Bandwidth connectivity is extremely low in many parts of the world, often delaying or even preventing people from accessing information. Although connectivity has increased by a factor of 10 or 15 over the last decade, average webpage size and number of objects has simultaneously increased by a factor of 60. This has resulted in what Subramanian terms an "unusable web"; the addition of video, audio and images has created huge web pages that take minutes or even hours to load in low connectivity worlds.

In his talk, Subramanian describes a range of techniques that he and his colleagues are developing to enhance information access in three scenarios of poor connectivity. These techniques illustrate the type of technologies being devised by a new group in Computer Science termed "Computing for Development." The focus of this group is on the design, implementation and evaluation of new computing innovations that enable global social and economic development. Since first world technology can often be a bad fit, this group instead seeks technologies that are locally appropriate, cost-effective, and easy to use.

The first low-connectivity scenario Subramanian discusses is that facing rural mobile users, who rely on low-end mobile devices and can thus only access voice and SMS services. To address the massive need for SMS services for this scenario, Subramanian and his colleagues have developed an SMS-based protocol stack for mobile applications that makes it possible to compress large quantities of information. The so-called UjU stack enables the compression of information into a 140-byte stack, while an affiliated UjU Create App interface enables anyone to create their own apps and forms. These forms are essentially turned into structured records (tables) that can be filled out and transmitted through a short message on a mobile phone. To date, UjU has already been used for microfinance applications in Mexico, mobile health data collection in India, and other applications in Ghana. Subramanian and his colleagues are also rolling out a live SMS search engine in Kenya and a data-over-GSM voice stack to support data connectivity over cellular voice.

Shared low bandwidth networks present a second low-connectivity scenario. In this scenario, an example of which might be a school where 2 Mbps of connectivity is used by 400 students, Subramanian suggests that a completely new Web architecture is needed. He and his colleagues have deployed an early version of such a system called Rural Café User Interface. Typically, a web browser sends dozens of requests when it is loading a particular page, since each site draws content from various sources and advertisers. Rather than being able to attempt to load as many windows and pages as possible, which results in even slower access, every user in Rural Café has a queue of what their search requests are. Users can search for anything at any point, but the interface acts as a planning tool by reporting how long (in seconds or minutes) the user would have to wait to load any particular site. The queue is persistent, so it doesn't change depending on how many new windows users try to open. This system is already being deployed in a few schools and universities in Kenya and India.

The third scenario is that of schools that have computer access, but no connectivity. To address this problem, Subramanian proposes the use of "vertical search engines" or contextual information portals that deliver a vertical slice of the Web in a hard-disk and provide an offline searchable and browse-able Internet. The portals are locally searchable and composed of many web-based services. Since the portals allow the user to search the local cache for the information they need rather than the URLs themselves, many local requests can be handled without browsing, supposing the local cache is strong and based on local interests and content. This is a good tool for either improving download times for people who have limited connectivity, or for enabling access for information for people without any connectivity at all. So far, the system has been piloted as an information tool for students and teachers in five schools with computers but no connectivity around Nairobi, Kenya.

The central premise underlying Bussell's discussion this week is the reality that public service provision is often flawed in the developing world. Reforming public services entails significant efforts to increase the quality of public service, and there have been two important recent trends in this space: privatization and public-private partnerships and increased use of information and communication technologies (ICTs). The one-stop services model (which has been adopted globally in various forms, from Singapore's eCitizen Centre to Germany's Buergerbuero or "Citizen's Bureau"), exemplifies current trends.

In her research, which focuses primarily on the case of India, Bussell asks the following questions: why do or don't governments reform public service delivery? And when governments decide to implement reforms, why do they or don't they employ ICTs in these reforms?

The Indian strategy to public service reform has been to create stand-alone, one-stop citizen service centers. These centers are computerized and frequently have private-sector participation. Aside from these typical characteristics, there is significant state-level variation on several points, including timing of policy adoption, quantity and type of services, degree of automation, and ownership and management models.

Bussell discusses a variety of hypotheses that offer insight into the potential factors influencing the character of Indian reforms. First, she notes that politicians faced mixed electoral incentives for and against reforms. After all, there is evidence that eServices can improve service characteristics and reduce corruption. On the other hand, reforms may also reduce politicians' opportunities to amass funds to run for re-election. After all, bureaucratic discretion and opaque processes enable the siphoning of funds and bribe taking in service delivery. Politicians use "transfer authority" over bureaucrats to access bribes. More transparent services, due the implementation of reforms, threaten this access to bribes. At the same time, reform may also offer new opportunity for rents. Private partnerships to run centers entail new contracting processes. Larger-scale bribes may be available from ownership and management negotiation.

Based on these mixed direct and indirect incentives, Bussell predicts that reform will require expected net benefits to ruling politicians. States with higher petty corruption should implement policy reforms later. Meanwhile, states with higher grand corruption might lead to more partnerships with private sector.

So what factors actually explain variation in when different Indian states adopted reform? According to Bussell's research, the level of corruption was the most statistically significant variable affecting the timing of reform adoption. However, whether or not there was a coalition government in power was also significant. This means that an increase of 1 point out of 10 in the state's corruption level led to a 63% decrease in the chance of a reform being adopted in a given year. The quantity of services covered varied widely in the sample of states from less than 10 to more than 40 services. On average, moving from a state with below average corruption to a state with above average corruption causes a drop of 14 in the number of services covered.

In addition to examining the variation in reform adoption across Indian states, Bussell also looks at the consequences of reform. Despite reforms in the state of Karnataka, to give one example, demand for services continues to outstrip supply. Visitors to Nemmadi (privately-run computerized centers) were able to access services faster, while paying less money, making less visits, spending less time waiting at each visit, and seeing a reduction (on average) in the number of days before receiving the service. Despite quantitative improvements in average efficiency, however, perceived efficiency actually declined. Now, there is a 7-day minimum for service delivery and a 21-day maximum. Although average has declined as a consequence of these new requirements, it is now impossible to bribe officials in order to achieve service delivery in one day.

According to Bussell, these results bring up some interesting policy implications. First, it is important to consider the institutional incentives underlying the established model in order to ensure that reforms are effective. The incentives faced by both top politicians and street level officials must be taken into account. Second, it may be best to design policies that establish a strong, if narrow, initial model. After all, growing citizen demand affects electoral benefits and the calculations of politicians. Third, more research must be done regarding the various factors that influence policy outcomes. According to cases in South Africa and Brazil, there is some evidence that differences in electoral competition can affect policy outcomes. In South Africa, for instance, less electoral competition has led to less incentive to reform. More research will help to clarify the factors of importance in public service reform implementation.

Drawing upon his experience with India's Right to Information movement, Vivek focused his discussion on how information and communication technology (ICT) tools could be designed and applied to strengthen people's movements to combat corruption. Of course, Vivek conceded, ICT cannot combat all kinds of corruption. These tools can be very effective, however, in combating types of corruption for which there is a paper trail attesting to something that never happened (such as the construction of a road or the provision of grain subsides or other goods).

In the past, it has been possible for members of people's movements working to combat corruption to request lists of all government programs going on in a village to monitor who received what benefits. After summing up this information over a long period and comparing notes with the villagers themselves, activists have then been able to expose inaccuracies in government records through public hearings.

Although activists can carry out this sort of fact checking without the use of advanced ICTs, the introduction of such ICTs has helped social movements work much more effectively to combat corruption. After all, an individual who goes to a government office to obtain public information will often face significant resistance. Requiring that government offices make information available online makes getting public records much easier. Additionally, cross-comparisons of data created by different government agencies (i.e. comparing ration card data against census information for each village) can be much more easily executed once this data is online.

In some cases, changing procedures can help reduce certain types of corruption. In Kathmandu, for example, public officials began to be required to wear shirts and pants with no pockets to reduce exchange of petty bribes. Creating procedures like these are very useful, Vivek emphasized, but they can also be enhanced through the use of ICTs. In the Indian state of Tamil Nadu, for example, public managers of public works projects began to be required to send text message at 10:30 in the morning, to ensure that officials accurately reported the number of persons employed at the site. Since these messages can be sent to all interested parties, anyone can photograph the site with their cell phone to expose officials' misreporting. This example illustrates how timely verification and dissemination of information can establish whether information is being falsified.

New technologies are also enabling the reporting of new types of information. New kinds of accounting include cross-verification, biometric verification, image-based processes (i.e. video and audio), and geo-specific information (i.e. through RFID, a low cost passive electric tag). Although what you can verify (i.e. teacher presence at the school) is not always the same as the indicator that is truly important (i.e. student learning), the use of these new reporting methods can often raise the cost of cheating.

In closing, Vivek noted that unless people are mobilized, they will not do anything to combat corruption. Once systems are in place however, technology can make any mobilized groups work more effectively. To maximize the ability of activists to extract information from the grassroots level, we need new forms of accounting and dissemination that are user-centered and not divided up by governmental department. Separating implementation agencies from payment agencies will be another positive approach in the attempt to reduce corruption. As entitlement programs grow due to the increased emphasis on a rights-based approach to international development, the need to combat various kinds of corruption is growing, and the application of ICTs offers a big step forward.