My doctoral research, conducted at the Université Polytechnique Hauts-de-France (DeVisu Laboratory, LARSH), examines how algorithmic systems reconfigure organisational communication, knowledge production, and the distribution of power within digital environments. It is anchored in Information and Communication Sciences (ICS), a discipline that treats technologies not as neutral instruments but as socio-technical configurations embedded in social logics, power relations, and organisational rationalities.
Core Research Question
The central question driving this research is: how AI reconfigures who holds tacit knowledge, whose knowledge is recognised, and how it circulates within organisations?
Artificial intelligence is often framed as a managerial optimisation tool. My work reframes this: AI is not merely a tool for managing knowledge – it is an epistemological dispositif that restructures what counts as knowledge, who holds it, and how it circulates. As a dispositif, AI operates as a pharmakon: the same system that can sustain coordination and shared understanding can also dispossess workers of their knowledge and narrow what may legitimately be said. When organisations deploy algorithmic systems, they do not simply automate tasks. They redistribute cognitive capital, reconfigure communicational authority, and alter the conditions under which collective meaning-making becomes possible.
Theoretical Framework
The research mobilises a multi-layered theoretical architecture that reflects the interdisciplinary character of ICS.
Power, Knowledge and Algorithmic Mediation
The research analyses how algorithmic systems function as socio-technical dispositifs that reconfigure the circulation and legitimation of knowledge within organisations. Drawing on Jeanneret and Souchier’s notion of architextual regimes, it argues that these systems do not simply process existing knowledge: they pre-structure what can be expressed and recognised, generating new categories of the knowable and the unknowable and reshaping what organisations treat as legitimate expertise (Cardon, 2015; Floris, 2001).
Tacit Knowledge and Organisational Epistemology
Nonaka and Takeuchi’s account of the conversion between tacit and explicit knowledge provides the analytical point of departure – but this research runs counter to its programme rather than extending it. Where their model treats the tacit as a resource to be elicited, formalised, and capitalised, the argument here is that algorithmic mediation does not convert tacit knowledge so much as reduce it: it produces impoverished formalisations that then harden into organisational norms. In info-communicational terms, this is an architextual operation — the formalisation pre-structures what can subsequently be expressed and recognised as legitimate knowledge (Jeanneret and Souchier). Workers thus experience what I describe as a double dispossession: the loss of their knowledge monopoly, and the simultaneous delegitimisation of the tacit knowledge that resists formalisation.
Ubuntu Ethics as an Analytical Framework
Ubuntu – the African philosophical principle that a person is a person through other persons – is not a cultural add-on in this research. In this research Ubuntu is mobilised in three distinct but articulated statuses. First, as a relational ontology (Ramose): being is co-constituted through others, which grounds an ethical and normative critique of systems organised solely around efficiency. Second, as an operational framework — « Ubuntu by Design » — in which Mbigi’s four interdependent dimensions (Survival, Solidarity, Compassion, Respect, and Dignity) become evaluative criteria for the relational and ethical qualities of AI-mediated communication, this register tends toward a dispositif. Third, and most distinctively, as a reflexive and transformative process that traverses existing dispositifs and reconfigures them from within, rather than standing beside them as an external moral supplement. Taken together, these three statuses differentiate the approach from mainstream AI-governance discourse, which typically treats ethics as a compliance layer external to the system.
Our recent research has demonstrated that Ubuntu reveals what standard AI evaluation metrics (BLEU, ROUGE, Perplexity) cannot capture: the structural marginalisation of collective interdependence – particularly Solidarity – in algorithmically mediated discourse.
Original Frameworks
This research has developed two original methodological tools for assessing GenAI-assisted communication.
STARA (Situation, Task, Action, Result, Self-Assessment) is a structured prompt model designed to generate comparable simulated interview outputs across different LLMs. It serves as generative scaffolding for road-testing interview guides and eliciting meta-level self-assessment from AI systems.
STRING is a post-interaction analytical framework that combines qualitative interpretation with quantitative scoring to evaluate three dimensions of response quality: coherence, plausibility, and relevance. It integrates conversational analysis with an Ubuntu-informed ethical lens and visual mapping of tensions between values and organisational dimensions.
Together, STARA and STRING offer ICS researchers a structured approach to assessing GenAI-assisted interviews – one that complements technical metrics with the kind of ethical reflexivity those metrics cannot capture.
Current Research Directions
Communicational regimes of LLMs. A comparative analysis of ChatGPT, Gemini, and Claude has revealed that each model produces a distinct communicational regime: ChatGPT stabilises meaning through logical closure, Gemini structures through systemic abstraction, Claude opens through reflexivity and ethical ambivalence. These are not merely stylistic differences – they reflect differentiated socio-technical mediation architectures.
Worker resistance and counter-conducts. An emerging direction examines how workers respond to AI-driven knowledge extraction: through knowledge retention, deliberate misinformation to AI systems, and the creation of new tacit knowledge domains post-implementation. These strategies are analysed as Foucauldian counter-conducts within algorithmic organisations.
Ubuntu as organisational design principle. Beyond its analytical function, Ubuntu is being explored as a framework for navigating organisational tensions between technological rationalisation and relational ethics – what I call Ubuntu by Design.
Institutional Affiliation
Laboratory: DeVisu – Laboratoire en Design Visuel et Urbain (LARSH)
Institution: Université Polytechnique Hauts-de-France (UPHF)
Discipline: Information and Communication Sciences (ICS)
AI does not just automate knowledge. It reconfigures who knows, what counts as knowing, and whose knowledge matters.