Abstract Community health programs in many developing countries are very similar: most involve semi- trained, semi-educated health workers (often fe- male), who provide health services in their own communities in a large hierarchically managed sys- tem across the country. In Pakistan, recent evalua- tions of the government's flagship community health program have revealed the need for im- provement in health workers' knowledge, and rec- ommended frequent refresher training courses. However, these have not been possible to imple- ment primarily for reasons of cost. It is our conjec- ture that telephone-based adaptive speech inter- faces, tailored to the specific information needs of health workers, are a viable alternative to current information access mechanisms such as handbooks and manuals. We will be designing, developing and testing such an interface (“HealthLine”) in the coming year in Pakistan. Initial results from dis- cussions with health workers (using video proto- types) indicate that speech interfaces may poten- tially be much more useful for health workers than traditional media. 1 Introduction Just under half the world’s population, or around 2.8 billion people, currently live on less than 2 dollars a day. 1 There exist a number of domains where Information and Commu- nication Technologies (ICTs) can provide real value to such populations, in a way that is both sustainable and appropri- ate [Mansell, 1998 & Tongia, 2005]. There are hundreds of such projects (as cataloged by the World Bank at http://www.infodev.org), most of which use existing, off- the-shelf technology. However, the direct transfer of “First World” technology has not been successful in most cases, primarily because of the mismatch between the intended environment the technology was designed for, and the ground realities of the environments in which they are de- ployed [Brand & Schwittay, 2006]. This is eloquently de- 1 This is in purchasing power parity (PPP) terms. scribed in [Brewer et al., 2005]: “Although it is clear that there are large differences in assumptions related to cost, power, and usage, there has been little work on how tech- nology needs in developing regions differ from those of industrialized nations. We argue that Western market forces will continue to meet the needs of developing regions acci- dentally at best”. Research on technology design tailored to the specific needs of emerging regions is needed to address this issue [Brewer et al., 2005, Brand & Schwittay, 2006, Tongia, 2005, Tongia 2006]. Many ICT-for-development (ICT4D) initiatives involve the use of standard PCs as the form factor, and standard web- based forms or Windows-based GUIs as the primary inter- face, and the Internet for connectivity. However, PCs and current GUIs were designed with a specific (“First World”) user in mind: a user who can afford a roughly $500 machine as well as Internet connectivity, has access to a stable elec- tricity supply, is literate, uses a language that has a written form, and finally can access and afford technical support when something fails or needs to be upgraded. These re- quirements are unrealistic for major portions of the develop- ing world, where users in many cases cannot afford such costly technology, do not have access to continuous electric- ity, are not functionally literate, may be fluent only in a lan- guage without a written form, and do not have access to any ongoing support for using unintuitive technology. While there are successful ICT projects using PCs (for example, the widely publicized e-Choupal initiative in India [Kumar, 2004]), there is a large part of the developing world for which such design is not viable. It is no surprise that rates of PC use in the developing world are dwarfed by those in the developed world. Cell phones, on the other hand, are a huge ICT success. Cell phone use across the developing world is increasing re- markably [ITU, 2003]. The extensive use of cell phones suggests that this is a fundamental mechanism through which underdeveloped regions are benefiting from ICTs, because they are easy to use, affordable, and suitable for non-literate populations; providers also find them profitable. Furthermore, the sustained use of cell phones in these re- gions also implies the existence of widespread ecosystems of supply, maintenance and technical support, which do not Towards Speech Interfaces for Health Information Access by Semi-literate Users Yousuf Memon, Mehtab Karim, Gregory Pappas Aga Khan University Community Health Sciences Dept. Stadium Road, Karachi, Pakistan {yousuf.memon,mehtab.karim,gregory.pappas}@aku.edu Jahanzeb Sherwani, Rahul Tongia, Roni Rosenfeld Carnegie Mellon University School of Computer Science 5000 Forbes Avenue, Pittsburgh, PA 15213 {jsherwan,rtongia,roni}+ijcai@cs.cmu.edu