Cognitive Engineering and Decision Making Technical Group

I have recently seen the phrase "functional allocation" making a comeback as our HFE community mobilizes to be prepared and relevant as AI Gold Rush II washes over all of us. Or maybe it's not a comeback - maybe I'm just now hearing again what our community has been saying for some time. I don't want to repeat the acrimony of the past (Yes, I know that I'm from Ohio State. Yes, I know that we have been part of this acrimony in the past), and I think some common grounding will (maybe) help to not repeat it.

Even though it's not really in the early celebrated reports by Fitts and Sheridan, our community ran with a very clean, tidy version of functional allocation, Levels of Automation, and capability attribution between people and the machines of the day. Unfortunately, other communities then ran with it, too, and it is seen as gospel truth among many of our computer science and software engineering colleagues. It has led the vast majority of the AI development community to think that a Level of Automation is a goal and a reliable description of performance, not just a description of the upper bound (i.e., max capability) of the machine counterparts. People that use this language usually (but not always, I know!) do not talk about how all activities are joint activities (thank you, Woods, Johnson, Vera, Bradshaw, Hoffman, etc.) and that even if a machine can independently perform the activities described by a certain LOA, or in a specific function allocated to them, doesn't mean that it effectively will perform that activity acceptably. This discourse is silient on the jointness that makes complex adaptive systems work. 

I implore our community to rally around jointness now. Now is the time. Focusing on innovations in joint human-machine configurations is NOT the same as functional allocation. The understanding that a joint activity architecture is not just (1) a functional allocation, (2) a joint human-machine configuration, or even (3) a set of configurations, but instead is (4) also how a system can fluidly (or not so fluidly) move between configurations is critical for AI Gold Rush II. This understanding will not organically come from MABA MABA conversations.

And I truly believe that using the words themselves affects our thinking in a way that obstructs us from innovating on these joint activity architectures. I really do.

Do you?