About

Patterns in Practice will explore how practitioners’ beliefs, values and feelings interact to shape how they engage with and in data mining and machine learning – forms of ‘narrow AI’. The project runs from August 2021 to February 2024 and is led by Jo Bates based at the University of Sheffield Information School, with Erinma Ochu, Helen Kennedy, Itzelle Medina Perea and Monika Fratczak.

Data and algorithms are becoming increasingly important resources for decision makers in organisations across sectors. Data mining and machine learning techniques allow analysts to find hidden patterns in the vast troves of data that organisations hold, producing predictive insights that can be actioned by others within the organisation or further afield. As applications of such techniques have become more common place, they have also become more controversial with concerns raised about, for example, discrimination and social manipulation. Across sectors practitioners are asking what good data practices look like and how they can be fostered. In 2017 the UK government launched the Centre for Data Ethics and Innovation to examine such issues.

While many data scientists are excited by these techniques and their potential to overcome perceived limitations of human judgement, for other groups of practitioners they can be perceived as an intrusive threat to privacy, an unwelcome challenge to professional insight, or dismissed as overhyped methods that produce poor quality information. Beliefs, values and feelings such as these, influenced by the cultures that practitioners are embedded within, are crucial factors that shape how the adoption and application of this type of AI unfolds in different contexts of practice. They also shape how different groups of practitioners come to relate to one another and the subjects of their data. Ultimately, practitioners’ beliefs, values and feelings shape how they come to understand what is desirable and ethical with regard to the application of such techniques in different contexts.

Research methods

In Patterns in Practice, we will use a combination of interviews, diaries, focus groups and observations to explore how the beliefs, values and feelings of different groups of practitioners shape how they engage with data mining and machine learning, and influence the evolution of cultures of data practice. We will examine the beliefs, values and feelings both of those developing and implementing applications that use data mining and machine learning techniques, and those being asked to use the outputs of such applications to inform their decision making. Since factors such as the novelty of application, individual and social implications, and the involvement of commercial interests can impact on people’s beliefs and feelings about the application of such technologies, we have decided to explore practitioners’ perceptions within three contrasting sectors in science, education and the arts: (1) mining chemical data to inform drug discovery in the pharmaceutical industry, (2) predictive learning analytics in UK universities, and (3) novel applications of data mining in the arts. Through exploring a diverse range of practitioners’ perspectives, we aim to build a rich picture about what they believe and how they feel about the application of data mining in different contexts. We will also engage an artist in residence on the project to create an artistic response to our findings using data mining techniques.

Building upon this empirical foundation, we aim to engage different groups of practitioners across the sectors to enhance their understanding of the ways in which their own and others’ beliefs, values and feelings can impact upon how they engage with data mining and machine learning applications and how this shapes how such applications become embedded, or not, into different organisational contexts.

Outcomes

Drawing on the deeper understanding we develop through our research, we aim to empower practitioners in the sectors we work with and relevant stakeholders to foster the development of critical and reflective data cultures that are able to exploit the possibilities of data mining and machine learning, while being critically responsive to their societal implications and epistemological limitations.