Experts have cautioned that as South Africa pursues digital transformation, it must ensure that it does not widen the already significant digital divide and digital inequality in the country.

While the benefits of AI and digital transformation are often highlighted, it does have the potential unintended consequence of expanding the digital divide.

Speaking at the Government Technical Advisory Centre-hosted Public Economics Conference on Tuesday, Research ICT Africa executive director Alison Gillwald said that the digital inequality paradox was getting worse.

“As increasingly advanced technologies are being overlaid on existing inequalities – themselves a reflection of deeper structural inequality – inequalities are actually being amplified.”

“This is not only between those online and those offline, as is the case in a voice and basic text environment. It is between those who have the technical and financial resources to use the Internet actively, or even contribute to national prosperity, and those who are barely online, passively and intermittently consuming a narrow set of digital services, which the majority of South Africans are doing.”

She commented that, contrary to much of the hype around AI and the activities in more mature economies, these advanced data-driven technologies and digitalisation were exacerbating inequality and would continue to do so until the underlying inequalities were mitigated.

The extreme structural inequalities were made more visible – and exacerbated – during the Covid-19 lockdown, when many people – the most marginalised and those barely online – were unable to “digitally mitigate” or “digitally substitute” the health and economic risks associated with the pandemic, Gillwald pointed out.

This is despite more people gaining digital access.

While the connectivity rates across sub-Saharan Africa, for example, look promising at about 65% to 70%, of that only about 35% to 40% are meaningfully online.

“We need to get to critical masses. These critical masses have been determined by the number of connections on assumptions that people are always online, that people are actually transacting across Africa. [However,] people are connected, but they are not doing much online. The use levels are very low,” she said.

Gillwald also pointed to a lack of digitalisation and datafication that would result in Africans largely being “invisible”, underrepresented or discriminated against in automated, data-driven technologies such as machine learning, large language modelling and AI.

She cited an example of generative AI and ChatGPT, which was effective when used in a very specific digitalised data environment. However, for a rural school, for instance, embedded with the various South African official languages and where local knowledge has not been digitised, ChatGPT had contributed significantly to inequality.

“Our own inability to address this, to have a digital policy, a transversal digital policy across government, possibly driven from the Presidency to ensure the levels of coordination needed between the public and private sectors, and across the public sector, to ensure that we can mobilise these technologies, deploy these technologies.”

She highlighted that there were data scientists at the University of Pretoria who were developing local, large language modelling, however, access to data sets to develop these languages for GIS to communicate in the constitutionally official languages of South Africa required a few "unblockages", and it could make “such a difference”.

The issues of digitalisation and datafication have been identified by the United Nations Secretary-General as one of two seismic shifts – the other being climate change – that will determine whether this century is going to result in greater inequality or greater equality.

“How we govern digitalisation will determine whether we are improving or exacerbating inequality,” she commented, noting that policies had fallen short in fostering equitable digital inclusion and more effective regulated markets and private-sector investment was required, besides others.

In addition to a need for an enabling environment, associated digital skills and capacity and digital innovation and entrepreneurship, there is a need for digital public infrastructure, including broadband and data infrastructure, as well as digital identities and online payments.

“We cannot deliver [digitalised] public services unless these are universally available.”

However, it is not only about ensuring people are digitally embedded, but also ensuring active digital engagement.

University of the Witwatersrand Link Centre director Lucienne Abrahams noted that digitalisation and digital futures were being driven by use cases.

“What is the usefulness, or the particular specific use, of digitalising anything in the public sector. Where is the social value? We understand something like mobile money, we know what the value is. What is the value of the digital future for education or health or any aspect of public service?” she questioned.

“If we are saying that each new technology leads to another digital divide – and a lot of research suggests that we now have four or more kinds of digital divide – we might add AI as a fifth or sixth digital divide.”

Abrahams pointed out that much being done with AI was substitution and enabling a very passive engagement, citing the example of education and online learning in South Africa.

“A lot of what we are doing with technology is pure substitution. Instead of having a teacher in the classroom, let us substitute that teacher with AI, or let us have a combination of AI-based content and the teacher in the classroom. But that is a very passive engagement with very advanced, very sophisticated tools.”

“The biggest problem with learning across South Africa and many parts of Africa, is that learning is passive, and passive learning is one of the reasons for the significant failure rates that we see.”

The learning experience can be augmented by making it more active.

“Augmentation, modification and redefinition really foresee much greater teacher and learner engagement, and what that requires is coming back to the point about digital multilingualism, with teachers also needing to be able to code and develop and build apps that make learning more meaningful, more contextual for learners in particular countries.”

“We need to foresee the end game, because the end game will influence the moves we make now, in the next three years, the next five years, the next 10 years,” she continued, concluding that the end game was about seeing into the future, seeing that digital future, to be able to organise better.