TechBridgeWorld

M. Bernardine DiasFall 2007

2

TechBridgeWorld

3

We are a multidisciplinary team embedded within Carnegie Mellon University,

spanning education, research, development, deployment, and outreach,

dedicated to defining the role of technology in sustainable global development.

www.TechBridgeWorld.org

4

Some day

By knowing them, we will learn to create technology of impact to them

5

Some day

The technology we create will help in the battle against poverty

6

Some day

Some of them will become technologists

our students and colleagues

7

Because today

We give students non-traditional educational opportunities!

8

Building Technology Bridges

9

Technology Consulting in the Community (TCinC)

Department: Computer Science DepartmentPrerequisites: typically 3rd or 4th year standingOffered: Fall and Spring since 1998Role: Satisfies other computer science requirementUnits: currently 9, considering going to 12 (equivalent to 3, going to 4 credits at other schools)Typical majors: CS, IS, ECEStudents per class: typically 12-25

10

PartnershipsEach student is partnered with a leader in a non-profit organization

E.g. Executive director, program director, technology coordinatorStudent works one-on-one with that community partner (and key staff people they designate)On site at the non-profitFor 3 hours every week

11

Technology Consulting in the Global

Community (TCinGC)

www.TechBridgeWorld.org/tcingc

12

Summer 10 week programStudents travel to consult with governments and non-profits in developing communities abroadUses the same consulting model as TCinC courseFinancial break-even deal for student consultantsProgram pays for travelCommunity partner provides accommodations and a small stipend to cover local transportation, food, and incidental expenses

Technology Consulting in the Global Community

13

The V-UnitLearning to Build a Vision

www.cs.cmu.edu/~vunit

14

FocusTechnology for non-traditional applications

Strong support networkV-Unit coordinatorsProject advisorsCommunity partnersDepartment, School, and University

DisseminationPublic oral presentation Written report

Key Elements

15

Poverty simulationSeminar seriesRobotics course in GhanaE-Village (online community)Wish listIndividually catered opportunities (internships, partners, funding, introductions, projects, etc.)

Other Activities

TFDC Examples II

M. Bernardine DiasFall 2007

17

Categories of TFDC Apps

EducationAgricultureHealth E-GovernmentEmploymentAssistive TechnologySafetyOthers

18

Education: ChallengesLack of resourcesLack of access to expertsProfessional and intellectual isolationLack of support for researchMistrust of technology (often due to inappropriate transfers of technology)Lack of access to professional organizations, literary works, academic journals, etc.Notion that education and technology are for the richTraditional and cultural barriers

19

Technology Enhanced EducationAutomated tutorsWeb-based toolsDistance educationTeaching/learning tools for the physically and mentally challengedComputing infrastructureRobotsTeacher training resourcesResources and tools for teachersMultimedia displaysVirtual experiences

20

Technology EducationTechnology coursesHands-on labsTrade/Job skillsTechnology teacher-trainingResearch opportunitiesWeb resourcesExchange/Study Abroad programs

21

Multiple Mice

Udai Singh Pawar, Joyojeet Pal and Kentaro Toyama, Multiple Mice for Computers in Education in Developing Countries, ICTD 2006

http://tier.cs.berkeley.edu/docs/ict4d06/multiple_mice-jp.pdf

22

IntroductionAddresses the challenge of lack of resourcesComputers can be expensive but mice arentOften many children use one computer in developing communitiesGroup engagement can be a positive element in the learning processShould allow several children to share a single computer with multiple miceNeeds relevant software applications

23

Results & Challenges

Developed several educational gamesA cursor with a unique color is attached to each mouseGames were implemented in English, Hindi, and KannadaPreliminary field trials were conducted

Questions answered:o Can children understand and

use the multiple-mouse paradigm when the number of mice is as many as five?

o How do children interact with each other with respect to multiple mice? How do they share or not share?

o Does the multiple mouse paradigm increase interest and engagement?

Observations:o Gender difference in sharingo One child didnt like

competitive element

24

Future WorkBuild more educational applicationsMore extensive field testsEvaluate tangible learning outcomesAdaptive software for simultaneous multiple usersJoint decision making mechanismCollaborative applications

25

Project Kan

M. Bernardine Dias, G. Ayorkor Mills-Tettey, and Joseph Mertz, The TechBridgeWorld Initiative: Broadening Perspectives in Computing

Technology Education and Research, 2005http://www.ri.cmu.edu/pub_files/pub4/dias_m_bernardine_2005_4/dias_m_bernardine_2005_4.pdf

26

IntroductionGhana is seeking means to improve English literacyOne major shortcoming with available resources is the opportunity for guided reading practiceCarnegie Mellons Jack Mostow and group had developed an automated English reading tutor project LISTEN

27

Results & ChallengesField study designed as TDC class projectPilot study was implemented over the following summer with good resultsSearched for follow-on funding and partnershipUNESCO funding and partnerships secured in 2006 summer

Many challenges in implementing follow-on study including very limited fundingDifficulty in ensuring attendance for long-term studyInitial phase was carried out in Internet caf; wasnt possible for follow-onBeing a remote partner is difficult

28

Future WorkComplete on-going field study comparison to guided practice with an older youth Seek creative ways to overcome recent licensing of the tutorSeek creative means of funding follow-on workFind other partnershipsLots of demand!English tutors for adults are also in high demand

29

Robotics Course in Ghana

G. Ayorkor Mills-Tettey, M. Bernardine Dias, Brett Browning and Nathan Amanquah, Teaching technical creativity through Robotics: A case study in

Ghana, 2007http://www.ri.cmu.edu/pub_files/pub4/mills_tettey_g_ayorkor_2007_1/mills_tettey_g_ayorkor_2007_1.pdf

30

IntroductionSustainable development requires increased diversity in the creators (and consumers) of technologyChallenges to successful technology education in developing communities include limited access to infrastructure, equipment, and expertiseChallenges can be mitigated by creative course design and collaborative partnershipsPartnership with Ashesi University in GhanaCourse was designed and co-taught with Nathan Amanquah head of CS dept.

31

Results & ChallengesDesign philosophy:o Encourage creativity o Use local resourceso Teach technical skillso Teach dissemination skillso Inspire with examples of

state-of-the-arto Encourage a broad

understandingo Impact and involve local

community

Outcomes:o Knowledge and

technical creativityo Technical skillso Confidenceo Awareness of

research and professional organizations

o Impact on other academic endeavors

o External interest

32

Future WorkFind ways to continue to offer the course at Ashesi UniversityDisseminate course to other universities E-Village: Online educational resource for TFDCFind ways to enhance access to professional organizationsAlum start-up company: e-wallet for mobile-phone banking!

33

Agriculture: ChallengesLack of resourcesLack of access to expertsMistrust of technologyChallenges with land ownershipObtaining good market pricesObtaining trustworthy, relevant, and timely informationTransportation and storage

34

Technology Enhanced AgricultureInformation kiosksAutomated soil mappingGM cropsAgricultural toolsOnline training and educational programsCrop protectionStorage and transportation devicesQuality controlSensing and monitoring systems

35

aAQUA Agricultural Portal

K. Ramamritham, A. Bahuman, S. Duttagupta, C. Bahuman, and S. Balasundaram,Innovative ICT Tools for Information Provision in Agricultural Extension, 2006

http://www.cse.iitb.ac.in/~krithi/papers/ICTD2006.pdf

36

IntroductionaAQUA: almost All QUestionsAnsweredAgricultural portal in IndiaMain goal to answer questions from farmersUses novel database and information retrieval techniquesExperts provide answers/solutions which are cached for future useFarmers can provide feedback and experts respondPictures and multiple languages are used

37

Results & ChallengesExperts response times average 36 hoursHelp-desk operator follows up with farmers a week after the solution is providedExpert calls the farmer to follow up if solution failedFarmers can call the center if online system cannot be accessed

Interface suitable for novice internet usersCan be accessed easily through low-bandwidth or intermittent connectivityAllows recycling of expert answersProvides crops library and crop doctor

38

Future WorkAccess through mobile devicesExperimenting with radio, SMS, and telephone accessFarmers can register for alertsExtend to providing more information such as weather reportsExtend to applications in education, healthcare, eGovernance, etc.

39

Statistical Soil Mapping

Juan Pablo Gonzalez, J. Andrew Bagnell, Simon Cook, Thomas Oberthur, Andrew Jarvis and Mauricio Rincon, Gaussian Processes for Statistical Soil Modeling of the Tropics, 2005http://www.ri.cmu.edu/pub_files/pub4/gonzalez_juan_pablo_2005_2/gonzalez_juan_pablo_2005_2.pdf

40

IntroductionStarted as a V UnitWorking with CIAT Intl. Center for Tropical Agriculture (situated in Columbia)Used Artificial Intelligence/ Machine Learning techniques to dramatically improve soil mapsSoil maps:o 1:24,000 ratio (30 m scale); 100%

coverage: USAo 1:5,000,000 only (over 5 km) based

on US soil taxonomy: FAO map

41

Results & ChallengesUsed:o 2500 soil samples from

Honduraso Digital maps of

Honduraso Derived variables

Honduras (112,000 km2)o 40 minutes @ 1kmo 3.4 days @ 90mo 30 days @ 30m

Applied Gaussian Processeso Quantitative predictionso Quantitative estimate of

confidenceo Complete coverageo Continued improvement

Matched/advanced state of the art in predictive soil mapping

42

Future WorkCompare performance with other learning techniques (mainly with regression techniques)Validate results in other locationsUse active learning for collecting new data and improving modelsPartner with CIAT on other projectso Weather insuranceo Species distribution

43

Text-Free User Interfaces

Indrani Medhi, Aman Sagar, and Kentaro Toyama, Text-Free User Interfaces for Illiterate and Semi-Literate Users, 2006

http://research.microsoft.com/users/indranim/Text-FreeUI.pdf

44

Introduction

User interface for novice/illiterate users (with no other help)Domestic laborers in Bangalore slums were chosenTwo applications:o Job searcho City navigation

45

Results & ChallengesDesign principles:o Avoid texto Numbers are oko Use semi-abstracted

graphicso Photo-realism with

deeper interactiono Pay attention to subtle

graphic cueso Provide voice feedbacko Provide help at all stages

Text-free but not click-freeLandmarks were important for navigationTesting conducted in homesUsed trusted contactsBollywood Method to encourage feedbackCollaborative use

46

Future WorkEthnographic process: 180 hours with 80 men and women from Bangalore slumsStrong preference for text-free interfaceHelp on every page was importantInclude short movie that loops at the beginningMove towards use-studies with no external assistance

47

Braille Tutor

N. Kalra, T. Lauwers, and M. B. Dias, A Braille Writing Tutor to Combat Illiteracy in Developing Communities, 2007

http://www.ri.cmu.edu/pubs/pub_5676.html

48

IntroductionCollaboration with Mathru School for the Blind > 90% of worlds blind population live in developing countries< 3% of them are literateBraille:o Tactile writing systemo Six dots per cello Write by embossing papero Traditionally with slate and stylus

49

Results & ChallengesSuccessful 6-week long field test in Summer 2006Tutor catered specifically to needs of students and teachers in under-resourced communitiesDesigned to be low-cost and robust

Currently has activities for 3 grade levels Quantitative predictionso dots - letterso wordso sentences

In use at MathruSecond version of tutor developed based on feedback

50

Future WorkBattery-poweredIndependent of computerAdaptive to skill level of userLonger, more extensive field study at MathruField studies at other locationsOther languages (Arabic Braille)Games to increase enthusiasm

51

Project for Healthcare in Pakistan

Speech-driven system for paramedical workers in Pakistan (e.g., Lady Health Worker Program)Background (Motivation)o Intersection of

PeopleMotivated faculty and graduate student from Pakistan

FundingMSR Digital Inclusion RFP

StakeholdersVisit to Pakistan found Aga Khan University, a leading medical

school Has substantial community health program

52

Research Questions[This is a research project, not a full deployment]

Will it work?o Urdu language speech recognition is limited

What interfaces and dialog systems are best?o Natural conversation, directed dialog, keyword

searching, etc.Are semi-literate users a compelling market?o Literate can read, instead; illiterate may not have a

substantial driver to use

53

Design SustainabilityTechnologyo Centralized design, using phones in the field, makes

maintenance and upgrading simplero Can be maintained by a motivated IT user (doesnt

require a speech researcher)Financialo Preliminary analysis indicates positive cost-benefit

analysis, indicating ongoing government or donor support would be justified

Stakeholder incentiveso If it works during the controlled trials, will users

continue to use it subsequently?

54

Ongoing and Future WorkDeveloping a Prototypeo Technical challenges

Vocabulary Acoustic models

Healthworker interactiono Dialog preferenceso Needs Assessment

Training, Diagnosis, etc.

Determining systems and incentive structures