ICTworks

“Political economy” has two senses. It can refer to a particular political economy of a country, industry or technology. It also refers to the kind of analysis that identifies the dynamics of political economy.

While political economy theory has its roots in classical economic theorists like David Ricardo and Karl Marx, its popularity in this second sense has waxed and waned in academic and policy research. The field certainly made a resurgence after the 2008 Great Recession as people sought to better understand the role of economic power and hierarchies of influence in bringing about that crisis in capitalism.

Now with appropriate concern about the role of artificial intelligence in re-shaping power relations between states, markets and citizens; the future of work and the labour process, as well as how AI might restructure firms and the provision of government services; applying a political economy lens can help policy researchers think more clearly about the interplay between power, classes, and technological change.

Through coming to fully perceive how the various statistical rules and weights of AI have political ramifications, the paper, The Political Economy of African AI: A Primer on Concepts, Contexts, Considerations and Capitalism, can help them take steps to avert another capitalist crisis of similar or greater magnitude. People usually understand political economy as a theoretical and applied framework that guides attention to several key areas.

This helps to better grasp the sum of social relations and their meanings. There are many schools of thought in the field, and they have an intensive, healthy intramural debate. Nevertheless, these schools tend to be unified around a concern of what existing social relations mean for the prospect of substantive social change.

7 Political Economy Aspects of AI

Typically, the key objective of a political economic analysis is to use the history of material organisation to explain relations, processes, institutions, and organisations. These seven ideas below can help improve public deliberation around AI and inequalities, in part by avoiding common intellectual cul-de-sacs that bring about platitudes.

One challenge in Africa is to realise that AI systems are not politically neutral or a panacea for growth and development; AI’s statistical rules and weights are situated judgements from those that work in Big Tech firms beholden to shareholder primacy.

Historically State-commissioned transnational companies in imperial centres were powerful, like the Dutch East India Company, which carried out colonial trade. Big Tech surpasses them. Though it arguably represents a new mode of production with specific and new implications for economy and society, AI is not a break with history.

There are strong continuities in this phase of advanced capitalism. The field of political economy shows how AI systems are contestable, in part because they are also social systems. This idea alone can bring energy for currently subordinated groups to pursue projects of negotiation, bargaining and struggle for a fairer political economy in which to live and work.

With this agenda, here are seven ideas to think about with respect to the political economy of AI.

1. Economic orders are constructed

First, political economy points to how economic orders are constructed by different groups within that order; how these groups negotiate, bargain and struggle against one another; and what resources they gather to defend or advance their particular projects and interests. AI-powered products are now among those resources.

Also, by being attentive to the role of wealth, how it is deployed in the market and how it is acquired, people can understand how the commoditisation of AI and the privatisation of data will mean that some groups will have dramatically more resources than others. Differentials like this beget social inequality.

2. Politics is about high stakes

Second, the field of political economy is open and honest about the ruthlessness of politics and the overriding power of economic interests. Adopting AI to automate parts of the labour process benefits some groups to the detriment of others, for instance.

So, it is a misleading to claim that AI can automatically create a better life for all. With so much ‘AI hype’ in public discourse pushing glowing one dimensional narratives (whether in marketing rhetoric or nationalistic proclamations), a political economy lens provides useful counterpoints that can clarify and elevate discussion by introducing new topics to public affairs.

3. The uneven geography of AI

Keeping social inequality in mind, our third point concerns how AI is implemented unevenly in different countries across the world. The adoption of AI is not a single process: there are multiple starting points, and points of disconnection and delay due to many factors including9 infrastructure, skills, resources, needs and interests.

It is even differently experienced by people across the world who seek different things, and approach AI from different cultural and political contexts.

This means that when we talk about AI, both from a capacity to implement and from an effects perspective, we are not necessarily all talking about the same thing. With most colonised spaces subject to unequal exchange in the 19th and 20th Centuries, and without diminishing the energies of post-independence governments to rectify that problem, unequal development remains a distinctive feature of our world system.

It therefore remains a distinctive feature when considering the implementation of AI.

4. Who is deemed to have expertise?

Fourth, the political economy of AI can help us see which actors are deemed to have appropriate expertise. “Faced with disorienting technological change,” Seth Lazar and Alondra Nelson recently wrote, “people instinctively turn to technologists for solutions.”

However, they add that “the impacts of advanced AI cannot be mitigated through technical means alone; solutions that do not include broader societal insight will only compound AI’s dangers”. Due to patterns of prejudice, bigotry, and the North-South world system, some groups’ voices will not be listened to when it comes to AI dangers.

Conversely, powerful shareholders are reluctant to disclose information about the inner workings of their AI products, products we already know can and do perpetuate racism and other inequalities. The consequences of adopting AI are too great to leave this matter with executives in New York and technologists in San Francisco.

Due consideration and lawmaking by democratically elected representatives can channel AI enterprises to address human needs.

5. Hidden artificial intelligence labour

AI models are developed using human labour – in the case of popular systems substantial human labour – but this is largely hidden from public view. AI systems are built on technologies reliant on rare earth metals, many of which are extracted by miners working in inhumane and dangerous conditions.

Generative AI is trained using the creative work of thousands of artists, writers or software creators, who have not consented to the use of their work. Because AI systems are trained on data generated by humans they may produce racist, sexist, explicit and abusive outputs.

To prevent this, Reinforcement Learning from Human Feedback (RLHF) involves humans identifying undesirable outputs to feed back into the system so that it does not produce them. RLHF is labour intensive and routinely outsourced to undercompensated workers in developing countries who suffer psychological trauma from exposure to undesirable content.

Concealment of human labour in technologies that are proclaimed to be carried out by machines is termed ‘fauxtomation’.

6. Experiences of profound technological change

Sixth, the political economy of AI can remind citizens about prior previous experiences of profound technological change. As with other great industrial transformations which shook up whole social orders, AI will have reactionary and revolutionary components as people try to make sense of and adapt to changing circumstances.

Advocates of ‘the status quo but more efficient’ may not have the imaginative capacities to anticipate the potential and pitfalls that AI introduces and could be swept aside as forces overtake them. Undoubtedly AI will lead to a “recalibration of the burdens of risk between capital and labour”; some people may turn to AI to strictly enforce hierarchy, stratification and mobility, while others think about liberation.

Put differently, the politics of AI will likely intensify in the coming decades as the stakes become more acute.10

7. Identifying major changes due to AI

Lastly, we endorse Henry Bernstein’s methodology for analysing the dynamics of social relations, processes, and change. He focuses on four aspects:

• Property rights over productive resources;
• Division of labour with forms of cooperation and conflict between and among different groups;
• Distribution of income and wealth and the mechanisms for extraction and accumulation;
• Patterns of consumption and investment along with their associated meanings.

In a succinct manner, Bernstein directs researchers to ask four questions:

1. Who owns what?
2. Who does what?
3. Who gets what?
4. What do they do with it?

A lightly edited synopsis of The Political Economy of African AI: A Primer on Concepts, Contexts, Considerations and Capitalism by by Scott Timcke, Sandra Makumbirofa, Andrew Rens, Nawal Omar and Hanani Hlomani for Research ICT Africa.