ADT person who was watching it to see if they had fallen, that was okay compared to maybe the video showing up on their daughter's desktop in a little frame or something. They didn't like that. Or their friends seeing it. But an ADT -- that's a service you're paying for, and they saw the value of that and they liked that.
But, again, I have to say, in general, people could see someone else using it. They have a very hard time thinking that they're going to need that and buy it. You have those buttons if you fall you press it and it sends an alert, even people who had fallen in our focus groups didn't think they needed it yet. It was always someone else.
So you have that barrier that people don't see themselves as aging. So actually getting them to use the technology or purchase the technology might be difficult.
Yeah?
>>: So talking about the same photo taking. Do You were talking about the same -- [inaudible] -- photo taking again, do you think that the children who are receiving the pictures even scaffolding -- either interpreting or sifting through these pictures.
-- all the data they're potentially getting --
>> Dr. Kay Connelly: Every time someone rings the doorbell, if it's that version, or if someone opens the door.
So, in some ways, if you have someone who's very active and always going in and out, that door opening one isn't going to be appropriate for them, all right, because they're going to have so many pictures that they're not going to know what to do. But the doorbell one, most people don't have a lot of people ringing their doorbells, so that would make more sense.
And it tends to be, I think people who would use the technology more are people who are becoming less capable of going out a lot more, you know. They might be starting to early stage
Alzheimer's or something like that. And so it really becomes, I mean it's the same thing as any kind of intrusion detection system. The false positives can make it so it's not very useful. You have to gauge that.
I think it's going to be on an individual situation. I don't know what kind of scaffolding we can do.
This is actually Jean Camp's project more than mine.
It's an interesting question. Is there a way to support it beyond just showing them pictures? Can we do any kind of processing to help them gauge what's actually going on would be interesting.
Thanks.
Any more questions, Dennis?
>> Dennis: So do you every think about how you're -- the elders for example have not really grown up the technology -- [inaudible] -- people now that kids that are do texting instantly how they were -- will any of your conclusions likely change?
>> Dr. Kay Connelly: Absolutely. It's something that we struggle with a lot. The current cohort of elders, you're right, they're not as technology savvy as the ones that are about to retire.
One of the things we're doing is we're actually working with a retirement community that are early adaptors. They are technology savvy and they're reflective of the majority of elders that are coming up.
But that doesn't necessarily reflect the entire population, and we recognize that. It's sort of the low hanging fruit. The easy population to target right now.
The next iteration, what we're looking for funding for right now, is to start looking at the other side of the digital divide. We're actually going to start looking at urban centers and people who have different community settings even.
Maybe their kids don't move away, so they're not as far away from the children as the ones we're looking at now. And they have different technology experiences as well.
But you're absolutely right, different cohorts are going to have different life experiences, and that is going to impact how they think about the technology and how they use the technology. And it's something we just are trying to account for, but until you, until you get old enough to use it we're not going to know, you know.
Any other questions?
All right.
Thank you.
[applause]