The article we recently published presents and discusses the findings of a review of multidisciplinary academic abstracts on the data-driven algorithms shaping (or with the clear potential to shape) decision making across several criminal justice systems. Over the years, increased attention has been paid to the possibilities for big data analytics to inform law enforcement investigations, sentence severity, parole decisions, criminal justice resource allocation, and other key decisions. In some jurisdictions, data-driven algorithms are already deployed in these high-stakes and sensitive contexts, with major real-life implications. While their use can potentially enhance the efficiency of certain routines and time-consuming tasks, they can also give rise to adverse outcomes (notably racial and other types of systemic bias, along with other ethical concerns). Given the complexity of human behaviour, and other problems such as the reliance of some algorithms on flawed administrative data, algorithmic outputs (e.g., risk predictions) can be spurious. But key stakeholders such as the justice systems that deploy them and the general public might overestimate the reliability of these results, as they might not fully understand the complex mechanisms behind them.
While working on our research on cybercrimes and cyber harms (Lavorgna), and conduits of bias in predictive algorithms deployed in justice systems (Ugwudike), which are often inherently interdisciplinary research area, allowing us to read across a wide range of publications from several disciplines, we noticed how the portrayal of data-driven tools was very different depending on the disciplinary take. Similarly, by reading declarations by policy makers, the security industry, as well as the creators, vendors, and other proponents of the algorithms, we could not avoid noticing a certain hype around the effectiveness of these tools, in contrast with our own research experience. Intrigued by this puzzle, we decided to carry out some research and move beyond our anecdotical experience.
In our contribution, we now propose a typology of frames for understanding how relevant technologies are portrayed, and we elucidate how notions of sociotechnical imaginaries and access to digital capital are of the upmost importance in explaining differences in how the value and impact of the technologies are framed. We hope that our work can help further critical debates not only on algorithmic harms, but also on the importance of truly interdisciplinary research to facilitate the inclusion of a broader range of perspectives in our current, and unavoidable, datafication challenge.