Futures Studies and Foresight for Science, Technology and Innovation: Trends of using Big Data and Machine Learning

Futures studies have shown an accelerated growth since the post-World War II period, in which governments and private companies have been attentive to the importance of "forecasting" new trends, mainly technological ones, for their security as an institution. Such studies have gained a new panorama from the proliferation of data on massive scales and the increasing processing capacity, leading to new approaches, mainly in data-driven studies. Big Data and Machine Learning (BDML) has become powerful tools to extract and analyze data for future-oriented activities. The central question about using BDML tools is to understand the specific impacts of these mechanisms on futures studies' conceptual and methodological approaches. This work intends to respond to these questions by analyzing academic publications about futures studies supported by BDML and the opinions of 479 futures studies experts. The proposed methodology aims to comprehend how these tools are employed, the future benefits and limitations of BDML in foresight. The bibliometric results point to a reduced but increasing number of prospective studies supported by BDML published in the past decades. In general, these studies employ BDML techniques such as text and data mining in at least one part of the foresight process. Futures studies experts' opinions suggested that 1) analytical competencies are essential to deal with the complexity of the digital revolution, and 2) robust data analysis and automated tools support the transfer of study results to policy- and strategy-making. However, 3) the lack of data reliability and manipulation can play an uncertain role in this environment. The thesis concludes that BDML impact future-oriented activities in three dimensions: 1) Data reliance, 2) Data-Method integration, and 3) Decision-making. Data manipulation may increase the perception of substantive uncertainty in futures studies. However, integrating BDML techniques in foresight methodologies strongly decreases procedural uncertainty and will support effective decision-making. The limitation of this work is mainly two. First, non-academic futures studies publications were not collected in the bibliometric analysis. Second, the expert's population and sample characteristics were not compared due to a limitation of population data in survey analysis.