By Luis Favela
In my previous post, I highlighted some of the defining features of phenomena investigated by complexity science. These features include an emphasis on system dynamics (i.e., how they evolve over time), nonlinearity (i.e., interactions of parts are not always additive), emergence (i.e., collective behaviors can be difficult to predict based on knowledge of constituents alone), and self-organization (i.e., collective behavior resulting from interactions of parts and not guided by a central controller). With my remaining posts, I will discuss some of the ways I’ve put complexity science to work in my own research. Today’s post centers on extended cognition.
Since at least the late 1990s, “extended mind” has been a controversial idea in the philosophy of mind. A literature review is way beyond the scope of this post. But, in short, ‘extended mind’ refers to the idea that mental states are not necessarily limited to the brain—constitutively and/or causally (depending who you talk with). In order to get a grip on “extended mind,” my colleagues and I have thought it more fruitful to talk about “extended cognition” instead. When talking about “mind,” one is bound to discuss the metaphysics of mental states. This can make empirical investigations of extended states enormously challenging because conceptions of mind that are metaphysically necessary do not always (ever?) line up with empirically manageable conceptions. Thus, we talk about “extended cognition” as the idea that cognitive systems extend beyond the boundary of the organism. Built into this definition are four claims: First, cognition is not restricted to the brain. Second, cognition is necessarily tied to sensorimotor processes. Third, although the body is a constraint upon cognition, it is not the constraint upon cognition; that is, cognition is not restricted to the boundaries of the organism. Finally, and following from the first three features, cognition is a phenomenon that spans brain, body, and environment.
That extended cognition is an issue that boils down to empirical facts is something that some proponents and opponents can agree on. Thus, with the set of claims above in place, our next step was to empirically study it. So we did. Following earlier work that set a precedent for empirically investigating extended cognitive systems my colleagues and I designed an experiment to see if sensory-substitution devices can become part of extended cognitive systems. Sensory-substitution devices are tools that allow a user to substitute the capacities of one sensory modality with capacities of another. Brail is a prototypical example of sensory substitution, where touch replaces vision when reading. Canes are common sensory-substitution devices, where touch replaces vision when navigating the world. In my experiment, I compared three ways of obtaining information about the world when making judgments about one’s action capabilities. One modality was normal vision, and the other two were haptic and involved the use of tools while the participants wore blindfolds and earmuffs. The first tool was a cane and the second was a nifty gadget called the Enactive Torch. Since we have not published this yet, I don’t want to give away too many details. What I will say is that participants performed just as accurately (statistically speaking) with the three modalities (vision, cane, and Enactive Torch). That participants performed as accurately with the three modalities suggests at least one reason to think cognition (i.e., in the form of perception-action capabilities) was extended—at least causally—with the haptic tools. But that’s not a super strong reason to believe in extended cognitive systems.
In order to bolster our case for extended cognitive systems, namely, that the results of our experiment suggests that sensory-substitution devices can become part of extended cognitive systems, we needed a more compelling reason. Moreover, we wanted a reason to suggest that not only could the tools be causally related to cognition, but that they partially constitute the extended cognitive system. In order to do that, we looked at the movement data generated while the participants explored the world during the task. Data analyses of participants’ movements suggest that the tools were part of their cognitive system. But what kind of data analysis and signal could show that? Detrended fluctuation analysis (a kind of fractal analysis) is one such method and pink noise is one such signal. It has been experimentally demonstrated that many natural phenomena exhibit pink noise (a.k.a. ‘1/f scaling or ‘1/f noise’), which is a type of variability in a data series. Pink noise is fractal structure in a data series, where the patterns of variability at smaller spatial scales or shorter timescales are statistically similar in structure to variability at larger spatial scales or longer timescales. This sort of structured variability is a predictable consequence of interaction dominance (mentioned in my previous post). In an interaction-dominant system variability/“noise” of individual component W propagates through the whole system, which alters the dynamics of other components, e.g., X, Y, and Z. Because of the connections among the system’s components, altering the dynamics of X, Y, and Z will lead to alterations to W’s dynamics. That initial random fluctuation of component W will reverberate through the system for some time. Pink noise should be expected when the components of a system are so tightly integrated with one another that their activity cannot be explained independently. Fluctuations in an interaction-dominant system percolate through the system over time, leading to the kind of correlated structure to variability that is pink noise. Exhibiting pink noise is initial evidence that a system is interaction dominant. For more on interaction dominance see Ihlen & Vereijken 2010, Van Orden et al. 2003, Van Orden et al. 2010, and Wagenmakers et al. 2005. Now back to the experiment:
Analysis of participants’ movement data while using sensory-substitution devices indicated the presence of pink noise. In other words, the tool can be thought of as component W, which partially constituted the system X, Y, Z. The system (i.e., participant wielding a sensory-substitution device) exhibited the kinds of signals (i.e., pink noise) seen in many natural systems such as heartbeats, gait patterns, and brain activity, just to name a few. So, we take the presence of pink noise during the task, together with the accuracy of judgments, to be indicative of an extended cognitive system, where the tools were not merely causally related to the system but were constituents of the system. If this is right, then cognition extended beyond the brain and the body. I think this is pretty cool. But, will it satisfy the metaphysician who cares more about “extended minds” than mere cases of “extended cognition?” Probably not—but that’s a topic for another day.