Information Communication Technologies (ICTs) such as cellphones have become a part of our daily lives. The technology has rapidly been adopted throughout the world, and this is particularly true in Africa. ICTs provide opportunities for development but the expectations they tend to raise, of improved services and a general modernization, are often not realized.

The failure rate of ICT systems in the developing world is astonishingly high (Heeks 2002). Initiatives such as telecentres (centres in rural villages with computers and internet access) have been shown to fail both in the software engineering sense (they are unsustainable and need repair), and in terms of human-computer interaction (no one uses the computers as they are seen as unnecessary) (Benjamin 2001). Disciplines such as Software Engineering and Human-Computer Interaction have been successful in creating ICT systems for the developed world. In Africa, however, the picture is very different.

Together with colleagues we have developed ICTs for under-resourced and rural communities in South Africa and other developing countries; a field that has become known as Information and Communications Technology for Development (ICT4D). Our focus has been on using devices such as cellphones to develop context-appropriate and easily accessible tools that can support communities and governments in advancing their developmental goals.

In this chapter we provide some examples of our successes and failures in designing ICT4D systems. In assessing these examples, we question basic concepts used in computing such as efficiency and effectiveness, as well as scalability and efficacy. In addition, we explain why we seek to include the voices of actors other than engineers in our design processes. Echoing the findings of Thesen and Cooper in this volume, we show that questions of assessment invariably involve issues of power and location, and that some voices, generally of the most marginal actors in a process, can easily be overlooked.

Some systems are reported as successful when they work in a single location, even if their deployment does not move beyond a single community (Sørenson et al. 2008). However, many existing system design methodologies do not respond satisfactorily when applied to specific development contexts. A major reason for this is that the assumptions on which such systems are based do not fit all contexts, and the outcomes they generate end up being unsustainable or inappropriate.