Like Pessoa (Reference Pessoa2013), many authors (e.g., Anderson Reference Anderson2007; Newell Reference Newell1990; Reisenzein et al. Reference Reisenzein, Gratch, Hindriks, Hudlicka, Dastani, Lorini and Meyer2013) have called for theory integration within psychological science. Indeed, the importance of integrating disparate theories of human behavior together is unlikely to be disputed. However, the path toward theory integration, on the other hand, can be quite thorny. Between research literatures, differences often arise concerning language, assumptions, and methodologies. For example, Pessoa discusses how the role of the amygdala reaches far beyond its typical description as a “fear center” for processing negative information. Similarly, within the book Pessoa describes numerous research findings that reveal (1) overlapping patterns of activation in brain structures previously considered specific to “cognitive” or “emotional” processing and (2) relevant interactions between affective states and cognitive tasks. Indeed, we applaud Pessoa for exploring the often rugged territory between established fields of research and for building a strong case in favor of an integrative approach.
Yet, we also challenge Pessoa on the grounds of defining an approach to theory integration. In what follows, we make three points. First, we suggest that cognitive architectures can provide steady ground for integrating theories of emotion and cognition, as well as for describing interactions among underlying processes of behavior. A cognitive architecture is a broad psychological theory implemented as a formal (e.g., computer) model that incorporates multiple facets of behavior, such as perception and memory, and that, ideally, can account for many different behavioral tasks, ranging from, for example, low-level eye movement behavior (e.g., Salvucci Reference Salvucci2001) to deliberate problem solving (e.g., Anderson Reference Anderson2005). Second, we provide examples of how affective components have previously been (Cochran et al. Reference Cochran, Lee and Chown2006; Ritter et al. Reference Ritter, Reifers, Klein, Schoelles and Gray2007) and might potentially be (e.g., Reisenzein et al. Reference Reisenzein, Gratch, Hindriks, Hudlicka, Dastani, Lorini and Meyer2013) implemented within cognitive architectures. Third, we also briefly describe how neuroimaging data can potentially be incorporated into these types of models, as well as into the process of theory development.
Although the book offers a strong “why” for theory integration between the study of cognitive and affective aspects of behavior, we suggest that one promising avenue for “how” to accomplish this theory integration is by way of developing affective features within formalized cognitive architectures. Pioneering artificial intelligence and cognitive science researcher Alan Newell, argued (e.g., Reference Newell and Chase1973; Reference Newell1990) that psychological science would benefit by moving beyond mere verbal (qualitative) hypotheses, such as simple dichotomies (e.g., nature vs. nurture), toward formalized (quantitative) hypotheses. Additionally, he suggested that one path toward a unified theory of mind is by developing cognitive architectures. A number of cognitive architectures have been developed, such as Soar (Newell Reference Newell1990), EPIC (Meyer & Kieras Reference Meyer and Kieras1997) and ACT-R (Anderson & Lebiere Reference Anderson and Lebiere1998; see Langley et al. 2008 for a review of different architectures). Take ACT-R, for example (see Anderson Reference Anderson2007 for details). This model incorporates decades of research to describe a full range of cognitive processes, from perception to action, and can provide fine-grained predications about reaction times, neuroimaging measurements, eye-tracking data, as well as behavioral responses. In our view, it is quite stunning that, thus far, there have been relatively few attempts to incorporate affective components into architectural models of cognition and behavior. For the purpose of this commentary, the most noteworthy aspect of cognitive architectures relates to understanding and hypothesizing about interactions between different perceptual, motor, and cognitive components that naturally arise while modeling behavioral tasks. Within Pessoa's book and elsewhere (e.g., McGaugh Reference McGaugh2000), affective aspects of behavior such as stress, motivation, and arousal have been shown to modulate cognitive processes such as attention and memory, and we believe that developing these affective components within cognitive architectures can afford researchers the ability to precisely define how and where these types of interactions may take place within a human system. Additionally, when one or more aspects of cognition are qualified based on an affective state and a possible system-wide chain of interactions occurs, cognitive architectures may be the best tool for dealing with the high level of complexity.
How can affective components be implemented within cognitive architectures? The approach that several authors have called for or begun working with is to define how affective states might modulate the underlying cognitive processes (e.g., attention, working memory) within the architecture (e.g., Belavkin Reference Belavkin, Altmann, Cleeremans, Schunn and Gray2001; Cochran et al. Reference Cochran, Lee and Chown2006; Dancy et al. Reference Dancy, Ritter, Berry and Klein2013; Hudlicka Reference Hudlicka, Lovett, Schunn, Lebiere and Munro2004; Ritter et al. Reference Ritter, Reifers, Klein, Schoelles and Gray2007; see also Gunzelmann et al. Reference Gunzelmann, Gross, Gluck and Dinges2009 for similar work related to fatigue). This can translate to adjusting certain parameters within existing architectures. For example, Cochran et al. (Reference Cochran, Lee and Chown2006) provide a relatively simple demonstration of this approach, in which they model the effect of one aspect of emotion (arousal) within one cognitive module of ACT-R (declarative memory). Cochran et al. (Reference Cochran, Lee and Chown2006) point out that the standard ACT-R model is not able to predict the results of the classic study by Kleinsmith and Kaplan (Reference Kleinsmith and Kaplan1964), which found that study of high emotional arousal stimuli led to short-term forgetting and long-term remembering compared with low emotional arousal stimuli. To implement this impact of arousal on memory within ACT-R, Cochran et al. (Reference Cochran, Lee and Chown2006) redefined and expanded certain parameters (specifically, within the declarative memory module) to produce a pattern similar to the behavioral data. Similarly, in another paper, Ritter et al. (Reference Ritter, Reifers, Klein, Schoelles and Gray2007) developed a model within ACT-R to predict performance on a serial subtraction task, in which certain cognitive mechanisms within the architecture (e.g., attention, working memory) were modified to represent the impact of stress. Much more, we suspect that it would be worthwhile to explore how the findings and theories presented within Pessoa's book can be modeled within cognitive architectures in similar ways.
Many cognitive architectures (ACT-R in particular) not only attempt to model the processes underlying human behavior, but they also incorporate neuroimaging findings to develop a brain-like system of structures and processes (e.g., Anderson Reference Anderson2007; Just & Varma Reference Just and Varma2007). Indeed, within ACT-R different cognitive modules are associated with certain brain structures. Because of this design approach, (1) neuropsychological findings can be used to guide and constrain model development, and (2) neuroimaging data (such as fMRI) can be used in conjunction with behavioral measurements to help validate models (e.g., Borst & Anderson Reference Borst, Anderson, Forstmann and Wagenmakers2014). Because ACT-R provides latency information for different cognitive processes (e.g., visually encoding a stimulus, retrieving information from memory, producing a motor response), this pattern of activity can be translated into predictions for neuroimaging data in correspondence with the brain areas associated with the different cognitive modules. We suspect that this facet of cognitive architectures may be especially compelling for the development of affective components because, as Pessoa describes, certain brain structures (such as the amygdala) are associated with a variety of processes. These types of neuropsychological research findings can be taken into account when exploring how affective aspects might modulate particular processes within an architecture.
There is, perhaps, no better way to conclude this short commentary than by turning to one of the conceptual founders of integrative approaches to behavior and cognition. In many ways, Pessoa's book echoes Newell's (Reference Newell1990) argument that, “A single system (mind) produces all aspects of behavior. It is one mind that minds them all. Even if the mind has parts, modules, components, or whatever, they all mesh together to produce behavior.…If a theory covers only one part or component, it flirts with trouble from the start” (p. 17). In short, Pessoa contends that, given the high level of overlap between aspects of cognition and emotion, the two should not be considered separately. We agree with this and believe that the ideal research approach for pursuing this integration of theories includes cognitive architectures.
You have
Access
Like Pessoa (Reference Pessoa2013), many authors (e.g., Anderson Reference Anderson2007; Newell Reference Newell1990; Reisenzein et al. Reference Reisenzein, Gratch, Hindriks, Hudlicka, Dastani, Lorini and Meyer2013) have called for theory integration within psychological science. Indeed, the importance of integrating disparate theories of human behavior together is unlikely to be disputed. However, the path toward theory integration, on the other hand, can be quite thorny. Between research literatures, differences often arise concerning language, assumptions, and methodologies. For example, Pessoa discusses how the role of the amygdala reaches far beyond its typical description as a “fear center” for processing negative information. Similarly, within the book Pessoa describes numerous research findings that reveal (1) overlapping patterns of activation in brain structures previously considered specific to “cognitive” or “emotional” processing and (2) relevant interactions between affective states and cognitive tasks. Indeed, we applaud Pessoa for exploring the often rugged territory between established fields of research and for building a strong case in favor of an integrative approach.
Yet, we also challenge Pessoa on the grounds of defining an approach to theory integration. In what follows, we make three points. First, we suggest that cognitive architectures can provide steady ground for integrating theories of emotion and cognition, as well as for describing interactions among underlying processes of behavior. A cognitive architecture is a broad psychological theory implemented as a formal (e.g., computer) model that incorporates multiple facets of behavior, such as perception and memory, and that, ideally, can account for many different behavioral tasks, ranging from, for example, low-level eye movement behavior (e.g., Salvucci Reference Salvucci2001) to deliberate problem solving (e.g., Anderson Reference Anderson2005). Second, we provide examples of how affective components have previously been (Cochran et al. Reference Cochran, Lee and Chown2006; Ritter et al. Reference Ritter, Reifers, Klein, Schoelles and Gray2007) and might potentially be (e.g., Reisenzein et al. Reference Reisenzein, Gratch, Hindriks, Hudlicka, Dastani, Lorini and Meyer2013) implemented within cognitive architectures. Third, we also briefly describe how neuroimaging data can potentially be incorporated into these types of models, as well as into the process of theory development.
Although the book offers a strong “why” for theory integration between the study of cognitive and affective aspects of behavior, we suggest that one promising avenue for “how” to accomplish this theory integration is by way of developing affective features within formalized cognitive architectures. Pioneering artificial intelligence and cognitive science researcher Alan Newell, argued (e.g., Reference Newell and Chase1973; Reference Newell1990) that psychological science would benefit by moving beyond mere verbal (qualitative) hypotheses, such as simple dichotomies (e.g., nature vs. nurture), toward formalized (quantitative) hypotheses. Additionally, he suggested that one path toward a unified theory of mind is by developing cognitive architectures. A number of cognitive architectures have been developed, such as Soar (Newell Reference Newell1990), EPIC (Meyer & Kieras Reference Meyer and Kieras1997) and ACT-R (Anderson & Lebiere Reference Anderson and Lebiere1998; see Langley et al. 2008 for a review of different architectures). Take ACT-R, for example (see Anderson Reference Anderson2007 for details). This model incorporates decades of research to describe a full range of cognitive processes, from perception to action, and can provide fine-grained predications about reaction times, neuroimaging measurements, eye-tracking data, as well as behavioral responses. In our view, it is quite stunning that, thus far, there have been relatively few attempts to incorporate affective components into architectural models of cognition and behavior. For the purpose of this commentary, the most noteworthy aspect of cognitive architectures relates to understanding and hypothesizing about interactions between different perceptual, motor, and cognitive components that naturally arise while modeling behavioral tasks. Within Pessoa's book and elsewhere (e.g., McGaugh Reference McGaugh2000), affective aspects of behavior such as stress, motivation, and arousal have been shown to modulate cognitive processes such as attention and memory, and we believe that developing these affective components within cognitive architectures can afford researchers the ability to precisely define how and where these types of interactions may take place within a human system. Additionally, when one or more aspects of cognition are qualified based on an affective state and a possible system-wide chain of interactions occurs, cognitive architectures may be the best tool for dealing with the high level of complexity.
How can affective components be implemented within cognitive architectures? The approach that several authors have called for or begun working with is to define how affective states might modulate the underlying cognitive processes (e.g., attention, working memory) within the architecture (e.g., Belavkin Reference Belavkin, Altmann, Cleeremans, Schunn and Gray2001; Cochran et al. Reference Cochran, Lee and Chown2006; Dancy et al. Reference Dancy, Ritter, Berry and Klein2013; Hudlicka Reference Hudlicka, Lovett, Schunn, Lebiere and Munro2004; Ritter et al. Reference Ritter, Reifers, Klein, Schoelles and Gray2007; see also Gunzelmann et al. Reference Gunzelmann, Gross, Gluck and Dinges2009 for similar work related to fatigue). This can translate to adjusting certain parameters within existing architectures. For example, Cochran et al. (Reference Cochran, Lee and Chown2006) provide a relatively simple demonstration of this approach, in which they model the effect of one aspect of emotion (arousal) within one cognitive module of ACT-R (declarative memory). Cochran et al. (Reference Cochran, Lee and Chown2006) point out that the standard ACT-R model is not able to predict the results of the classic study by Kleinsmith and Kaplan (Reference Kleinsmith and Kaplan1964), which found that study of high emotional arousal stimuli led to short-term forgetting and long-term remembering compared with low emotional arousal stimuli. To implement this impact of arousal on memory within ACT-R, Cochran et al. (Reference Cochran, Lee and Chown2006) redefined and expanded certain parameters (specifically, within the declarative memory module) to produce a pattern similar to the behavioral data. Similarly, in another paper, Ritter et al. (Reference Ritter, Reifers, Klein, Schoelles and Gray2007) developed a model within ACT-R to predict performance on a serial subtraction task, in which certain cognitive mechanisms within the architecture (e.g., attention, working memory) were modified to represent the impact of stress. Much more, we suspect that it would be worthwhile to explore how the findings and theories presented within Pessoa's book can be modeled within cognitive architectures in similar ways.
Many cognitive architectures (ACT-R in particular) not only attempt to model the processes underlying human behavior, but they also incorporate neuroimaging findings to develop a brain-like system of structures and processes (e.g., Anderson Reference Anderson2007; Just & Varma Reference Just and Varma2007). Indeed, within ACT-R different cognitive modules are associated with certain brain structures. Because of this design approach, (1) neuropsychological findings can be used to guide and constrain model development, and (2) neuroimaging data (such as fMRI) can be used in conjunction with behavioral measurements to help validate models (e.g., Borst & Anderson Reference Borst, Anderson, Forstmann and Wagenmakers2014). Because ACT-R provides latency information for different cognitive processes (e.g., visually encoding a stimulus, retrieving information from memory, producing a motor response), this pattern of activity can be translated into predictions for neuroimaging data in correspondence with the brain areas associated with the different cognitive modules. We suspect that this facet of cognitive architectures may be especially compelling for the development of affective components because, as Pessoa describes, certain brain structures (such as the amygdala) are associated with a variety of processes. These types of neuropsychological research findings can be taken into account when exploring how affective aspects might modulate particular processes within an architecture.
There is, perhaps, no better way to conclude this short commentary than by turning to one of the conceptual founders of integrative approaches to behavior and cognition. In many ways, Pessoa's book echoes Newell's (Reference Newell1990) argument that, “A single system (mind) produces all aspects of behavior. It is one mind that minds them all. Even if the mind has parts, modules, components, or whatever, they all mesh together to produce behavior.…If a theory covers only one part or component, it flirts with trouble from the start” (p. 17). In short, Pessoa contends that, given the high level of overlap between aspects of cognition and emotion, the two should not be considered separately. We agree with this and believe that the ideal research approach for pursuing this integration of theories includes cognitive architectures.