This article examines how ethnographic fieldwork is made possible under conditions of limited transport, uneven mobility, and uncertain institutional access. Drawing on 12 months of fieldwork in rural hamlets in Kuyu District, Central Ethiopia, it develops the concept of social infrastructure of access to explain how existing relationships support and sustain ethnographic fieldwork over time. The findings show how a locally embedded schoolteacher acting as an interlocutor enabled the researcher to build trust, broaden participation, and gain access to people, institutions, and documentary materials that would have been difficult to reach through formal procedures alone. The study further demonstrates that ethnographic access develops through ongoing relationships, flexible field practices, and repeated engagement rather than a single moment of entry. The article contributes to methodological debates by conceptualizing access as a social infrastructure, sustained through interconnected relationships rather than a one-time event. More broadly, it shows how fieldwork under constraint can illuminate the relational conditions through which ethnographic access and knowledge production become possible.
This article describes a visual and participatory study exploring postsecondary students’ conceptualisations of feminism. We conducted the study in two phases: the first involved individual object elicitation interviews and the second involved group participatory data analysis sessions. Here, we focus on the participatory data analysis phase and the ‘amazing space’ some participants felt it engendered. We discuss two features that we see as related to this interpretation of the research process. The participatory data analysis sessions were, on the one hand, encounters suffused with intimacy and vulnerability, and on the other hand, dialogues that exposed participants to interpretive disagreements and tensions. We also reflect on how the study’s culminating research exhibit worked to extend the ‘amazing space’ cultivated during data collection.
In this paper, we report an exploratory collaboration in data analysis between a human researcher and Generative AI (GenAI). Specifically, we reveal how human-AI collaboration might contribute to alternative interpretations in theory-informed policy analysis. We draw on Karen Barad’s philosophy of agential realism as an anchor point to consider the ethico-onto-epistem-ological affordances and limitations of incorporating GenAI in social sciences research to guide the human-AI collaborative analysis. The exploratory collaboration reported is based on a case involving how a human researcher (the first author) engaged with Copilot, drawing upon a relevant analytical method, to analyse a Chinese education policy document. Through this example, we illustrate how a more intra-active human-AI collaboration could enable the human researcher to think the unthinkable , that is to think beyond the discursive or conceptual boundaries of his own understanding of the policy logics through the analytical framework. We argue that we can benefit from the analytical conversational collaboration process between AI and human intelligence, rather than simply the AI-generated output per se. This study is significant as it demonstrates a much more ethical, interdependent, and indeed intimate form of human-AI collaboration, in which the agency between humans and GenAI is recognised and respected. This more intimate form of human engagement with GenAI might also yield new ways of knowing and being as researchers in conducting qualitative policy analysis.
This paper compares two popular R packages for structural equation modelling (SEM) – lavaan and seminr – to help researchers understand not only how they work, but also when each is most appropriate. Although both tools allow users to estimate the same structural models, they are grounded in different methodological traditions and are designed to support different research goals. Using an identical model, we estimated results with both covariance-based SEM ( lavaan ) and variance-based SEM ( seminr ) and compared their outputs, including model specification syntax, evaluation criteria, and reporting conventions. The results show that both approaches lead to substantively similar conclusions regarding the relationships between constructs, while differing in emphasis: lavaan provides richer global model-fit diagnostics, whereas seminr places greater emphasis on prediction-oriented assessment and convenient access to latent variable scores. The contribution of this study lies in its practical, hands-on demonstration rather than in a theoretical or simulation-based comparison. The findings reinforce that there is no universally “better” SEM approach; instead, methodological choice should be guided by the research objective. Researchers focussed on theory testing may benefit more from lavaan , while those prioritising prediction or exploratory analysis may find seminr more suitable. Ultimately, considering both perspectives can support more transparent, robust, and methodologically appropriate SEM applications.
