Enhancing Impact Measurement of Philanthropic Organisations: A Human-AI Collaboration Framework

Abstract

Philanthropic organisations increasingly face pressure to demonstrate the impact of their work, yet existing impact measurement practices remain fragmented, resource-intensive, and often ill-suited to capturing both qualitative and quantitative outcomes. This thesis addresses these challenges by proposing a human–AI collaboration framework designed to enhance the efficiency, traceability, and usefulness of impact data in the nonprofit sector. Building on principles of Design Science Research (DSR), the study integrates semantic technologies (ontology and knowledge graphs), natural language processing (NLP), and automation tools within a prototype system aimed at structuring and querying unstructured impact data. The research is informed by a two-phase empirical process: initial exploratory interviews to identify key challenges and requirements, followed by evaluative interviews assessing the system’s perceived usefulness, usability, and ethical acceptability. The results confirm the relevance of established models such as the Technology Acceptance Model (TAM) and Human-Centered AI (HCAI) in this context, highlighting the importance of transparency, trust, and organisational fit. The proposed framework was found to effectively support common impact measurement needs, such as aggregating indicators, linking data to strategic goals like the SDGs, and making qualitative insights more analysable. This work contributes both a functional prototype and a set of design recommendations for responsible AI implementation in the social sector. It also responds to documented gaps in the literature regarding integrated, context-sensitive AI tools for nonprofits. The findings underscore the potential of AI to support evidence-based decision-making in philanthropy, provided that technical innovations are embedded within participatory, ethical, and user-centred processes.