Listening at scale: AI-assisted causal mapping for inclusive evaluation
Revista Brasileira de Avaliação · 2026
DOI: 10.4322/rbaval.202600222026 ↗Kurz gefasst
Fachartikel in der Zeitschrift Revista Brasileira de Avaliação, 2026. Verfasst von Gabriele Caldas Cabral und 1 weiteren Autor:innen. Die Arbeit nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Evaluation practice often faces a trade-off between depth and metrics, frequently resulting in the marginalisation of local voices. Traditional methods of coding texts, such as interviews, risk filtering diverse narratives through pre-defined, donor-driven frameworks, perpetuating a form of epistemic violence that prioritises ‘data’ over ‘story’. We propose an alternative: AI-assisted causal mapping as a method for ‘listening at scale’ using a zero-codebook approach where Generative AI serves as a ‘low-level’ assistant to extract causal links from large volumes of unstructured text without an imposed codebook. A case study of Tree Aid in Burkina Faso demonstrates how this method reveals complex outcomes which might be missed by models that use a pre-determined codebook. We introduce ‘soft recoding’ as a technique to align local narratives with donors’ conceptual frameworks without erasing original data. We conclude by reflecting on the tensions between bottom-up storytelling and top-down accountability, arguing that while AI has its own biases, rigorous human oversight can use it to amplify rather than silence marginalised voices.
Zitieren
APA
Cabral, G. C., & Powell, S. (2026). Listening at scale: AI-assisted causal mapping for inclusive evaluation. Revista Brasileira de Avaliação. https://doi.org/10.4322/rbaval.202600222026
Vancouver
Cabral GC, Powell S. Listening at scale: AI-assisted causal mapping for inclusive evaluation. Revista Brasileira de Avaliação. 2026. doi:10.4322/rbaval.202600222026
BibTeX
@article{Cabral2026,
author = {Cabral, Gabriele Caldas and Powell, Steve},
title = {Listening at scale: AI-assisted causal mapping for inclusive evaluation},
journal = {Revista Brasileira de Avaliação},
year = {2026},
doi = {10.4322/rbaval.202600222026}
}
EUVIMED