The challenge: Searching for Research on Red AI, the Environmental Impacts of Artificial Intelligence

We report this case to serve dual ends. First, we share our lessons learned to inform how others might investigate Red AI, a term for the environmental impact of artificial intelligence drawn from a seminal paper on this topic (Schwartz et al., 2020).1 Second, we seek to raise awareness of the complexities and trade-offs in “tech mining” certain research domains. Novel to us was the degree to which the target in this case confounded with other literature, necessitating human screening of many papers (abstract records) to determine suitability for inclusion in the study data set. We report a mixed data sampling approach and a multifaceted, iterative search refined over more than two years that does not lend itself to straightforward replication.