Dataset for the paper: An Outlook on the Labor Market in relation with Artificial Intelligence in Chile: how Labor and training policies should prepare to meet this new challenge?

Abstract

Phase 1: Selection of the data source. Four online job portals were selected (i.e. CompuTrabajo, LinkedIn, Trabajando.cl and ChileTrabajos.cl). These sites were chosen according to three hierarchical criteria (in terms of representativeness and amount of available information) (Gontero & Menéndez, 2021): i) priority for the most visited websites in Chile in the category of job portals, according to Similarweb metrics (2023); ii) Websites where the most vacancies related to AI appeared during the systematic review process (exploration); and iii) data that contain a significant amount of relevant information in terms of the identified variables. Although LinkedIn was not among the five most visited job portals in Chile, it was chosen for its inclusion due to its satisfactory fulfilment of the rest of criteria. No metasearch engines were included to avoid duplicity in the advertisements. Overall, a total of 1,100 online job vacancies were collected during the three months of monitoring. Out of these, 45.45% were obtained on LinkedIn, followed by CompuTrabajo (28.18%), ChileTrabajos (15.73) and, finally, Trabajando (10.64). Phase 2: Collection of unstructured open data from the selected sources The data collection was carried out systematically every day in 2023 from August 5 to November 5, using the Web scraping technique in the same way as other studies in this field (Gontero & Menéndez, 2021; Mamani, 2022). An in-depth review of the literature and the analysis of the exploratory results was launched with the purpose of optimizing the search process for job demands related to AI, as a result, the following keywords were established as search criteria: ‘Artificial intelligence’ ‘Prompt or prompt engineer’ ‘Neural networks’ ‘Machine Learning’ ‘Deep learning’ and ‘IA or AI’ (in two languages, English and Spanish). As mentioned above, this made it possible to collect information on 1,100 vacancies. These numbers are considered suitable for content analysis methodologies (Gardiner et al., 2018; Verma et al., 2019, 2022). Phase 3: Database processing and construction It is important to highlight that the advertisements have no standard structure for the different websites, which makes the processing and construction of the database more complex. In this sense and based on the data collected in the previous step, four activities were carried out: cleaning the data, identification of variables; construction of the database, and identification of skills.