Abstract
We used the folk theory perspective to investigate Internet users’ understanding of algorithms during their Internet use. Empirically, we conducted a mixed-method study. First, we carried out semi-structured in-person interviews with 30 German Internet users. Our analysis of these interviews enabled us to identity five folk theories – economic orientation theory, personal interaction theory, popularity theory, categorization theory, and algorithmic thinking theory. In a second step, we created a standardized survey questionnaire with 19 illustrative statements for these five folk theories, relying on participants’ explanations in the interviews to develop statements that reflected lay users’ ideas as much as possible. Participants (N = 331) were recruited through a commercial online access panel using quota criteria for age, gender, and education level to have a sample representative of the German population. Our survey findings indicate the prevalence of such folk theories among a broader population of Internet users, except for the algorithmic thinking theory, which is likely due to it being based on inaccurate assumptions about algorithms’ capabilities.
Notes
1 Comparable to folk theories, mental models are sets of ideas developed by lay individuals to understand what social world around them is, how it works, and to describe, explain, and predict events they encounter (Banks 2020). Research on mental models shows that individuals tend to form analogies based on similar domains to explain a phenomenon with which they are unfamiliar. This process involves the transfer of an existing relational structure into another domain (e.g. to explain electricity systems through the operations of water-flow systems) (Jones et al. 2011).
2 Data used in this study is part of a larger research project inquiring Internet users’ awareness, perceptions, and strategies regarding the operation of algorithms in different areas of Internet use. In the semi-structured interviews, we used a combination of open-ended and standardized questions, e.g. several sorting tasks to assess to what extent participants believe algorithms impact their own and other’s Internet use. The interviews further covered users’ risk perceptions, coping strategies associated with these risks, and regulatory preferences, which are not covered in the present article. Readers who are interested in other findings of this project are encouraged to read our article focusing on the relationship between user’s awareness of algorithmic decision-making and their autonomy (Dogruel, Facciorusso, and Stark 2020).