Conceptions of Cognition for Cognitive Engineering

Abstract

Cognitive processes, cognitive psychology tells us, unfold in our heads. In contrast, several approaches in cognitive engineering argue for a shift of unit of analysis from what is going on in the heads of operators to the workings of whole socio-technical systems. This shift is sometimes presented as part of the development of a new understanding of what cognition is and where the boundaries of cognitive systems are. Cognition, it is claimed, is not just situated or embedded, but extended and distributed in the world. My main question in this article is what the practical significance is of this framing of an expanded unit of analysis in a cognitive vocabulary. I focus on possible consequences for how cognitive engineering practitioners think about function allocation in system design, and on what the relative benefits and costs are of having a common framework and vocabulary for talking about both human and technical system components. I argue for what I call an expansive but deflated conception of cognition, primarily on pragmatic grounds. In addition, I claim that the important lesson of the “boundaries of cognition” debate in cognitive science is the negative claim that there is not anything special about the biological boundary of the skin and skull per se, rather than some positive claim about where the boundaries of extended or distributed cognitive systems really are. I also examine the role of the concept of cognition in the theoretical frameworks of distributed cognition, joint cognitive systems (also known as cognitive system engineering), and cognitive work analysis.

Notes

I am grateful to Gavan Lintern, Andy Clark, and Richard Harper for helpful comments and questions. I also want to thank Microsoft Research for funding.

¹As Halverson (1994) pointed out, knowledge about the opportunistic uses of meaningful structure in the environment such as these might be useful for engineers and system designers who are trying to replace or improve workplace tools. For another excellent case that illustrates the importance of such knowledge in air traffic control, see the literature on the role of paper flight strips and the various attempts made to introduce systems that get rid of them (Berndtsson & Normark, 1999; Fields, Wright, Marti, & Palmonari, 1998; Harper, Hughes, & Shapiro, 1989; Mackay, 2000; Marti, 2000; Rognin, Salembier, & Zouinar, 1998).

²Including related fields such as human–computer interaction (HCI), human factors, and computer-supported cooperative work. There are differences between these fields, but I will use the label cognitive engineering to refer to them collectively. Clark and Chalmers (1998) were by no means unaware of the connections between extended cognition and ideas in cognitive engineering. They mentioned what they called “the theory of situated cognition,” which they attribute to Suchman (1987)and Hutchins (1995a) as research that “reflects” their philosophical position. Both Suchman's and Hutchins's theoretical frameworks have been influential in cognitive engineering circles. Clark (2008, p. 96) also suggested that HCI houses “nascent forms” of a science of extended cognition.

³This view has its roots in the work of Newell and Simon (1976).

⁴Systems that have been described within the framework include the cockpit of a commercial airliner(Hutchins, 1995b; Hutchins & Klausen, 1996), a telephone hotline group (Ackerman & Halverson, 1998), software programming teamwork (Flor & Hutchins, 1991), work practice in an engineering company (Rogers & Ellis, 1994), and trauma resuscitation teamwork (Sarcevic, Marsic, Lesk,& Burd, 2008).

⁵This aspect of distributed cognition has not received a lot of philosophical attention, but see Rupert(2004, pp. 391–392, 425, n. 59) for some remarks about the relation between distributed and extended cognition views. In Hutchins (1995a), see the following pages: p. 196, regarding memory; pp. 182, 194, regarding perception and detection; chapters 4 and 5, passim, regarding expertise at navigation; and pp. 239–261, regarding hypothesis testing, interpretation, and confirmation bias. The claim that these are cognitive properties of a socio-technical system rather than human agents are not, as far as I can see, intended to be read metaphorically. Hutchins (1995a) wrote that “there is nothing metaphorical about talking about the bearing record book as a memory, or about viewing the erasure of lines drawn in pencil on a chart as forgetting” (p. 363).

⁶ Clark and Chalmers (1998) reject the idea that the debate is merely about terminology: “Thus, in seeing cognition as extended one is not merely making a terminological decision; it makes a significant difference to the methodology of scientific investigation. In effect, explanatory methods that might once have been thought appropriate only for the analysis of ‘inner’ processes are now being adapted for the study of the outer, and there is promise that our understanding of cognition will become richer for it” (p. 10).

⁷It might be that the best way of resolving the philosophical issue is to work out an answer to this latter more pragmatic question. I believe that this is what the debate between Rupert (2004, 2009) and Clark (2007) has ultimately boiled down to.

⁸Button (1997, 2008), a social scientist primarily involved in computer-supported collaborative work, argued that the distributed cognition framework results in distortive redescriptions of social phenomena in what he characterized as a “redundant cognitive vocabulary.” Part of his critique against Hutchins is based on a conflation between the two senses of representation at play in Hutchins's work.Button (2008) criticizes Hutchins, wrongly I think, for appealing to internal personal-level representations in his descriptions of situated human conduct (by “personal-level representations,” I mean representations that must be interpreted by persons). However, the conflation is understandable, as Hutchins does not explain what the relation is between the external personal-level representations that cognitive ethnographers are attending to and the internal representations referred to in cognitive scienti?c explanations pitched at the “subpersonal level” of explanation. For an interesting discussion about how representation should be understood in distributed cognition, see Osbeck and Nersessian (2006).

⁹ Hutchins (1995a) wrote,“I believe that humans actually process internal representations of symbols. But I don't believe that symbol manipulation is the architecture of cognition” (p. 370).

¹⁰Research on the role of flight strips seems to be one of the few points of exchange between aviation psychology and philosophy of cognitive science. The ethnographic work of Mackay (2000) and Mackay, Fayard, Frobert, and Médini (1998) was discussed by Noë (2009, pp. 84–87) and by Wheeler(2010), both in connection with the hypothesis of extended cognition.