Mandiaye Diallo (Sénégal): Biais Algorithmiques Menacent Souveraineté Afrique

• Mandiaye Diallo, a Senegalese digital expert, warns that artificial intelligence systems can amplify racial prejudices, leading to algorithmic discrimination affecting millions.
• Experiments showed AI tools providing different responses based on perceived origin or skin color, with some leading to security concerns for individuals identified as Algerian or Black.
• Africa, despite holding 18% of the global population, possesses less than 1% of worldwide data center capacity, contributing to its digital dependence.
• Diallo advocates for African countries to develop their own AI models, research centers, and databases that integrate local languages, histories, and cultures.
• The expert stresses the critical need for Africa to build technical capacity and maintain control over the human choices encoded within AI technologies to ensure digital sovereignty.

Artificial intelligence tools are not neutral or autonomous; they reflect human choices, from data selection to defining objectives and evaluation criteria, and can perpetuate existing societal power dynamics.

Mandiaye Diallo, a prominent digital expert from Senegal, has issued a stark warning regarding the pervasive threat of algorithmic discrimination, emphasizing its potential to impact millions within mere seconds. He highlights how artificial intelligence systems are not only capable of reproducing but also significantly amplifying existing racial prejudices. Diallo substantiated his concerns through experiments conducted with several well-known AI systems, which revealed disparate responses based on the perceived origin or skin color of individuals mentioned in prompts. For instance, scenarios involving an Algerian or a Black person sometimes triggered security alerts and even suggestions to contact law enforcement, whereas a similar situation referencing an Italian individual typically elicited advice related to social interaction and exchange.

The implications of such biases extend far beyond simple chatbot interactions. Artificial intelligence is increasingly integrated into critical decision-making processes across vital sectors, including recruitment, justice, healthcare, security, and information dissemination. Consequently, the inherent representations embedded within the data used to design and train these AI models could directly influence the life trajectories and fundamental rights of vast populations. Diallo underscores that these advanced tools are neither neutral nor autonomous; rather, they are direct reflections of human choices, from the initial selection of data to the definition of objectives and evaluation criteria. What might appear as a purely technical decision can, in reality, serve to perpetuate existing power dynamics and societal prejudices.

This urgent call to address algorithmic bias is intrinsically linked to a broader discourse concerning Africa's digital sovereignty. Despite accounting for 18% of the global population, the African continent possesses less than one percent of the world's total data center capacity. Furthermore, a significant concentration exists, with over two-thirds of the 223 identified data centers across Africa located in South Africa. This substantial material dependence on external infrastructure exacerbates a deeper reliance on the data, computational models, and cultural references that are integrated into the digital tools utilized daily across the continent.

Mandiaye Diallo, also the author of "La Quatrième Blessure – Pourquoi l’Afrique doit gagner la guerre numérique," articulates how this technological dependence extends beyond mere infrastructure and data storage to encompass the very cultural frameworks embedded within digital technologies. With computing and hosting capabilities distributed so unevenly, Africa faces the persistent risk of remaining primarily a consumer of systems developed elsewhere. This consumption model inherently limits the continent's ability to fully control the operational conditions of these technologies or to mitigate the potential consequences of their embedded decisions, thereby reinforcing existing biases and hindering genuine digital autonomy.

To counter this growing challenge, Mandiaye Diallo strongly advocates for African nations to transition from being passive users to active shapers of digital innovation. He proposes a strategic imperative for the continent to develop its own artificial intelligence models, establish dedicated research centers, and build comprehensive databases that authentically incorporate African languages, historical narratives, diverse cultures, and unique realities. This vision entails fostering an environment where AI can truly "speak" African languages and be "nourished" by the continent's rich imaginaries.

However, Diallo also issues a critical warning: even if machines eventually master African languages, there remains a significant risk that they could simultaneously reproduce existing racist biases if not developed with intentional, culturally sensitive frameworks. The fundamental challenge for Africa, therefore, is two-pronged: it involves not only developing robust technical capabilities and infrastructure but also, crucially, maintaining firm control over the human choices and ethical considerations that are encoded within these advanced technologies. This dual focus is essential for securing genuine digital sovereignty and ensuring that AI serves the continent's best interests.

Lawyers and compliance officers must be aware of the significant legal and ethical risks posed by algorithmic bias, particularly in critical sectors like recruitment, justice, and healthcare. This includes potential discrimination claims and the need for robust due diligence on AI systems to ensure fairness, transparency, and compliance with evolving ethical AI standards, especially given the call for African-centric AI development.

Source: Reporting based on an article from APS.