| dc.description.abstract | Contemporary scholars and civil organizations increasingly associate Artificial intelligence (AI) with colo-nialism. How should these normative comparisons be interpreted? This thesis conducts a philosophical investigation into three different ways of understanding the indictments of ‘AI colonialism’. First, AI systems or their development and deployment could literally instantiate the injustices constitutive of co-lonialism. Second, AI colonialism could be a metaphor that expresses a similarity between AI and colonialism. Third, it could be a case of conceptual engineering, prompting either the amelioration of the existing concept of colonialism or the expansion of the lexicon with the term ‘AI colonialism’. In order to evaluate which interpretation is correct, an operational definition of ‘colonialism’ is pursued. The linguistically normative core of the definition established here is the ‘institutional dehumanizing subversion of self-determination’, while its descriptive periphery is found to consist in exploitation, the taking of land, cultural imposition and violence. Subsequently, the contours of AI colonialism are investigated. Based upon a technical understanding of machine learning algorithms, six grounds for associating AI with colonialism are found in the literature: extractivism, exploitation, objectification, cultural imposition, epistemic violence and racialization. Based on these grounds and colonialism’s operational definition, it is argued that the literal interpretation is invalid because AI systems cannot literally colonize and AI development and deployment doesn’t instantiate colonialism’s linguistically normative core. The thesis’ corresponding positive argument is threefold. First, it is argued that AI colonialism is a conceptual metaphor that doesn’t warrant colonialism’s conceptual amelioration, but is a candidate for lexical expansion. Second, it is argued that AI colonialism, as conceptual metaphor, expresses insightful interrelations between AI and colonialism’s contingent properties: its projection ‘stretches and twists’ colonialism to disclose a mean-ingful coherence in AI injustices. This claim stands in between the metaphorical and conceptual engineering interpretations because it is currently a metaphor but could develop into a full neologism over time. Finally, the political metaphor that is AI colonialism is found to be epistemically successful because it is inference preserving and invites taking a theoretically fruitful political perspective. | |
