The exploitation of natural resources has become an essential financial lifeline for non-state armed groups in current conflicts. In Colombia, paramilitary forces, left-wing guerrilla groups and criminal organisations have increasingly combined illegal activities around drug trafficking with extortion and exploitation of mineral resources. In the unlicensed extraction of gold, children perform multiple forms of labour, including armed and unarmed roles that benefit armed organisations. However, academic attention and institutional assistance have been mainly directed to child combatants. The widespread stereotyping of children used to participate in armed conflict as manipulated armed fighters oversimplifies the complex realities of young people during natural resource conflicts and overlooks other equally vulnerable children. This ethnographic study focused on three mining-and-conflict-affected areas in Choco, from October 2016 to June 2017, seeking to understand the lived experiences of working children and their possibilities for action and survival. It combined participatory observation, interviews and a collection of methods based on storytelling and imagination to better engage children in the research process. Besides giving particular value to local perceptions and children's voices, this methodology offers an original approach to conduct research with young people in conflictive settings. In order to understand the complex nature of childhood experiences in the research areas, I connect the literatures on the sociology of childhood, anthropology of conflict, and humanitarianism as guiding analytical frameworks. This enabled me to uncover multiple childhood experiences, as well as identify shortcomings in theoretical and practical responses. This thesis found that, in contrast to mainstream conceptualisations of child labour in extractive economies, young people in the research areas do not perform a set of fixed, observable roles. Rather, they adapt their networks of interaction, identities and meanings ascribed to childhood in their quest for survival. The complexity of their experience navigating combined forms of labour problematizes existing categorisations, the division between armed and unarmed roles, and the labels use to acknowledge their experiences. In addition to methodological strategies, this thesis makes three significant contributions to theoretical debates around childhood, child labour, and humanitarianism. Firstly, I propose a networked interpretation of child labour that embraces the fluidity and 'messiness' of natural resource conflicts. Secondly, I expand the conceptualisations of children's agency and social age by including the relationships of working children with illegal actors. Finally, I challenge the way vulnerability is currently used to categorise disadvantaged children within the humanitarian architecture. I demonstrate that, despite existing policies and the moral urgency to protect all children at risk, in practice intervention agencies give greater attention to child combatants, neglecting others just as vulnerable in mining-and-conflict-affected areas.