A group of cognitive scientists, initially drawn from cognitive ergonomics and anthropology, have lobbied for a shift in the main unit of analysis to understand thinking (e.g., Hollan et al., 2000
). They reject a traditional model of the mind where cognition is sandwiched between perceptual inputs and behavioral outputs (to adapt Hurley, 2001
). Instead they argue that cognition is the product of a distributed system that reflects the dynamic meshwork of resources internal to the reasoner (such as cognitive capacities, acquired knowledge) as well as resources external to the reasoner (such as artifacts, people, cultural beliefs; Kirsh, 2009
; Hutchins, 2010
). A key notion in the systemic perspective is that people interact with external resources to augment and facilitate thinking. From a distributed-cognition perspective, thinking is the product of embodied and embedded mental and physical activities. In other words, people do not just “think
with their heads,” they also “think
with their eyes and hands” in an environment that affords interaction. This results in an extended cognitive system (Wilson and Clark, 2009
), akin to an ecological niche (cf. Laland et al., 2000
) enabling people to exceed the capacities of their unaided, non-extended mind.
People act upon their environment when they think, and more specifically when they evince a preference, in a rich and varied manner; yet this activity is rarely the focus of research. People, generally, do not choose their homes or their cars from written descriptions. Rather, they walk in potential flats, project and sketch furniture placement, open and close wardrobes, sit on the terrace to help simulate what it would be like to live in the place. In other words, they do not only adapt to their environment, they actively shape, manipulate, and interact with it to support their decision-making.
The distributed perspective has been the subject of ethnographic analysis “in the wild” (Hutchins, 1995
), but it can also guide more controlled experimental work (Fioratou and Cowley, 2009
; Weller et al., in press
). For example, we recently examined performance on matchstick algebra problems which present participants with a false algebraic equation made of matchsticks and require them to move one matchstick to form a true equation (Knoblich et al., 1999
). Adopting a distributed-cognition perspective, we compared performance on the traditional paper-and-pencil version of the task with performance in an interactive version where participants could physically manipulate the matchsticks, using a modifiable, three-dimensional, physical presentation of the equation. Participants in the interactive group were significantly more likely to achieve insight to transform these expressions into true equations. Moreover, whereas numeracy predicted performance in the paper-and-pencil group, performance was best predicted by visuo-spatial reasoning skills in the interactive group. These results suggest that different types of resources and skills were recruited in the interactive and non-interactive versions of the task, respectively.
The distributed-cognition perspective may also offer a novel way to conceive the role of the environment in the construction of preferences. The theoretical frameworks reviewed earlier assume that the environment shapes cognitive activity. In experiments used to test these approaches, however, the environment is often presented in a two-dimensional, fixed presentation akin to the non-interactive version of the matchstick algebra task, offering linguistic or numerical information that is presented in essentially inflexible and intangible formats. These environments severely limit individuals’ natural tendency to think with their eyes and hands. The distributed-cognition perspective could offer a new window onto the process of preference construction, focusing on the coupling between people's cognition and the strategic and opportunistic manipulation of the information populating their immediate physical space. As Weller et al.’s (in press
) study illustrates, adopting a distributed perspective on cognition does not necessitate studying cognitive activities in naturalistic settings. In fact, we believe that the potential of this approach resides in its promise to better capture the essence of cognitive processes in general, and preference construction in particular, within laboratory settings.
Adopting a distributed-cognition perspective also highlights a potentially invalid assumption underpinning alternative accounts of preference construction, such as SDS theory (Montgomery, 1998
) and the DiffCon theory (Svenson, 1996
) reviewed above – and more generally, numerous theories accounting for higher level cognitive processes – namely, the assumption that the mental restructuring
of a rigid presentation of the informational landscape is equivalent to the physical restructuring
of this landscape, in the individual's immediate environment. It is not: an inflexible physical problem presentation exerts gravity on people's effort to depart from and transform their representation of the information; whereas a dynamic one may better support the development of a productive representation of the problem information.
Concretely, better understanding how preferences may be constructed in the physical world will involve designing experimental settings where participants are no longer limited to alter the information presented to them mentally. This, we surmise, will lead to a revision of the amount of information that people are actually capable of computing when constructing preferences. For example, a canonical representation of the information in choice framing tasks such as the Asian Disease problem (Tversky and Kahneman, 1981
) requires taking into account all outcomes of concurrent decisions. Such a bias-free representation has been previously ruled out as psychologically implausible, assuming that it would exceed human computational capabilities (Kahneman and Tversky, 1984
). Maule and Villejoubert (2007
) surmised that participants might instead mentally switch between a gain-framed representation and a loss-framed representation, in a similar manner to the perspective-switching occurring when people are presented with ambiguous figures such as the Necker cube. Choice behavior would then be determined by the dominant representation at the moment of choice. Taking a distributed-cognition approach to study choice framing, one could use playing cards presenting a positive or negative outcome associated with each of two alternatives. Probabilities of outcomes would be presented as the relative proportion of positive and negative outcomes. This would enable participants to manipulate, spread, arrange and rearrange the cards, and perhaps contrast losses and gains while constructing their preference. Importantly, rather than constrain thinking, the manipulability afforded by the material presentation of the information would instead support – if not augment – people's computational abilities. In such a situation, the mental switch of focus between a gain-framed and a loss-framed representation (Maule and Villejoubert, 2007
) could then be supported by the physical presentation of the information and thus, considerably reducing the mental efforts required for switching focus. Moreover, this would make the process of restructuring directly accessible to the researcher, through the observation and coding of the actions and eye-gazes executed by the decision-makers.
To conclude, Simon's (1956
) emphasis on the major shaping role played by the environment within which decisions are made has often been used to explain how preferences are constructed. Simon's argument has often been summarized as focusing on the “interaction” between individuals’ mental activities and their immediate environment (e.g., Brighton and Todd, 2009
, p. 339; Lichtenstein and Slovic, 2006
, p. 23; Bettman et al., 1998
, p. 187). However, interactivity as such never figures in either Simon's (1956
) account or in subsequent theoretical efforts. Some have developed theories explaining how decision-makers may select choice heuristics that are fitted to
the structure of the environment. Others have stressed the importance of the mental restructuring of the information in preference construction. In this article we sought to illustrate how neither approaches can fully account for the essence of preference construction as it may occur in natural settings. We propose that this is because past research has neglected an important aspect of cognition – viz., how interactions with the world may influence and support mental processes. Whether, under what conditions, and by which processes, freeing up decision-makers’ hands may indeed affect the way they construct their preferences, may thus prove to be an important new avenue for research.