The ‘Plasticity’ of the Algorithm: Affective Labour, Machine Inference, and the Defence of Kibagenge in Kenya’s Creative Economy

Abstract

While the global expansion of Generative Artificial Intelligence (Gen-AI) has produced heated debates about the use of personal data and rights to authorship; the current state of knowledge on Gen-AI’s effects on ‘Affective Labour’, following Hochschild’s (1983) concept of emotional labour and Hardt’s (1999) affective labour, used here to denote the relational and emotional dimensions of creative work; globally is still at an embryonic stage. This paper explores the theoretical tensions inherent in the epistemic difference between machine-learning and human intuition through a decolonial lens to the West-centric world of Affective Computing. It examines the implications of Gen-AI for East Africa’s performative arts. We contend that Western algorithms are deficient relative to the realities of two interlocking East African relational constructs, Kibagenge, and Utu, in Kenya. Employing a thematic analysis of qualitative data from 16 Kenyan practitioners, our research highlights a large generational divide among participants in terms of how they navigate ‘Synthetic Empathy.’ Veterans identified AI's context blindness as being inherently ‘plastic’ (participants’ vernacular term for inauthentic or synthetic, distinguished below from the technical sense of algorithmic ‘plasticity’ as adaptive capacity), and ‘robotic’ and therefore deliberately retreated to unrecorded, off-line spaces to preserve the shared spirit of their labour. Moreover, younger digitally native individuals employed a hybrid strategy using Gen-AI. They utilised a ‘50/50’ practical hybrid workflow, using Gen-AI to manage the structural aspects of creating art while preserving their relational and intuitive affects. Our findings indicate that participants experienced this not as a mere technological ‘reaction delay’ but as what they described as an unbridgeable ontological void created by what we term the algorithmic ‘reaction gap’: the mismatch between a system’s inferred emotional output and the affective cues it is responding to. We report this as a finding about participants’ experience of current systems, rather than a claim about what machine inference can achieve in principle. Therefore, protecting the future of labour related to African performance will require transcending Western theories of discrete emotions and developing forms of technology governance that respect collective relational epistemologies.