Some forms of human-algorithm interaction result in the suppression of human agency. Slot machine players in Vegas, for example, describe their experiences in terms of feeling “paralyzed” or “hypnotized” by the machines’ loops, to the point of not being able to get up to use the bathroom. Nevertheless, by distinguishing two structures of human-algorithm interaction, I aim to show how algorithms can either suppress or augment our agency.

I define an algorithm as any fully automated procedure for transforming an input into an output—meaning, a process involving just the abilities to read and write symbols and move between functional states. An algorithmic decision system (ADS) is a system of algorithms executable with just these abilities. When a human is incorporated into an ADS, she uses just this subset of her abilities; in doing so, she acts “automatically” or “mechanically,” and her role in the ADS is substitutable.

What I term a human-algorithmic decision systems (HADS), in contrast, is a process that can be algorithmically decomposed into a set of subprocesses that includes at least one discursive practice-or-ability. As Brandom (2008) demonstrates, this necessitates the involvement of discursive creatures. I conclude by discussing aviation checklists as an example of HADS.