In 2019, APC and ARTICLE 19 partnered to publish a groundbreaking edition of Global Information Society Watch (GISWatch) exploring emerging issues around artificial intelligence (AI) and how algorithm-based decision-making impacts the Global South. Rooted in human rights, social justice and development, this edition brought together 40 country reports, three regional reports and eight thematic reports from contributors around the globe, influencing policy makers and human rights defenders from the grassroots level to key governance spaces.
As AI was gaining widespread global traction, with many unquestioningly touting it as the way of the future, many civil society organisations were advocating for a more measured and informed approach, even if it meant taking time for more thorough analysis. Seven years since the launch of the GISWatch edition on AI, what changes have we seen? How have some contexts improved and others worsened? What are the recommendations that authors called for that still need our urgent attention today, perhaps more than ever?
Moving the conversation forward, in April this year, APC, GenderIT.org and Global Voices launched a new series on AI, questioning what could be done to create a human rights approach to AI. Grounded in lived realities and backed by evidence, authors from different countries and backgrounds have provided a very practical demonstration of the ongoing violations in this field, while pointing the way towards curbing them. A call for accountability runs through them all.
It is in this spirit that we revisit the 2019 edition of GISWatch on “Artificial intelligence: Human rights, social justice and development” with a view to renewing our commitment to addressing these important issues and exploring what concrete steps we can take for a more equitable and rights-based digital future. We highlight a selection of calls to action that are still relevant and where civil society has a critical role to play.
In doing so, we hope to revive our activism through the research conducted by authors on the ground, in spaces where important discussions are taking place or perhaps have yet to manifest. At the same time we observe a hope for the positive potential of AI to serve humanity, but only when done through meaningful engagement, transparency and intentionality.
Exploring AI across the APC network in 2019
In its country report on AI in education in India, Digital Empowerment Foundation (DEF) examines how AI data over-represents biased socioeconomic profiles based on caste, gender and religion, meaning marginalised populations are neglected by statistical models. DEF notes, “given that AI systems are increasingly aiding state institutions in the allocation of resources, it becomes imperative that they align with principles of non-discrimination rather than perpetuate existing misallocations by creating pervasive systems of privilege being trained on unrepresentative data sets and models.”
To that end, one straightforward and practical action step proposed by the organisation focuses on representative data choices, specifically that “the data on which a machine-learning system is trained must be representative of the population in which it is to be deployed.” While it may not eradicate discrimination bias entirely, having this kind of representation in machine learning is a practical step toward circumventing the prioritisation of certain voices over more marginalised ones.
In Costa Rica, Cooperativa Sulá Batsú looked at AI applications in health care and the impact on the country’s public universal health care system. The technical platform that contains a broad range of healthcare data within the country, including patients’ medical histories, is seen as an important public management strategy and a public good.
However, despite the importance of this system for healthcare services, Sulá Batsú states that “it is necessary to incorporate key actors and citizen organisations in the discussion about the advantages and challenges of incorporating AI into public health.” They furthermore affirm, “It is an extremely important issue for the exercise of the right to health that should not be treated only as a technological development.”
PROTEGE QV from Cameroon also looked at the integration of AI in the national healthcare system. In this case, however, it notes that “while artificial intelligence offers the potential to enhance health care throughout the country and to overcome an acute shortage of medical personnel in rural areas, Cameroon has been slow in its adoption.”
To address these shortfalls, one clear action step calls for government to meet its promised targets, specifically that “Cameroon should increase the health budget so that it meets the Abuja Declaration requirement of 15% being allocated yearly to the health sector.”
The labour sector is also heavily impacted by hastily implemented AI policies and practices. As a case in point, Bangladesh is one of the top garment exporters in the world, and the vast numbers of employees who work at the bottom of these production processes are society’s most vulnerable. This disproportionately includes women and girls, according to Bytes for All Bangladesh. “Artificial intelligence can enhance productivity through automation, machine learning and precision,” the authors explain. “But to automate tasks through an intelligent system that can reduce time, costs and the need for labour also has the potential to diminish jobs."
How to improve industry efficiency and output while protecting the jobs of those living in economic precarity? One key action step is to recognise that “just as sewing machines break and need calibration, so, too, will the apparel printers and packaging systems of the future,” the organisation states. “To help ease the transition from manual to modern manufacturing, businesses and governments must begin improving the tech literacy of current employees.”
The question of how AI will recalibrate working conditions was also a topic of consideration for Nodo TAU in Argentina, who explored how unions are addressing this shifting reality across the country. One of the big challenges was finding a consistent definition of AI, a task described as “a kind of moving target”. Like Bytes for All Bangladesh, Noda TAU also recommends that it is vital to “promote the debate around training and reskilling.” To that end, they advise that “unions should be consulted regularly about skills and the government should promote training in AI technologies in the education system.”
Reviewing our lessons while looking ahead
Looking across these three key sectors – education, healthcare and labour – we can see that there are dialogues that are still urgently needed concerning the adoption of AI practices and platforms. Advocacy around more responsible AI policies by civil society is critical in order to push states and corporations toward accountability. Understanding the lived realities on the ground is integral in developing laws, policies and treaties that are rights-respecting.
This means also taking into consideration who has a seat at the table where decisions are being made. “Despite the application of AI in geographically diverse contexts, conversations on AI have been driven largely by Western and Global North predictions and perspectives,” write APC and ARTICLE 19 in the preface to the 2019 edition. “Yet the assumptions, values, incentives and socioeconomic environments within which AI technologies function vary greatly across jurisdictions and, as a consequence, the very real effects of AI are also more diverse.”
As the pace of technology intensifies, we call for a reexamination of lessons learned to inform our action steps moving forward.
Discover the full edition of GISWatch 2019 on “Artificial intelligence: Human rights, social justice and development.” This edition can also be downloaded online in PDF, MOBI and ePUB formats.
