17
views
0
recommends
+1 Recommend
0 collections
    0
    shares
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      An Active Inference view of cognitive control

      editorial
      1 , 2
      Frontiers in Psychology
      Frontiers Media S.A.

      Read this article at

      Bookmark
          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

          Abstract

          A commentary on What ever next? Predictive brains, situated agents, and the future of cognitive science by Clark, A. (in press). Behav. Brain Sci. The Active Inference framework (Friston et al., 2009; Friston, 2010) argues that the brain's generative models continuously produce predictions and goals that guide its action (active inference) and perception (predictive coding) through free energy minimization. In this framework, most studies have focused on the on-line prediction of perceptual events and the control of overt behavior. We propose that the framework can be extended to explain cognitive control. We assume that architectures of cognitive control are elaborations of the predictive architectures of sensorimotor behavior in early living organisms. As the sensorimotor control system of early organisms evolved (to face increasingly harder individual and social problems), it gradually began predicting increasingly long-term and abstract consequence of behavior and—critically—doing so off-line and without overt behavior. This permitted rehearsing action sequences without executing them. In turn, off-line predictions opened the doors to higher cognitive abilities, such as planning, emulation, imagery, mental state inference, goal-directed decision-making, prospection, and the acquisition of declarative knowledge (Hesslow, 2002; Grush, 2004; Jeannerod, 2006; Pezzulo and Castelfranchi, 2007, 2009; Moulton and Kosslyn, 2009; Pezzulo, 2011; Clark, in press). We argue that off-line predictions opened the doors to executive functions and cognitive control, too. Executive functions, typically linked to prefrontal cortex, are self-directed actions functional to self-regulation and the coordinated planning of present and future actions and goals toward distal objectives. Executive functions (and prefrontal cortex functioning) have been linked to a plethora of processes, including the self-generation of goals and plans, their maintenance in working memory, their monitoring, the inhibition of prepotent but inappropriate actions, the regulation of attention, and valuation processes. A reconciling perspective is that prefrontal cortex supports “cognitive control”: the control of thought processes and the top–down guidance of overt behavior toward distal goals (Fuster, 1997; Miller and Cohen, 2001; Botvinick, 2008). We cast cognitive control within the Active Inference framework. We argue that cognitive control consists in a nesting of optimizations (i.e., free energy minimization loops) over time; in addition to the usual overt loop of active inference, one (or more) covert loop(s) help optimizing distal goals. The left part of Figure 1 shows the Active Inference framework, in which predictions and goals steer perception and action via free energy minimization. Note that here predictions are relative to the present situation. The right part of Figure 1 shows an extension of the framework that includes Controlled Predictions and a nesting of optimizations (for simplicity we only show two loops). Figure 1 Left: Active Inference framework. Right: “Controlled Predictions” and cognitive control. The covert loop works off-line via the suppression of overt sensory and motor processes (in the Active Inference framework, this requires the suppression of proprioception). This permits running imaginary actions that produce a sequence of fictive actions and of predictions relative to future (rather than present) situations. Fictive actions and predictions can be optimized via free energy minimization but without overt execution: they are not just “mind wandering” but are truly controlled toward goals specified at higher hierarchical levels. Prospection and planning are thus optimization processes that support the generation of distal and abstract goals (and associated plans), beyond current affordances. The covert loop supports also the recall of learned contextual goals and rules rather than always forming them de novo via active inference. The selected goals can thus be set as priors for the overt loop. In the Active Inference framework, this has the same effect as goal maintenance in working memory, and affords the top–down guidance of overt behavior beyond stereotyped responses, which is the hallmark of cognitive control. [Note that specifying contextual rules or plans to distal goals requires setting strong priors over sets of states and transitions rather than just one set point (Friston et al., 2012)]. Active inference in the overt loop ensures that the goals (and plans) generated by the covert loop are achieved in practice. In turn, feedback from the overt loop is informative of (changing) environmental constraints; it permits revising and re-situating imagined goals, and ultimately achieving them in the current context. In some cases, the overt loop is recruited for planning and thinking, too, such as for instance when Tetris players rotate blocks to better decide where to place them (Kirsh and Maglio, 1994). Our view is compatible with theories of cognitive control that highlight the active maintenance of goals in working memory (Miller and Cohen, 2001). However, we assume that cognitive control and mental operations are internalized forms of overt actions and recruit the same brain structures, rather than involving separated neural representations and dedicated, modular processing (Diamond, 2000; Barkley, 2001; Cotterill, 2001; Ito, 2008; Pezzulo, 2011). We propose that the nesting of optimization processes is mainly realized in prefrontal hierarchies and is functional to the achievement of goals at different time scales. Compared to overt action, cognitive control optimizes at longer time scales, and requires maintaining goal representation (priors) for a long time. It is thus not surprising that cognitive control processes (and in our proposal, covert loops) mainly involve high hierarchical levels. This creates a potential confound, suggesting that cognitive control might also require dedicated neural resources or modularized processing. However, our proposed architecture can be implemented within a homogeneous (cortical) hierarchy performing free energy minimization; which ones of the hierarchical levels are recruited to work covertly depends on task demands (say, abstract thought vs. the mental imagery of rotating an object with the left hand). Our view suggests that executive functions and forethought might have resulted from an internalization of the process of predicting the consequences of actions, which permitted to endogenously steer and control covert predictive loops. Partial support for this view comes from the close relationships between the neural systems for motor preparation and mental imagery (Cisek and Kalaska, 2010). Furthermore, in our proposal cognitive control requires the coordination of overt and covert processes. Disruption of this process might result in the execution of imagined actions, as it was reported by Schwoebel et al. (2002) in the case of a patient with bilateral parietal lesions who was unable to refrain from executing imagined (hand) movements.

          Related collections

          Most cited references13

          • Record: found
          • Abstract: found
          • Article: not found

          Control of mental activities by internal models in the cerebellum.

          Masao ITO (2008)
          The intricate neuronal circuitry of the cerebellum is thought to encode internal models that reproduce the dynamic properties of body parts. These models are essential for controlling the movement of these body parts: they allow the brain to precisely control the movement without the need for sensory feedback. It is thought that the cerebellum might also encode internal models that reproduce the essential properties of mental representations in the cerebral cortex. This hypothesis suggests a possible mechanism by which intuition and implicit thought might function and explains some of the symptoms that are exhibited by psychiatric patients. This article examines the conceptual bases and experimental evidence for this hypothesis.
            Bookmark
            • Record: found
            • Abstract: found
            • Article: not found

            Close interrelation of motor development and cognitive development and of the cerebellum and prefrontal cortex.

            Motor development and cognitive development may be fundamentally interrelated. Contrary to popular notions that motor development begins and ends early, whereas cognitive development begins and ends later, both motor and cognitive development display equally protracted developmental timetables. When cognitive development is perturbed, as in a neurodevelopmental disorder, motor development is often adversely affected. While it has long been known that the striatum functions as part of a circuit with dorsolateral prefrontal cortex, it is suggested here that the same is true for the cerebellum and that the cerebellum may be important for cognitive as well as motor functions. Like prefrontal cortex, the cerebellum reaches maturity late. Many cognitive tasks that require prefrontal cortex also require the cerebellum. To make these points, evidence is summarized of the close co-activation of the neocerebellum and dorsolateral prefrontal cortex in functional neuroimaging, of similarities in the cognitive sequelae of damage to dorsolateral prefrontal cortex and the neocerebellum, of motor deficits in "cognitive" developmental disorders, and of abnormalities in the cerebellum and in prefrontal cortex in the same developmental disorders.
              Bookmark
              • Record: found
              • Abstract: found
              • Article: not found

              The emulation theory of representation: motor control, imagery, and perception.

              Rick Grush (2004)
              The emulation theory of representation is developed and explored as a framework that can revealingly synthesize a wide variety of representational functions of the brain. The framework is based on constructs from control theory (forward models) and signal processing (Kalman filters). The idea is that in addition to simply engaging with the body and environment, the brain constructs neural circuits that act as models of the body and environment. During overt sensorimotor engagement, these models are driven by efference copies in parallel with the body and environment, in order to provide expectations of the sensory feedback, and to enhance and process sensory information. These models can also be run off-line in order to produce imagery, estimate outcomes of different actions, and evaluate and develop motor plans. The framework is initially developed within the context of motor control, where it has been shown that inner models running in parallel with the body can reduce the effects of feedback delay problems. The same mechanisms can account for motor imagery as the off-line driving of the emulator via efference copies. The framework is extended to account for visual imagery as the off-line driving of an emulator of the motor-visual loop. I also show how such systems can provide for amodal spatial imagery. Perception, including visual perception, results from such models being used to form expectations of, and to interpret, sensory input. I close by briefly outlining other cognitive functions that might also be synthesized within this framework, including reasoning, theory of mind phenomena, and language.
                Bookmark

                Author and article information

                Journal
                Front Psychol
                Front Psychol
                Front. Psychology
                Frontiers in Psychology
                Frontiers Media S.A.
                1664-1078
                05 November 2012
                2012
                : 3
                : 478
                Affiliations
                [1] 1Istituto di Linguistica Computazionale “Antonio Zampolli”, Consiglio Nazionale delle Ricerche, via Giuseppe Moruzzi Pisa, Italy
                [2] 2Istituto di Scienze e Tecnologie della Cognizione, Consiglio Nazionale delle Ricerche, via S. Martino della Battaglia Roma, Italy
                Author notes

                This article was submitted to Frontiers in Theoretical and Philosophical Psychology, a specialty of Frontiers in Psychology.

                Edited by: Shimon Edelman, Cornell University, USA

                Reviewed by: Shimon Edelman, Cornell University, USA

                Article
                10.3389/fpsyg.2012.00478
                3488938
                23133435
                d442f60c-6bfa-4833-a4e6-296d10508ee1
                Copyright © 2012 Pezzulo.

                This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.

                History
                : 01 October 2012
                : 17 October 2012
                Page count
                Figures: 1, Tables: 0, Equations: 0, References: 20, Pages: 2, Words: 1568
                Categories
                Psychology
                General Commentary Article

                Clinical Psychology & Psychiatry
                Clinical Psychology & Psychiatry

                Comments

                Comment on this article