joshiThe interview with Dr. Anirudha Joshi

December, 2015

Interviewer: Prajish Prasad

Interviewer: Welcome to this IEEE TCLT interview at the 7th IEEE International Conference on Technology for Education 2015. I am Prajish Prasad, currently a PhD student, in the Indian Institute of Technology, Bombay. We are pleased to have Dr. Anirudha Joshi here with us today. Prof. Anirudha Joshi is professor in the interaction design stream in the Industrial Design Centre, IIT Bombay, India. He works in the field of Human-Computer Interaction. He teaches HCI and related topics in IIT Bombay and in academic institutions all over India. Prof. Anirudha is involved in designing interactive products for users in developing economies.

Thank you Professor for having accepted our invitation and during this time with you, we would like to hear your thoughts on the topic “Developing ICT for people with less education”.

Question: An Interaction Design for Indian Needs Lab has been set up at IDC. What got you interested in the idea for setting up such a Lab?

Dr. Anirudha Joshi: Actually, I have been working in this space for quite a few years now, because we believe that there are several problems that developing countries generally face, India specifically faces, which are potentially solvable with the help of new and emerging technologies. So we just recently called it a new Lab, but we have been working in this space for about 15-16 years now. So the basic idea of this lab is to bring together people, interns, researchers, PhD students, MDes students and let them work on interaction design related issues for Indian needs, particularly focuses with less education. So, if you look at the current set of information and communication technologies, these have been designed with the objective of supporting the needs of people in the developed countries, for urban, westernised, typically English-speaking, office-going, that kind of (people). But, now technologies are becoming cheaper and more accessible and more widespread, so there is an opportunity to design for people with less education. And because technology is reaching these people, except that these technologies have not been designed, these products have not been designed with these people in mind. So that is what we try to do in this lab.

Question: Can you share with us some of the key projects which this lab is undertaking or has been undertaking in the past?

Dr. Anirudha Joshi: So, basically, if we look at the population in India, we can broadly divide it into 4 categories, 4 groups. They are approximately equal (in) size. So, there are those who have reasonably good amount of education, and by that I mean 10 years of schooling or more, so that’s about 20-25% population in that space in India today. Then there’s the next group of people which has less than 10 years of schooling, maybe between 5-10 years of schooling, and this group we actually call as Non-English literate. So they are literate, they can read and write, but they are not very comfortable with English, their exposure to English is not sufficient, so they need services or products in the area of localization or specific apps, so they need a certain kind of products. But textual literacy is not a problem in their case. Okay, so that’s one group that we are working for. So we have done products for truck drivers etc. We do a lot of work in the area of text input. So we have this keyboard called Swarachakra, which we launched a couple of years ago on the Android platform. We have several other products also in text input, but this is the one that we made available to a large audience, and we have now over 600,000 downloads. So that’s a product that we targeted for this group of people. If we kind of go beyond that to people with primary or less than primary school of education, so these people usually they are still literate but they are badly literate, I mean they often have lost their literacy or they don’t have practice with their literacy. So this group of people actually needs additional hand-holding as far as literacy is concerned, maybe minimal use of text, although I will not say zero use of text, because that is often counter-productive. People think that, oh, this is not meant for me, I can read and write, although I struggle. So actually, we should have some text, but the interface should not rely too much on text. And in fact, by providing text in such interfaces, we are actually providing literacy practice to these people. So that is also a benefit. But other than that, actually, they tend to be involved in different professions. So they need different kinds of applications. So that’s a second group of people. The third group of people is completely illiterate. So, some of the/our work in that area is to support, help them learn literacy, so help them acquire literacy using ICTs. That’s one group of work. Other group of work there is to actually… so for example we we’re involved in a healthcare application where we help HIV patients manage their regimen better with the help of IVR calls. So IVR, for example, does not require any literacy. But IVR in itself is not very usable even by literate or educated people. It has a separate set of problems, so we kind of designed and optimized IVR system so that they can be usable by people with no education, no prior experience to interactive devices. So these are some of the projects that we are working on.

Question: So as you mentioned about this IVR system to support people living with HIV especially in resource-limited settings. So can you explain its workings and some of the challenges which you faced in designing such a system?

Dr. Anirudha Joshi: So there are some very interesting challenges. So one big advantage of IVR systems is that it is a directed-dialogue system which means… so there are many ways of looking at it, but the way we look at it as far as illiterate users are concerned, the dialogue tells you what to do, so it uses phrases like ‘if you want to do x, then press 1, if you want to do y, then press 2.’ So it tells you specifically.

If I kind of convert this into a visual system, I will put a button for x and a button for y. So now I have to make a decision that the button looks like something that I can press on, so I have to, you know, apply… so it doesn’t direct me to do that, I have to interpret that, so it’s sort of an interpretative system. So, we actually decided to take advantage of the fact that IVR does not require literacy, and we started using this and we started creating, we had to create several bunch of prototypes. Currently the product that we designed actually does four of five things.

Firstly, it gives people reminders for their pill times and particularly in case of HIV, but in case of lot of medications being very adherent to the medication is very important. So, it gives a call at pill time and it says by now you should have taken a pill and, so if you have taken a pill press 1, else press 2. So if you press 2, then it gives certain kind of feedback. If you press 1, then you get a feedback about your adherence, which is ‘keep it up, your adherence is 96%, so that is good and so on. So it gives you that… so it gives you like a run rate in cricket, you know, so it gives you a target to achieve, it keeps you on track. So, that’s one thing that it does. We can also, if they’re not feeling well, if they’re facing side effects or if they are facing any symptoms, they can look up the problem in the IVR and then very often they can save a day’s trip to the clinic or they get reasonable advice or sometimes people tend to ignore minor symptoms. Some of these symptoms can be very life-threatening also in cases… so it can off escalate that. It also does a few other things. It does… it provides, OK, when somebody is detected HIV positive or is put on ART treatment, he’s given a huge amount of information through counselling and the counselors do their best to provide the information in the way that people can absorb but ultimately, at the end of the day, it’s like drinking from a waterfall. So, you are just given this whole amount of information on your head so usually you do not absorb it all or retain it and so on. So we break down that similar information into chunks of 30 seconds and everyday provide a 30 second chunk. And then after a while it repeats so it can fully re-emphasize it. It’s also ordered in a meaningful manner so that in the beginning it gives you certain kind of tips and little later it gives certain kind of tips. So, the other things it does is it provides appointment reminders and reminds you on the day of…, so it first reminds you to take an appointment, later reminds you to go for the appointment. So, it does things like that. So, what are the challenges? That was the other question that you asked. Actually, there are a lot of interesting challenges, but I just want to highlight one or two in this case. So, a lot of interesting challenges came in the area of scripting. For example, we did some of our early evaluations in Marathi and we found (and I am kind of loosely translating) – “If you have taken a pill, press 1; if you are going to take it later, press 2. The language was fairly straightforward and we thought that this was as simple as it can get, but we found that this actually failed with some of the users. Then we went and we changed the dialogue and we made it a little more expansive. So we said “If you have not yet taken your pill, but do plan to take it after some time, then press 2” and that worked so it’s not just a question of, you know…, so ‘later’ sounds like… or in marathi the word “nanter” sounds like “this will be OK”, but I mean most people would understand it, but probably something didn’t work that the way it was getting communicated, and we just made expansive, more clearer, more actionable, so going to take it after some time, so then it kind of worked. So there were many such subtler details which actually just clarified the communication.

Question: Another project which you mentioned and which has been a huge success, as you said, is the Swarachakra, so how did the idea of creating this app come about?

Dr. Anirudha Joshi: This is something that has been sort of close to our heart for a very long time, and text input is a work that one tends to do in a lifelong sort of a way. So we’ve been actually working in the area of text input for a while. We started originally with physical keyboards then when phones started getting smarter, we started designing for slightly smarter phones, basically phones that were Symbian OS or earlier version of smartphones or what we can probably call today is future phones. Then at some stage smartphones and touchscreens particularly started becoming available and we saw lot of potential in that, for particularly Indian languages and we sort of designed subsystems for Indian languages. So, Swarachakra was one of the five different concepts that one of my students had designed at that time, and we evaluated all the concepts and we found this one to be doing the best in the field and, therefore we decided to prototype it a little more rigorously and then we tested it with some of the other competing products and that worked pretty well. Is it a huge success? So relatively speaking in the sense that I think smartphones in general have changed, have done a lot of change for the way people input text which did not happen, say for example on the desktop computers. So today a lot more people and a lot more text input happens on mobile phones, on smartphones in India, than it ever happened on desktop computers. So, yeah, in a sense it’s a huge success, but still, yes, able to reach only a very small percentage of the population. So Swarachakra itself has been downloaded about 600 thousand times by people in 12 languages, so it’s kind of small community. So we are supposed to have about a 20 crore i.e 200 million smartphones in India in this year, growing really fast. But I think in relative terms, yeah, there is some success, but I think there is a fairly long way to go.

Question: Can you mention some certain key design decisions which you made while planning to create for the development of this app?

Dr. Anirudha Joshi: OK, so there are several interesting design decisions. So, first of all, this has been our basic philosophy so at one time Indian text inputs in Indian languages was described as a puzzle. People could just not figure out how to do it. So, at that time, we said that, you know, we have to use the basic conceptual organisation of the Indian scripts and that should be reflected in the design and then that is the only way that this puzzle will be solved. But this was like about, say 15 years ago and we’ve kind of moved a little ahead from there. So, now we have solved that puzzle, at least on smartphones, the puzzle seems to be reasonably solved not only by us but also by many other keyboards that are out there on smartphones, so seems to have solved them reasonably well. However… so, before I go into the however part of it… There are some other decisions that we decided too. So because you are designing for the touch screen, we decided to take advantages of touch screen in a significant way so there are two main things I would like to point out – one is The Swarachakra. So, it throws a wheel around the consonant and so Indian languages are what are called alphasyllabaries, the abhigudha scripts, so which means there are several consonants which are immediately followed by a word modifier for that consonant so we wanted to make that process as quick and as easy as possible, and not necessarily a reflection of Unicode, but a reflection of the script. When you put your finger down the chakra Swarachakra appears and then you drag your finger outwards. So you are taking advantage of a touchscreen. The other thing we take an advantage of a touch screen is the formation of the conjuncts. So, conjuncts are particularly difficult and confusing for lots of people. What Swarachakra does is provide a preview of how the conjuncts will appear and that seems to help users, particularly beginners enormously. So, these two things actually take advantage of the touch screen and are close to the script. I wanted to mention one other thing which is what is now the current challenge? So, the original challenge was that text input is confusing or was a puzzle for Indian languages. The current challenge perhaps is that of speed and accuracy. So the confusion part we have now sort of managed to leave behind, and I think it’s an important thing too. The next challenge is to actually reach and even now text input though it’s possible is fairly slow we want to see how that can be ramped up, particularly since a lot of text inputs is now happening. So now, the focus on speed is actually relevant. So, yeah that’s what we are working on now. Improving speed.

Question That leads to my next question so what are the future directions which Swarachakra can take or plans to take? Is one of them maybe increasing the speed?

Dr. Anirudha Joshi: Yes, so speed is a function of two things, one is the performance of the device, which probably includes the technology capabilities of the device like how smooth the touchscreen is and so on, and how powerful the processor is, how good the display is, and things like that. But it’s also a function of the user so how does the user ramp up, how do you help the user, how do we save the user’s efforts, how do you reduce his cognitive load, the flash on the screen you know so those kind of things so it’s kind of also challenging in some other interesting ways. For example, one of the challenges in Indian languages is we have large number of characters. Well not large as Chinese or Japanese but still much larger than say the roman scripts. So one solution that many keyboards use is a shifted layout to solve the problem. So, this poses lot of cognitive and performance-related confusions and poses lot of cognitive problems. So if you take that problem away and you unshift everything, the keyboard becomes big and this increases the finger travel and also occupies a lot of space on the screen. So you have to balance between these parameters and then try to increase the speed. We are working on certain ideas right now. We have done fairly detailed studies to understand what is it that is actually slowing people down after they have reached a certain level of proficiency. So, as people try to increase their speeds, they start doing certain kind of errors which slows them down. So we are trying to reduce those errors and as errors are reduced, speed automatically improved to some extents. So we are doing that. We are removing, or adding and removing certain features which were actually slowing people down. So, we hope we are getting some breakthroughs and not getting all of them but hope we are on the verge of releasing a new version which will hopefully solve some of the problems. We are hopeful about that. We don’t have concrete result yet, but I think we will be there.

Question: In all these projects, when you have developed applications for people with less education, what are some of the important lessons you have learned, maybe in these explorations in general?

Dr. Anirudha Joshi: So, one important lesson that we always learn is that actually there is a lot more diversity in the user groups than what we are used to in the educated user groups. Education tends to sort of even out or make people lot more similar. But when people are not very educated, they have not done lot of school education or college education, at some stage there is a lot more variety which is a great thing because it’s actually not a bad thing at all it’s the richness of the world that we like. But, we have to actually design for those diversities so we particularly look at these three groups – the non-english literate, the badly literate and the illiterate and they have specific needs, but within these groups, there are further needs and abilities that actually we need too. So, actually, there is a lot of work that needs to be done in terms of understanding these users, modeling them, theoretical theories behind how different people use and so on. So, diversity is the one of the lessons what we have learned. Second thing that actually we have learned, particularly in the text input space is that we really find that user behavior or user performance tends to be very surprising or on a very unexpected line. A simple example is we have evaluated a whole bunch of keyboards which use prediction styling interfaces, prediction completion etc., so smarter techniques and others which don’t use them. And we are very surprised to find that users tends to perform slower on keyboards that use prediction techniques or completion techniques, than those that don’t. So, by just looking at the keyboard, I mean we have evaluated even keyboards which are identical in all other respects except for the predictions and so it happens, without prediction, keyboards do much better. This is probably because the cognitive load the prediction interface poses on the user is actually taking away all the advantage of efforts saving. This is for us a counter-intuitive and a surprising finding. I mean we can explain it after having done the study through some mechanism or through some theory but we really didn’t expect it to happen. So, similarly any design change that we make, tends to have an unpredictable effect on one’s users, particularly in text input. So, text input and some of the other performance oriented activities are actually heavily dependent on actually performing empirical evaluation. Otherwise, we cannot… Just by looking at the keyboard, it’s very difficult to tell whether it will… Well, this is, in a sense, known in the field of HCI for several years, but it still comes as a surprise when it goes in the opposite direction of what it expected to do.

Question: Do you see an increase in ICT usage in primary education and if so, according to you, what should be some of the design considerations for ICTs in such settings?

Dr. Anirudha Joshi: We have done some experiments in ICTs for primary education, school level education and we believe that actually, there’s a lot of opportunity out there. In fact, in India, we have some of the highest teacher to student ratios. Individual attention from the teacher to the student is actually effectively that much less, because we have just so few teachers. So, we have to come up with tools and devices which will engage the students in self-learning, in discovery, and also will aid the teachers in paying more attention to the students who need, while the other kids are still engaged in some way. So, there is a lot of opportunity for doing that. We have not really done any publicly available products, but most of our experiments have shown that there’s a lot of opportunity in self-learning and evaluation. I think we had the presentation today by the MHRD representative, and I think he talked primarily about higher education, but actually in primary education… In India, we have a large amount of school dropout, we have a large amount of people who are not able to go to schools, or schools are far away or too few and so on… So, in another way, our society is getting what is called “informationalized” which means before people get basic needs like roads or electricity or even a toilet, before that, they actually get an information or a communication device. It’s kind of a lopsided view of development perhaps in some way, but here are opportunities, so we can make use of these opportunities to provide better education. So I believe that there is a lot that we can still do and I also believe that we are actually on this cusp of take-off where many of these things are going to take-off. In 1995, everybody had this feeling about the internet, so those who were in the field of internet, we had this feeling that things might suddenly explode, and they did in the next five years, so 2015 feels like the 1995 as far as developing countries are concerned. Maybe in the next five years suddenly things might take off. What shape they will take is difficult to predict but it looks like all the underpinnings that you need for large scale digital adoption of technologies in all spheres of life, particularly in developing countries are all now in place. We have the networks, we have the devices, so many of the things are coming in place. The question is how do we put these together, how do we use them, I think that’s where the crux lies.

Question: Can you in brief talk about some of your future projects or ideas that you are really excited about?

Dr. Anirudha Joshi: We want to actually do several things. One of the things is of course we want to expand on everything we are doing right now. So, text input, for example, we are very keen to solve the speed and accuracy problem next. We want to expand some of the work we have been doing earlier in the medical domain to include other kind of illnesses, or other kind of conditions, say for example, there’s a huge demand or huge need for doing something in terms of maternal health, so we are kind of hoping to get into some projects in that space. I think there are several other, so India is supposed to be a diabetes capital. So, there’s a lot that can be done in that space. So, in the healthcare area, we want to do some things. Then, we are actually working on, we are now expanding our language work into something called the Love Your Language projects. So we urge people, and anyone from India watching this, you are welcome to come and work with us if you love your language and want to do something for it. One of the things, for example, that we are doing right now is creating tools or support systems to help people learn in other Indian languages. We are a large country with so many languages, but relatively speaking we have very few resources within our country to learn other languages. I am a Marathi speaking person, I know a little bit of Hindi, I also have a little bit of smattering of Gujarati, because I stay in Mumbai. Maybe because I’ve seen a lot of Bengalis, I understand a little bit of Bengali. But if I have to learn Telegu or if I have to learn Tamil, then I just don’t have any resources available. And what is interesting is that these languages are surprisingly close to each other. But we still don’t have the opportunity to go and learn them. So we have done some basic experiments in this area and we want to expand upon that work. Then, science in schools is a very important area for us and we hope to do a lot more experimentation and product development as we go ahead. So, these are some other spaces. I don’t know what else, but there are some things that we are kind of doing which actually will take off in the next five years we believe. We are going to see!

Interviewer: Thank you so much for your time and we wish you all the best for your future endeavours.