ICTD2007: Session1 - Designing Applications Session Chair - John Canny Braille Tutor Slates, developing communites and literacy * Literacy Rate in blind population is 3% * Slate and sylus as a writing tool of choice - uses no power, portable, affordable * But using it is hard! Learning what problem to solve * started openended collaboration - what was life like for people in these communities - received videos of students from schoool - questionairs abou what parts of their life were difficult * Work with blind adults in Pittsburg, as wlle as CMU - form design goals for tech apporac low cost low power easy to understand easy to poerate locally maintainable robust transferrable learning First generation hardware: * Stylus wire, input area - pressing a dot isth a stylus * economical, few chips, about $40 * on/off switch * power plug, serial cable First gen software * three levels of software 1. second standard tutor: feedback on braille cell dot and letter input 2. third standard tuto: additional feedback repearts entered words, teachis positioning of letters in cells for spelling 3. fourth standard tutor In the field in Mathro * six week pilot study, 12 students in process of learning bralle writing * classroom environment -learning * first time collaborating in person * under observation during study: studnt accpetance, changes in proficiency, ease of setup and use Demonstration of tutor in hallway area Results * not statistical analysis * but explanation? * proficiency changes - 4 students already procificent and used tutor for practic - 5 students showed measured performance gains during the study period 2 students already good, hard to know if improvement is natural or result of the tutor. but 3 or 5 showed conceptual advancements 1 was having problems - revealed that they knew letters but didn't know that they were supposed to move from cell to cell - 3 students no benefit * Acceptance - hope students would be excited - no clear if entusiastic * Observations at mathru led to several improvments - noticably smaller - computer interface - serial connector was difficult, move to usb - button arrangement as 4+styleus to 7 buttons, with 6 in a large cell - kids don't have dexterity to punch holes - students instructed ot avoid wires - so wires scary A goal of software - move from instant feedback to microlearning aspects of teaching to software. Use the computer to monitor... New softwaer: towards adaptive tutoring Don't take over for the teacher - instead teacher picks a learning mode from a set of box. software tracking students within a mode. Help them focus on tasks Dynamically figure out level and work from there, rather than individual identities Deploying again at Mathru this summer Collaborating with institutions in the US to release software etc Start: 10.03 5 min: 10:22 2 min: 10:25 Stop!: 10.27 Design of early warning flood detection systems Elizabeth Basha CSAIL Working in Honduras * northeastern area * annual hurricanes, project instigated by Hurricane Mitch * villages had no knowledge that it was coming * also flooding due to heavy rain Collaboratin wtih CTSAR * they initiated * began jan 2004 through class project * 9 trips to date. Problem Defn * warn people of flooding 24 hours in advance * need - prediction of event - large georgraphic area - low cost - limited recurring costs - ability to adapt to socio-political changes * focus on technical solutions - somewhere between tech and policy side (policy is more partners Soln: autonomous sensor network Related work * zebras in kena * road monitorring in sri lanka * arsenic monitorin in bangladesh Event prediction problem sensing -> models and computation -> comunicate prediction -> -> authorities and NGSOs -> Communities System overview (graphic) Problem: sensing * What sensors to use? * how to protect from weather and theft? * sensors, electronics, etc. Electronics * computational servics * sd cards for data storge * expansion board for various sensors * water resistant boxes Problem: Models * statistical learning models * one hour prediction, * + also lookup table based on community historical knowledge Problem: Offices * how to commmunicate - alerts, system status * local vs remote * solution: offices - wireless infrastrucuter, and backup solar for offices Problem: Communities * type of alert * illiteracy? * authorization needed for notification Radio Testing * 53 km River level measurements? * various prototypes * running long term tests in dover, massacusets, - 3 sensors - testing algorithms, etc conclusion * flooding warning is complicated * but can use automated sensor networks (how much does it cost? to maintain?) Start: 10.28 5 min: 10:47 2 min: 10:50 Stop!: 10.52 Digital Green :Participatory Video for ag ex (Rajesh's project) Rikin Gandhi Particaptory video for highly local and highly relevant infor to farmers Current sources of info * other progressive farms * existing social networkers not private ealers * radio, tv, * extension workers (100.000 across the country) Farmers are not getting expert info that they need. Technology interventions * aAgua: pull pbased question and answer (almost all questions answered) * eSagu: pushed based expert review of digital photos (krishna reddy) * Kiosks with internet access for farmers Field Reality * Collaboration with GREEN Foundation in southeastern karnataka india * Traditional extension - farmers not present, not receiving data - awarness, training, actions - intervention systems are costly with hard to find results Digital Green Production * Video centric Trustworthy Comparisons * two plots with different types of fertilizer on same land * longer time horizons to see changes * use video time-lapse to see But! * don't want to spend lots of time producing quality videos * want fresh data with new farmers * people who want to be seen as resource people * differ from traditional broadcast - don't include experts - produced in univerity setting * Juxtapose expert content with familiar farmers from local vicinity!! Producing videos for last year experiments * video producer x receptiveness - just the farmer or just the facilitator didn't helpt - various skilled faciliators with farmers - positive reception - expert facilitator was potentially detrimental Traditional Schooling * Digital StudyHall - deliver videos of good teachers to village teachers -> improve pedagogy - traditional settings - * digial green scholling: create informal school stystm using tv and video for classes - fluid audience of 20-30 farmers after working in the field * Take the video to the farmers - traditional broadcast at set hour and set place - but with video can take to different community segments Tutored video mediation * humans engage farmers - field questions, capture feedback, encourage participation - use pause and rewind to help delivery of info * social group structure around videos Also lloking at other existing social groups: women in self help groups Extension staff * also educate staff - teacher training * training to use video to motivate and educate others * prof interested in showing videos to students Distribution experiments: mediator vs receptiveness * hand out supplies, low/med/expert skill mediator, no mediator * no mediator is bad! * hole in wall television elicits small curiousity (esp children) for 2 weeks * but perception that tv cannot be valuable * experts don't connect well with farmers because farmers lookinf for familiar Local Area Network Reality * strucurin gmeetings * sharing experiences * supplying equipment * catching mistakes - film/capture, and show how to work around them * farmers innovate on what they see Network Effect * farmers are incentivised to adopt practices so they can also be on tv * youtube/idol effect Phase 1: Evaluation/Selection * tv penetration is very high (50-75%) * tried to distribute videos through local network - hole in wall cable - farmers couldn't connect with videos Phase 1: Results * measure farmers and adoption of practices * t&v control vs digital green * 6-7x more adoptions through digital green digitalgreen.org Start: 10:48 5 min: 11:07 2 min: 11:09 Stop!: 11:11 Q: To digitalgreen: What are practices that are being designated through videos Why not use kiosks? What are ways in which info can be shared? A: At this point we think of ourselves as a digital pipe for distribution of content - so videos are about Green Foundations content. They have a specific suite of interventions. Are open to kiosks. but tvs and dvd players have been cost-effective, and found that unmanned kiosk wasn't working. Q: Comparing to tv? Is there difference in info provided by normal agex officers and what you are providing? A: Practices are not that different from agext worker practices. Govt ext also teaces same practices. For compare to control villages - we try to match content on a calendar basis. Uniform seasonal centric office A: (from green foundation) There is a difference between agex and green foundation. Interest in sustainability. Knowledge systems. HOw can we communicate info that the farmers already know! Q: To early warning: drainage facilities and other things contributing to floods To digital green: gathered data on best practices, what is being done? especially with regards to sustainable agriculture? A: (early warning) Tak into account different locations (rural and urban) and how terrain affects implementaion of the system. Not looking into addressing flooding via civil engineering. Tradeoff between idenfication of flooding and actually doing something about it A: (Rikin, digital green) Green Fdn does participatory researche before going out with practices. Followup surveys after one and two years to measure changes and benefits. But it depends on practices and how long it takes to see improvments. some take a month, some take longer. try to include sequencing. Q: How does it affect debt for the family, etc - actual affects on poverty, not just measurement of the intervention uptake. Q: (Tap, to Elizabeth/early warning) How to combine with other early warning system. Envison other applications besides flood detection. A: Volunteer basis: looking at rain levels and bring that info back. current is going out on a regular basis and report water levels - low cost but unreliable. In US, expensive system requiring expertise to run hardware. looking for soemthing in between. Other applications - designed with other applications in mind - river pollution, etc. Q: Taking data vs prediction A: We are taking data every hour and predicting for 8 hours Q: Regarding wireless network - what problem? Value of petrol? ITU and broadband. What is the bandwidth for data - can it carry voice? Showed laptop being used: developed for rural - laptops in developed country don't have lifetime in developikng country conditions. How about low cost laptops. A: 140 MHz range. Free in US, permission in Honduras. Laptops don't drain power as much as PCs. Laptops are in offices of people trusted. dont'w worry as much about Q: (tom/braille) Big problem with mechanical pen. Is it it better to devleop a low cost printer? and teach them how ot use it? Q: (to digital green) participation and power traditiona l mass media could it be as effective? A: People at broadcasts, etc