Philip Alston, the UN rapporteur on extreme poverty, warned yesterday the members at the UN General Assembly about the dangers to which human rights are subject with the emergence of ’digital welfare states’. He went further to highlight the burden of the externalities levied on society as a result of the prevailing logic of the market in digitalisation, which disregards human rights. He, too, warned of the diversity crisis in the AI sector, pointing at it as a key driver of discrimination based on gender and race. The externalities of these systems are affecting the poor and the vulnerable more than those whom are better off. As these technologies are deployed globally, it is reasonable to think that those loosing out the most are those living in states where the law and human rights are most fragile.
Payal Arora, professor at the Erasmus University Rotterdam, documented in 2010 the implementation of a medical diagnostic in the Himalayas. Her work showed how the skewed development of the software resulted in the majority of the illnesses reported by villagers not finding any appropriate categories in the software. These illnesses ended up being categorised as ’other’. The software’s deficit in terms of socioeconomic and cultural contextualisation led to a non-representative value system for diagnostic.
Examples abound about the inaccuracy, inappropriateness, and discriminatory outputs of data enabled processes. Despite this, the hype of datatisation in development goes on. Philip Alston has called for effective government regulation, based on the respect for human rights, to counter this reality. The question here is if regulation in itself will suffice for everyone in the absence of thorough analysis of other factors deepening the various departure points for the different countries. Such an analysis would include at least:
- Considerations about the availability of data and the representativeness of data collection processes. The availability of historical data, which can be used to populate big data sets, in the Global South is limited. Current processes for data collection most often rely on internet usage based processes. The After Access 2018 report published by Research ICT Africa, found out that South Africa was the only country in sub-Saharan Africa with an internet penetration rate over 50% while the norm for other countries was 10-30%. This would mean that as of last year the data collected in some countries in the African continent does not represent between 70-90% of its population, and the population represented is not likely to be poor and vulnerable whom often lack the resources and literacy for access. These unevenness in representation in datasets is likely to affect both, datasets representation of the Global North over the Global South, and the representation of those better off over those worse off in the datasets. This raises the question of how data-enabled development processes for decision making and prioritisation can deliver appropriate outcomes for the poor and vulnerable if these are not represented in the datasets. Are we sure we will not be ’leaving anyone behind’?
- Considerations, as flagged by Alston, of how the values and life experiences of those shaping algorithmic data processing are embedded in those processes and often leading to discrimination and exclusion.
- Considerations of the different approaches to data by the different governments in the Global South. The differences in how China, India or Brazil have approached digitalisation and its objectives, from surveillance to service delivery, is representative of the heterogeneity in the Global South.
- Considerations of the unevenness of existing capacities to generate, access, purchase, store and process large data sets. If considered, the current state of affairs tilts the balance towards the Global North, having those capacities, a vast majority of the population residing in the Global South, whose data is being collected, lacking those capacities. This asymmetric relation between those who have it and those who do not is what Andrejevic has called the big data divide. This big data divide, Andrejevic argues is likely to contribute to exacerbating power imbalances in the digital era. While Andrejevic’s focus is on the individual, similar imbalance can be imagined at the macro level when thinking about data and development. On the one hand, there will be those holding the data and setting the agenda, and on the other hand those whose data is being captured and thereafter told what to prioritise. How will this power imbalance affect the Global South’s governments capacities to regulate if they cannot set the agenda of what machines are looking for or how they do it?
- Considerations of how our current knowledge on the impact of data has been formed and where it is coming from. In this regard, the Big Data from the South Initiative, argue that most of this knowledge has been generated in the Global North, with its interests and concerns as drivers. Stefania Milan and Emiliano Treré thus call for a ’de-Westernisation of critical data studies’. This epistemological shift would allow for re-starting the discussion about data and development taking into account the variety of understandings, experiences and responses to datatisation processes arising from the Global South. This would also allow a de-colonial reading of datatisation.
There is no doubt regulation is needed to address the negative impacts that the use of data and datatisation processes are having on those less well off. However, it is reasonable to believe that given the power imbalances in knowledge and representation and the universalist tendencies of these datatisation processes, regulation might not be enough. The structural causes behind the current failures need to be addressed. To succeed, it is crucial to recognise the variety of departure points and heterogeneity of contexts and practices in the Global South, so these can meaningfully influence the datatisation debate and contribute to reshaping its future development.