Researching algorithms and recommendation systems inside out using reversed engineering

Abstract

It is difficult for a lone researcher to obtain an unfiltered perspective of algorithms in digital media and recommendation systems in digital platforms because these companies tend to be reluctant to share insights into their algorithms and business models. Researchers therefore need to develop new methods to obtain knowledge. The method of reversed engineering, which explores an algorithm from the inside out by “re-engineering” how algorithms are set up, has been recommended for gaining empirical knowledge about how algorithms work and what they do. This paper uses reversed engineering to illustrate how algorithms in dating platforms calculate matches between single persons. One of the findings is that the matching algorithms in dating platforms follows a psychologica l discourse based on similarities in personality and other personal aspects between candidates when recommending matches, and thus ignoring socioeconomic principles (e.g. economic income, social class, educational level) that social science find important when choosing a life-long partnership. Another observation is that an algorithmic matching machinery based on similarities follows a 'more of the same' logic, which risks limiting the pool of single candidates the user gain access to.