Integrating nature as a stakeholder in strategic governance: A goal-driven Large Language Model approach applied to the Fanyatu Case

Details

Supervisors

Kolp, Manuel

Faculty

Louvain School of Management

Degree label

Master [60] en sciences de gestion (horaire décalé)

Abstract

enOrganizations increasingly seek to integrate ecological considerations into strategic governance. Existing approaches typically rely on predefined sustainability objectives, such as ESG targets, which assume that ecological interests can be specified in advance and evaluated against fixed criteria. However, such concerns are often partial, context dependent, and difficult to anticipate prior to a concrete decision situation. This thesis investigates an alternative perspective in which Nature is considered as a stakeholder whose concerns are articulated in response to specific governance decisions to-be taken rather than a generic strategy that has been predefined. Drawing on concepts from Requirements Engineering and goal-oriented modeling, the study explores how such a perspective can be operationalized in practice through a model-driven approach. The research is conducted as a case study within the Fanyatu reforestation initiative. Fanyatu is a Belgian non-profit organization active in the Congo Basin that needs to decide upon the continuity of its activities in the zone of the Kivu--Kinshasa Green Corridor. A board level decision regarding the representation and valuation of ecological impact in the framework of that specific region is first modeled using the i* framework. The latter allows to capture actors, alternatives, and governance-level trade-offs independently of Nature. Based on this representation, Nature's perspective is subsequently articulated using large language models under structured constraints, and the resulting concerns are analyzed and reintegrated into the model documenting the governance decision to-be taken. The findings show that this instantiation makes explicit several resulting ecology tensions that were not fully visible in the initial deliberation. These include tensions between standardized reporting and ecosystem complexity, short-term measurable indicators and long-term ecological processes, and external accountability and community-based stewardship. Rather than prescribing decisions, the approach contributes to structuring deliberation by rendering implicit trade-offs explicit and traceable. The decision thus benefits from an increased documentation giving Nature the opportunity to advice board level decisions. Eventually, the thesis demonstrates how a model-driven approach can support the integration of non-human stakeholder perspectives in strategic governance while preserving human decision authority. While the study is conducted in a reforestation context, the approach may be applicable to other governance settings involving stakeholders whose interests cannot be fully specified in advance.