This conceptual paper addresses the uses of epistemological metaphors (Thagard & Beam, 2004) in an emerging educational domain with the key areas of computational thinking, programming and data science. The point of departure of the analysis is descriptions of abstraction, widely considered to be a core aspect of computational thinking. Abstraction in this context is often described as the removal of ‘irrelevant’ details to make a problem accessible to algorithmic solutions. Some authors, most notably Stephen Wolfram, further claim that computational thinking makes the world and by extension content in different school subjects more transparent and easier to understand. This yields the impression that the result of abstraction is simply a better or more useful picture of the world or subject matter, not a picture from a very specific point of view. The use of visual metaphors by Wolfram are further analysed drawing on Robert Romanyshyn´s (1989) study of the development of the linear perspective in art and its radical consequences for cultural understandings of the relationship between the world, humans and technology. The conceptual analysis also describes alternative metaphors grounded in empirical work in the field of data science, including Roth´s (2013) analysis of scientists’ recontextualizations of abstracted data and Philip, Olivares-Pasillas and Rocha’s (2016) study of racial aspects of visualizations that emphasizes dispute and antagonism. In response to these and Wolfram´s approach, a reflexive pedagogy of computational thinking is considered that raises the question, can what is lost in abstraction become the figure?