The next speaker at AoIR 2018 is Noemi Festic, whose focus is on algorithmic content selection processes by automated systems. This includes search applications, recommendation systems, and a broad range of other automated tools; these govern user behaviour by limiting and shaping activities but thereby also provide a space for new forms of engagement.
Such applications have an effect on the social order in human societies at the macro-level, but the extent of that effect is debatable and needs to be tested by empirical research. Part of the question here is how aware users are of the agency of these algorithms, and what opportunities and risks do they associate with them.
The study pursued this through some 60 interviews with Swiss Internet users across four domains of everyday life including politics, commercial transactions, socialising, and recreational uses. In each domain, the project worked with a wide variation of users for its 15 interviews, and the interviews also led to the development of a further quantitative online survey.
In each domain, offline sources remained more important to respondents than online sources; of the online tools, search applications were most important across all four domains. Generally, the duration of users’ engagement with automated applications did not necessarily correspond to the subjective relevance they assigned to them.
Experiences with such applications built awareness of algorithmic selection; personalised advertisements were especially important in raising users’ awareness, and this awareness crossed domains. Risk awareness emphasised different aspects across the different realms; the most important risk overall was a loss of privacy, and the domain where users perceived most risks was that of commercial transactions.
Users cope with such risks by engaging in advanced physical and cognitive practices (e.g. leaving their phone at home in order to avoid tracking); digital practices (e.g. deleting cookies or using multiple search engines); and platform-specific practices (e.g. confusing the platform about their preferences). The extent to which users engage in such practices varies with perceived risk and digital skills.