Dataset Electrical crisis, dwellability, and psychosocial distress: Discursive evidence of infrastructural solastalgia on X/Twitter during Ecuador’s 2024 power outages

This dataset documents digital conversation about Ecuador’s 2024 electricity crisis and was assembled for the study of discursive markers of psychosocial distress in X/Twitter posts. It includes a corpus of 3,953 posts in Spanish, collected between 22 September and 19 December 2024, together with a daily table of outage hours constructed from official government announcements of scheduled power cuts by date and region. Because outage schedules were not identical across all cities, the daily outage variable was defined using the duration reported for the majority of the population on each date.
The deposit also includes a labeled subcorpus of 1,581 posts, extracted from the full corpus and organized as a quasi-balanced set by category, month, and outage hours for model training and evaluation. Annotation was carried out through close reading of each post, taking into account tone, subjectivity, insults, profanity, Ecuadorian localisms, and references to disrupted everyday routines such as sleeping, cooking, working, studying, and inhabiting the home or neighborhood. This process was formalized in a codebook with operational definitions, inclusion and exclusion criteria, and examples.
The annotation scheme includes seven main emotion categories—Support/Guidance, Stress/Overload, Frustration/Powerlessness, Informative/Neutral, Anger/Indignation, Fear/Insecurity, and Sadness/Discouragement—as well as three binary discourse markers: Solastalgia, Anxiety, and Depressive Affect. In this framework, solastalgia captures posts expressing disrupted habitability, persistent alteration of routine, deterioration of the everyday environment, and loss of stability associated with prolonged outages; anxiety captures discursive traces of worry, nervousness, threat anticipation, and sustained insecurity; and depressive affect groups expressions of discouragement, exhaustion, persistent sadness, and deeper emotional wear. These variables were designed to support computational social science analyses of how the daily duration of outages related to digitally expressed psychosocial distress.
The repository contains the base corpus (x_data.csv), the daily outage-hours table (horas_apagones.csv), the labeled training subcorpus (corpus_etiquetado_40pct_quasi_balanceado.csv), the full heuristically annotated corpus (corpus_etiquetado_completo_heuristico.csv), the codebook (codebook_etiquetas_apagones_ecuador.csv), the labeled-sample summary (resumen_muestra_etiquetada.csv), and the two main analysis notebooks used for training/evaluation and article outputs. Tweets/posts, labels, documentation, and supporting materials are in Spanish.