The Telecommunications for Development (T4D) project is designed to develop ICT resources that are directly relevant to the South African context with the explicit aim of socio-economic development. The essential idea is to develop software frameworks and processes for producing reliable, cost-effective, scalable, and efficient solutions on a variety of network platforms. The solutions themselves use, are based off and feed back into designing better digital infrastructure. The focus on developing prototypes means that industry partners could have a ready idea of how the research will be useful and exploitable. Outputs from the project include open-source artefacts and projects for use by the wider practitioner and research communities, and that can serve as the basis for further development by industry partners. Most, if not all, work packages will result in the development of software prototypes e.g., the SimpleDL toolkit, routing data analyser/visualization platform from ARDA project, and the iNethi platform and services.
The T4D project can be split into five work packages: T4D-WP1, T4D-WP2, T4D-WP3, T4D-WP4, and T4D-WP5. The research theme of these work packages is predominantly ‘Environmental Sustainability and Social Impact.’ A few of the work package projects also align with the themes ‘Connected Intelligence’ (T4D-WP4) and ‘Data-Driven and Computational Research and Applications’ (T4D-WP5).
The T4D work packages are detailed below:
4D-WP1: Community Centred Communication, Education and Content Creation
Theme: Environmental Sustainability and Social Impact
While many development projects are top down, with goals, funding, and messaging coming from donors and government organizations, it is also well known that grassroots development has great potential for better targeting the needs and realities of community members. In this project, we propose to take a community-centric view on development of technological interventions intended to help people in low-income rural and urban areas.
This has three main aspects. Firstly, we investigate the mechanisms by which people communicate, proposing that locally hosted services on community networks might help make networked communication more affordable. In this component of the project, we seek to establish community networks, to co-ideate, co-design and co-deploy local services, and to study community conceptions around internet pricing and usage. We will also study how these local services can synchronise and integrate with global services, opening the way for collaboration between communities as well as greater community representation on the Internet. Much of our research is being realised in the iNethi project, which provides a set of local services to communities where connectivity is unaffordable. Our flagship deployment is in Ocean View where we are carrying out research on co-designed services and analyse the impact by studying network usage and behaviour. At the request of local communities, we also have iNethi deployments in Sweetwaters, KZN; Kilifi, Kenya; and in Limpopo. Each of these deployments serves as a living lab for rethinking networked services for communities. A key agenda emerging from our engagements so far is the need for more community- and human-centered network management interfaces. We bring methods and knowledge from human-computer interaction in concert with networking to explore how to make community networks easier to manage, maintain and sustain.
Secondly, we seek to leverage mobile devices as a tool for community education in health and other sectors. Our aim is to develop and test design mechanisms for improving retention of material and persuasion of the recipients towards more healthy behaviours. Finally, we target the very significant gap of improving bottom-up messaging, enabling creation of multimedia content, as well as bridging the communication gap between community members and the local and global community. This work dovetails with the iNethi project; but we also explore alternative structures to support content creation and dissemination in communities.
For all this research, we recognize that for many of our users, the mobile phone is the first and often only means of access to the Internet, so we also seek to understand the resultant ramifications for design of mobile devices and applications. In addition, we observe the ways in different communications technologies might be leveraged in complementary and evolving ways to result in a more effective braided communications channel. Our research will add to theory on braided communications, leveraging it in more holistic interventions that recognize that mobile phones are not a panacea.
T4D-WP2: Low-resource Education Technology
Theme: Environmental Sustainability and Social Impact
Low-resource environments are environments where resource limitations have a fundamental impact on the design of technology. These constraints include electricity, Internet bandwidth, availability of skilled staff and availability of equipment. Low-resource environments include so-called developing countries but solutions that are efficient and robust in using resources can benefit virtually any environment; the low-resource constraint results in a focus on better algorithms and designs that all users can benefit from.
Educational Technology refers to software and hardware solutions designed specifically to support the teaching and learning processes. Traditionally, this was defined to include Computer Aided Instruction/Learning (CAI/CAL) and, more recently, Learning Management Systems. Now, however, a much broader working definition includes mobile applications for learning, intelligent assessment systems, school management systems and tools to assist teachers in enabling learning. Educational technology has been widely adopted at both universities and schools, with the goals of either improving efficiency or keeping abreast of trends and/or effecting interventions to enhance the teaching and learning processes and outcomes.
The recent advances in AI are a new area of investigation because the higher degrees of automation substantially lower the cost of traditionally labour-intensive tasks (like giving students feedback, grading large numbers of assignments, and detecting plagiarism).
T4D-WP3: ICT for bandwidth-constrained users.
Theme: Environmental Sustainability and Social Impact
Early efforts around ICTs for development have focused on access as a primary goal, positing that access to internet resources was sufficient to bridge capabilities and empower people to learn and advocate for themselves. And yet, despite mandates for universal broadband, and wide accessibility of mobile internet, inequalities persist. Assumptions of always-on unlimited data built into mobile applications make them unusable for people with data caps and pay-per-kb plans. A possible answer to this includes zero-rating of key services – and yet research has shown that zero-rated services are often seen as degraded experiences, and that despite zero-rating, charges still happen due to the multi-sourced nature of today’s websites. Universal broadband also fails to be an answer – as broadband becomes more available, applications and systems demand more data, with the last mile always having slower and more costly per-kb internet. In this research we propose a different point of view. What design modifications can we make to apps to make them more inclusive of bandwidth constraints – not using data unnecessarily and being aware of the differences between uncapped WiFi and mobile data (or even capped WiFi). What recommendations can we make to organizations seeking to be more inclusive of bandwidth-constrained (and often otherwise-resource-constrained) people? This starts with simple design patterns such as download-on-demand (e.g., photos and video on WhatsApp). In this research we work with bandwidth constrained users to identify and design ways to help them use their bandwidth more effectively, with a goal of offering recommendations to app and service designers on how to be more inclusive in their designs without compromising functionality.
T4D-WP4: Networking for Resource-Constrained Environments
Theme: Environmental Sustainability and Social Impact; Connected Intelligence
As the Internet continues to be a critical tool for socioeconomic development, the need to understand and address performance challenges in developing areas becomes increasingly urgent. This research is particularly focused on studying Internet access and performance issues, such as limited capacity in certain parts of the network, suboptimal routing, security vulnerability, and the impact of traffic management middle-boxes like proxies and firewalls.
This research aims to design adaptive, data-driven network architectures and explore mechanisms to enhance the effectiveness of Internet systems in Africa and other developing regions, addressing unique resource constraints and network challenges. The overarching goal is to improve the quality of service (QoS) for Internet users in these regions.
The research is two-pronged:
- Building Pan-African Capacity: The first objective is to establish a robust framework for collecting, collating, visualizing, and evaluating Internet-related data across Africa. This bottom-up approach aims to identify specific technical and socioeconomic problems, such as Internet bottlenecks, that hinder network performance. Activities will include longitudinal Internet measurement studies, especially in rural and peri-urban areas, with a focus on automatic detection of key access and performance bottlenecks and resource optimization for constrained networks.
- Technical Exploration and Innovation: The second objective involves exploring advanced techniques for improving network performance and security. This includes leveraging AI techniques and Software Defined Networking (SDN) for performance engineering in low-resource networks and investigating the intersection between Internet performance and security, privacy, and censorship. The aim of this research is to design adaptive data-driven network architectures that are responsive to resource constraints and network challenges that exist in Africa and other developing regions. The specific research goals are:
- To explore data-driven network engineering using AI and SDN in low-resource network environments.
- To examine the intersection between Internet performance and issues of security, privacy, and censorship.
- To extend previous work on network measurement and analysis of Africa’s Internet topology, with a new emphasis on access, user-centric performance measurements (quality of experience), and the study of performance bottlenecks and longitudinal trends.
T4D-WP5: Digital Libraries for Development
Theme: Environmental Sustainability and Social Impact; Data-Driven and Computational Research and Applications
Modern digital libraries are built using reusable software toolkits such as DSpace and EPrints or commercial service provider systems. Most South African universities use DSpace for their institutional repositories and similar archival systems, in a bid to pool resources and share expertise. However, the use of these tools is still a daunting prospect that has led to universities like UCT outsourcing their management to overseas companies. Clearly, the fundamental software architecture is too complex for use in low-resources environments, and the net effect is the failure of archives in poorer institutions across the continent.
In addition to the archives, search technology has rapidly progressed in other parts of the world because of large datasets combined with deep learning, producing popular language models such as BERT and GPT. These are, however, not readily applicable in low resource environments because of the lack of data and questionable bias embedded in the global models.
The goal of this project is the development of appropriate systems that are easier to install/configure/customise/maintain/scale/migrate and are better suited for local communities e.g., by supporting local languages or using algorithms to overcome inherent systemic bias. One specific focus is on the use of machine learning for retrieval and mining of datasets in local languages and representing local cultures and local problems.