1. I draw this metaphor from Collins and Pinch (1998), The Golem: What You Should Know about Science. It is very similar to E. T. Jaynes’ 2003 metaphor of statistical models as robots, although with a less precise and more monstrous implication. 
2. There are probably no algorithms nor machines that never break, bend, or malfunction. A common citation for this observation is Wittgenstein (1953), Philosophical Investigations, section 193. Malfunction will interest us, later in the book, when we consider more complex models and the procedures needed to fit them to data. 
3. See Mulkay and Gilbert (1981). I sometimes teach a PhD core course that includes some philosophy of science, and PhD students are nearly all shocked ...