AI and a Politics of Truthfulness and Compassion

The wealthiest one per cent controls more wealth than the majority of humanity. A politics of intelligence, rooted in truthfulness, algorithmic fairness, and compassion is urgent. Our own Aimé Césaire from Martinique reminds us: “There is not in the world one single poor lynched bastard, one poor tortured man, in whom I am not also murdered and humiliated.”

In a handful of countries, the human rights ecosystem is vilified, shunned, and distorted. Artificial Intelligence (AI) brings new speed and scale to these terrorisations. In our modern world, we must invoke a critical human rights-centred criminology, as civilisations can no longer accept the normative claims of states at face value.

We must seek to uncover every practice that undermines the human rights of workers and intends to reinforce inherited inequalities and intergenerational immobility.  Social tensions increase as injustices breed hostility, often targeting the most vulnerable. The human rights edifice, like the cast iron frame of the Minor Basilica of Our Lady of the Immaculate Conception in Castries, raised painstakingly over decades, is stressed. The past is always in front of us. Never behind. Every decision we make today – whether it nurtures impunity or justice – will shape our digital futures.

Buried deep beneath the spectre of AI and its lofty accomplishments are the Data Labelling Factories cramped with annotation workers. These invisible workers are the élan vital of the burgeoning AI industry, which is expected to be worth $407 billion by 2027. Data labelling is the process of annotating raw data like images, short videos, and text to allow AI systems to recognise patterns and make predictions.

Data labellers work in dusty, overcrowded environments that pose a serious threat to their well-being. Once deployed, large-scale AI models require continuous investment in data labelling, refinement, and real-world testing. These piece-workers are paid an hourly rate that ranges from US$0.90 to US$2 in some jurisdictions. In other places, the rate is between US$10 and $25 per hour.  The mental strain of data labelling is significant, with repetitive tasks required to meet strict deadlines and stringent quality controls. Data labellers who have repeatedly reviewed hate speech, egregious acts of human violence, and abusive language have suffered psychological trauma.

While many notable efforts focus on ethical AI and AI for social good, the work of data labellers that underpins the AI industry requires equal attention.  Annotation workers are denied access to performance data, and this lack of transparency impedes their ability to improve or contest job losses and pay cuts.

Data labellers also work as independent contractors, lacking access to healthcare and benefits. It was not until the UK Supreme Court affirmed that Uber drivers are “workers” and not independent contractors, hence entitled to employment rights like minimum wage, holiday pay, and rest breaks after logging into the app, that the status of these workers changed.  One of the positive outcomes of AI is its potential to identify supply chains, regions, and specific factories that may be abusing their workers by manufacturing products through forced labour.

AI can assist retailers and consumers in avoiding these manufacturers, as it is illegal to import these products into certain countries; moreover, most people and companies with a conscience would not wish to contribute to human rights violations.  The ILO defines forced labour as contexts in which individuals are coerced into working through the abuse of vulnerability, isolation, deception, intimidation, threats, violence, excessive overtime, abusive living conditions, withholding of wages, debt bondage, retention of identity documents, and restriction of movement.

In our modern world, approximately 50 million people are living in modern-day slavery, according to data from the International Labour Organization (ILO). The scope of forced labour within global supply chains necessitates that all stakeholders assist in alleviating human suffering and safeguarding human rights.  As AI supply chains proliferate in the Global South in pursuit of maximising profits, the need for ethical AI supply chains is growing increasingly urgent.

One approach to intervene is to apply a human rights-centred design and oversight strategy to the entire AI supply chain. In the US, trade laws such as Section 307 of the 1930 Tariff Act and the more recent Uyghur Forced Labour Prevention Act are designed to prevent products coming into US ports that have been extracted, fashioned, or mass-produced wholly or in part through forced labour.

However, when the product, the data, can be transmitted via electronic mail, the policies and enforcement actions currently in place may prove inadequate. These laws were initially intended to govern tangible goods rather than intangible goods that do not use traditional ports of entry.

States must also draft regulations mandating that these practices support a Rawlsian sense of Justice as Fairness, ensuring transparency in performance evaluation and personal data processing. This will enable annotation workers to understand how they are assessed and contest potential inaccuracies. Governments must also promote the formation of Digital Labour Unions or cooperatives to give piece workers a voice. Just as we now reward green and fair trade producers of goods, we must advocate for change by choosing digital services that adhere to human rights standards, ethical AI brands, and a politics of truthfulness and compassion.