Scholarly article on topic 'Interpreting psychological notions: A dual-process computational theory'

Interpreting psychological notions: A dual-process computational theory Academic research paper on "Psychology"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"Dual process" / CLARION / "Cognitive architecture" / "Cognitive modeling" / Implicit / Explicit / Instinct / Intuition}

Abstract of research paper on Psychology, author of scientific article — Ron Sun

Abstract The distinction between implicit versus explicit processes (or “intuitive” versus “reflective” thinking) is arguably one of the most important distinctions in cognitive science. Given that there has been a great deal of research on explicit processes (“reflective” thinking), it is important in studying the human mind to consider implicit processes, treating them as an integral part of human thinking. A cognitive architecture (a comprehensive computational theory) may be used to address, in a mechanistic and process-based sense, issues related to the two types of processes (including their relation, interaction, and competition) and their relevance to social and organizational research.

Academic research paper on topic "Interpreting psychological notions: A dual-process computational theory"

' '"iffHW^M MIIILE IN PRESS

journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

ELSEVIER

Interpreting psychological notions: A dual-process computational theory

Ron Sun *

Cognitive Sciences Department, Rensselaer Polytechnic Institute, Troy, NY 12180, USA

ABSTRACT

The distinction between implicit versus explicit processes (or "intuitive" versus "reflective" thinking) is arguably one of the most important distinctions in cognitive science. Given that there has been a great deal of research on explicit processes ("reflective" thinking), it is important in studying the human mind to consider implicit processes, treating them as an integral part of human thinking. A cognitive architecture (a comprehensive computational theory) may be used to address, in a mechanistic and process-based sense, issues related to the two types of processes (including their relation, interaction, and competition) and their relevance to social and organizational research.

© 2014 Society for Applied Research in Memory and Cognition. Published by Elsevier Inc. All rights

reserved.

Contents lists available at ScienceDirect

Journal of Applied Research in Memory and Cognition

journal homepage www.elsevier.com/locate/jarmac

ARTICLE INFO

Article history: Received 15 October 2013 Accepted 23 September 2014 Available online xxx

Keywords: Dual process CLARION

Cognitive architecture

Cognitive modeling

Implicit

Explicit

Instinct

Intuition

1. Introduction

The distinction between "intuitive" and "reflective" thinking has been, arguably, one of the most important distinctions in cognitive science. There are currently many dual-process theories (two-system views) out there. However, although the distinction itself is important, the terms involved have been somewhat ambiguous. Not much fine-grained analysis has been done, especially not in a precise, mechanistic, process-based way (Kruglanski & Gigerenzer, 2011; Sun, 1994, 2002). In this article, toward developing a more fine-grained and more comprehensive framework, I will adopt the more generic but less loaded terms of implicit and explicit processes (Reber, 1989; Sun, 2002) and present a more nuanced view of these processes, centered on a computational "cognitive architecture".

Given that there has been a great deal of research on explicit processes ("reflective thinking") and the apparent significance of implicit processes (Sloman, 1996; Sun, 1994), it is important, in studying the human mind, to more seriously consider implicit processes. I have argued that we need to treat implicit processes as an integral part of human thinking, reasoning, and decision-making, not as an add-on or an auxiliary (Sun, 1994,2002). This point applies

* Tel.: +1 518 276 3409. E-mail address: dr.ron.sun@gmail.com

not only to studying the individual mind, but also to studying collective processes involving multiple individuals (and the interaction of their minds) such as in organizational research (Sun & Naveh, 2004).

In this short summary article, a brief review of the background concerning implicit and explicit processes will be given. Then a theoretical framework (based on a computational cognitive architecture) will be presented that addresses, in a mechanistic, process-based sense, issues concerning dual-process theories. Specifically, issues concerning different types of implicit processes, their relations to explicit processes, and their relative speeds may be addressed within the framework. The notions of instinct, intuition, and creativity are important in this regard and will be briefly discussed also. This framework will then be applied to social and organizational modeling where its relevance will also be demonstrated. Connections will also be made to the notion of rationality in economic and organizational research.

2. Some background

To better explore the distinction between implicit and explicit processes, let us examine some background first. There are many dual-process theories (two-system views) currently available (e.g., as reviewed by Evans & Frankish, 2009). One such two-system view was proposed early on in Sun (1994,1995). In Sun (1994), the two systems were characterized as follows:

http://dx.doi.org/10.1016/jjarmac.2014.09.001

2211-3681/© 2014 Society for Applied Research in Memory and Cognition. Published by Elsevier Inc. All rights reserved.

R'liff^M iRIIILE IN PRESS

R. Sun / Journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

"It is assumed in this work that cognitive processes are carried out in two distinct 'levels' with qualitatively different mechanisms. Each level encodes a fairly complete set of knowledge for its processing, and the coverage of the two sets of knowledge encoded by the two levels overlaps substantially." (Sun, 1994, p. 44)

That is, the two "levels" (i.e., the two modules or components) encode somewhat similar or overlapping content. But they encode their contents in different ways: symbolic and subsymbolic representation is used, respectively. Symbolic representation is used by explicit processes at one "level", and subsymbolic representation is used by implicit processes at the other. Different mechanisms are involved at these two "levels" due to the representational differences. It was hypothesized that these two different "levels" can potentially work together synergistically, complementing and supplementing each other (Sun, Slusarz, & Terry, 2005; Sun, 1994). This may in part explain evolutionarily why there are these two levels.

However, some more recent two-system views are somewhat different, and their claims seem more contentious. For instance, a more recent two-system view was proposed by Kahneman (2003, 2011). The gist of his ideas was as follows: There are two styles of processing: intuition and reasoning. Intuition (or System 1) is based on associative reasoning, fast and automatic, involving strong emotional bonds, based on formed habits, and difficult to change or manipulate. Reasoning (or System 2) is slower, more volatile, and subject to conscious judgments and attitudes.

Evans (2003) espoused a similar view. According to him, System 1 is "rapid, parallel and automatic in nature: only their final product is posted in consciousness"; he also notes its "domain-specific nature of the learning". System 2 is "slow and sequential in nature and makes use of the central working memory system", and "permits abstract hypothetical thinking that cannot be achieved by System 1". Moreover, in terms of the relation between the two systems, he argued for a default-interventionist view: System 1 is the default system, while System 2 may intervene when feasible and called for (see Evans, 2003, for more details).

However, some of these claims seem, in a way, simplistic to me. For one thing, intuition (System 1, and implicit processes in general) can be very slow, not necessarily faster than explicit processes (System 2) (see Bowers, Regehr, Balthazard, & Parker, 1990; Helie & Sun, 2010). For another, intuition (and implicit processes in general) may sometimes be subject to conscious control and manipulation; that is, it may not be entirely "automatic" (Berry, 1991; Curran & Keele, 1993; Stadler, 1995). Furthermore, decisions made implicitly can be subject to conscious "judgment" (Gathercole, 2003; Libet, 1985). In terms of the relationship between the two systems, implicit and explicit processes may be parallel and mutually interactive in complex ways instead of being limited to being default-interventionist (Sloman, 1996; Sun, 1994, 2002) and so on.

It thus seems necessary that we come up with more nuanced and more detailed characterizations of the two systems (the two types of processes) in order to avoid painting the picture in too broad strokes. To come up with a more nuanced, more detailed, and more justifiable characterization, it is important that we ask some key questions. For either type of process, in any given situation, the following questions, for instance, may be asked:

• How deep is its processing (in terms of precision, certainty, and so on)?

• How much information is involved (how broad is its processing)?

• How incomplete, inconsistent, or uncertain is the information available?

• How many processing "cycles" are needed considering the factors above?

And there are many other similar or related questions. See also Evans and Stanovich (2013) and Kruglanski and Gigerenzer (2011). Asking such questions may lead to better characterizations of the two systems and useful interpretations of related notions. But in order to do so, one has to rely on some basic theoretical frameworks to begin with.

3. A theoretical framework

In order to sort out these issues and answer these questions in a manageable way, below, I will describe a theoretical framework that can potentially provide some clarity to these issues and questions. The framework is based on the CLARION cognitive architecture (Sun, 2002, 2003, 2014), viewed at a theoretical level, used as a conceptual tool for generating interpretations and explanations (Sun, 2009).

The framework consists of a number of basic principles. The first principle is the distinction and division between procedural (i.e., action-oriented) and declarative (i.e., non-action-oriented) processes, which is rather uncontroversial (see, e.g., Anderson & Lebiere, 1998; Tulving, 1985). The next two principles concern implicit and explicit processes, but not just one simple division as in many other dual-process theories. Thus the next two principles are unique to this theoretical framework, and may require some justifications, which have been argued in, for example, Sun (2012) and Sun (2014). The second principle is the distinction and division between implicit and explicit procedural processes (e.g., Sun et al., 2005). The third principle is the distinction and division between implicit and explicit declarative processes (e.g., Helie & Sun, 2010). Therefore, in this framework, there is a four-way division: implicit and explicit procedural processes and implicit and explicit declarative processes. These different processes may run in parallel and interact with each other in complex ways (e.g., as described in Sun, 2014).

The divisions above between implicit and explicit processes may be related to some existing computational paradigms, for example, symbolic-localist versus connectionist distributed representation (Sun, 1994, 1995). As has been extensively argued before (e.g., Sun, 1994, 2002), the consciously (relatively) inaccessible nature of implicit knowledge may be captured by distributed connectionist representation, because distributed representational units are subsymbolic and generally not individually meaningful. This characteristic of distributed representation, which renders the representational form less accessible computationally, accords well with the relative inaccessibility of implicit knowledge in a phenomenological sense. In contrast, explicit knowledge may be captured by symbolic-localist representation, in which each unit is more easily interpretable and has a clearer conceptual meaning.

4. A sketch of a cognitive architecture

Now with the basic principles outlined, I will sketch an overall picture of the CLARION computational cognitive architecture itself, which is centered on these principles. Only a quick sketch is possible here (without getting into too much technical details though); for details, the reader is referred to the references cited below.

CLARION is a generic cognitive architecture - that is, a comprehensive, generic model of psychological processes of a wide variety, specified computationally. It has been described in detail and justified on the basis of psychological data (Sun, 2002, 2003, 2014).

CLARION consists of a number of subsystems. Its subsystems include the Action-Centered Subsystem (the ACS), the Non-Action-Centered Subsystem (the NACS), the Motivational Subsystem (the

' '"li^Pf^M MIIILE IN PRESS

R. Sun / Journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

MS), and the Meta-Cognitive Subsystem (the MCS). Each of these subsystems consists of two "levels" of representations, mechanisms, and processes as theoretically posited earlier (see also Sun, 2002). Generally speaking, in each subsystem, among the two "levels", the "top level" encodes explicit knowledge (using symbolic-localist representation) and the "bottom level" encodes implicit knowledge (using distributed representation; Rumelhart, McClelland, the PDP Research Group, 1986).

Among these subsystems, the Action-Centered Subsystem is responsible for procedural processes, that is, to control actions, utilizing procedural knowledge (Sun et al., 2005). Among procedural processes, implicit procedural processes are captured by MLP networks (i.e., Backpropagation networks; at the bottom level of the ACS). Explicit procedural processes, on the other hand, are captured by explicit "action rules" (at the top level of the ACS).

The Non-Action-Centered Subsystem is responsible for declarative processes, that is, to maintain and utilize declarative (non-action-centered) knowledge for information and inferences (Helie & Sun, 2010). Among these processes, implicit declarative processes are captured by associative memory networks (Hop-field type networks or MLP networks, at the bottom level). Explicit declarative processes are captured by explicit "associative rules" (at the top level).

The Motivational Subsystem is responsible for motivational dynamics, that is, for providing underlying motivations for perception, action, and cognition (in terms of providing impetus and feedback). Implicit motivational processes are captured by MLP networks for drive (implicit basic motive) activations (at the bottom level; Sun, 2003, 2014). Explicit motivational processes are centered on explicit goal representation (at the top level).

The Meta-Cognitive Subsystem is responsible for metacognitive functions, that is, for monitoring, directing, and modifying the operations of the other subsystems. Implicit metacognitive processes are captured by MLP networks (at the bottom level), while explicit metacognitive processes are captured by explicit rules (at the top level; Sun, 2003, 2014).

The two levels within each subsystem interact. The interaction between the two levels includes bottom-up and top-down activation flows. Bottom-up activation is the "explicitation" of implicit information, through activations of nodes at the top level by corresponding nodes at the bottom level. Top-down activation is the "implicitation" of explicit information, through activations of nodes at the bottom level by corresponding nodes at the top level. For example, there may be an inhibitory role for explicit processes to suppress implicit information (Gathercole, 2003).

The interaction between the two levels also includes bottom-up and top-down learning. Bottom-up learning means implicit learning first and explicit learning on that basis (Sun, Merrill, & Peterson, 2001). This may be viewed as online knowledge extraction from neural networks: Implicit knowledge may be learned implicitly (through reinforcement learning within the neural networks at the bottom level), and then may be explicated to become explicit knowledge (at the top level). Top-down learning means explicit learning first and implicit learning on that basis. Explicit knowledge may be explicitly learned, and may be assimilated into implicit processes, through a gradual process (e.g., through reinforcement learning at the bottom level on the basis of explicit rules).

The interaction between the two levels also includes the integration of the results from the two levels. See Fig. 1 for a sketch of the CLARION cognitive architecture. The proportion of implicit versus explicit processing in the integration may be determined by the MCS taking into consideration a number of factors (Sun, 2003, 2014).

Fig. 1. The four subsystems of CLARION. ACS stands for the Action-Centered Subsystem. NACS stands for the Non-Action-Centered Subsystem. MS stands for the Motivational Subsystems. MCS stands for the Metacognitive Subsystem.

5. Interpreting folk psychological notions

Based on the framework above, we may re-interpret some common folk psychological notions, to give them some clarity and precision.

For instance, the notion of instinct may be interpreted and made more specific by appealing to the CLARION framework. Instinct, according to its common, colloquial usage, involves mostly implicit processes and is mostly concerned with action. Within CLARION, instinct may be roughly equated with the following chain of activations: stimulus ^ drive ^ goal^ action. This chain goes from stimuli received to the MS, then the MCS, and eventually the ACS. That is, stimuli activate drives (especially those representing essential innate motives; Sun, 2014), drive activations lead to goal setting in a (mostly) implicit, direct (and often innate) way, and based on the goal set, actions are selected in a (mostly) implicit way to achieve the goal. Instinct is mostly implicit, but it may become more explicit, especially with regard to the part of "goal ^ action" (Sun et al., 2001).

For another instance, the notion of intuition can also be made more concrete by using the CLARION framework. Intuition, according to CLARION, is roughly the following chain: stimulus ^ drive ^ goal ^ implicit thinking. This chain goes from stimuli received to the MS, the MCS, the ACS, and the NACS. As such, intuition mostly involves implicit declarative processes within the NACS (at its bottom level, as directed by the ACS), including those common functionalities within the NACS such as associative memory retrieval, soft constraint satisfaction, and partial pattern completion. Intuition is often complementary to explicit reasoning, and the two types are used often in conjunction with each other (Helie & Sun, 2010).

Some other folk psychological notions may be re-interpreted and made more precise in a similar manner. For example, the notion of creativity may be captured within the CLARION framework. Creativity may be achieved through complex, multi-phased implicit-explicit interaction, that is, through the interplay between intuition and explicit reasoning, according to Helie and Sun's (2010) theory of creative problem solving - a theory derived from the CLARION cognitive architecture. It includes (1) the explicit phase: processing given information (mostly) explicitly through reasoning using explicit declarative knowledge (at the top level of the Non-Action-Centered Subsystem); (2) the implicit phase: developing intuition using (mostly) implicit declarative knowledge (at the

R'lfflffW^M IRIIILE IN PRESS

4 R. Sun / Journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

bottom level of the NACS); (3) finally as the intuition emerges into explicit processes, the explicit phase again: verifying and validating the result using (mostly) explicit declarative knowledge (at the top level of the NACS). See Helie and Sun (2010) for further details. This theory has been successful in accounting for a variety of empirical data related to creativity (e.g., in relation to incubation or insight generation).

What about the relations among these different types of processes, especially in terms of their relative time courses, as alluded to earlier (Evans, 2003; Kahneman, 2011)? An issue was raised earlier concerning fast versus slow processes with regard to different two-systems views. The twin divisions in CLARION, procedural versus declarative and implicit versus explicit, definitely have implications for identifying slow versus fast processes. We may question the conventional wisdom on a number of issues in this regard, instead of simply assuming the seemingly obvious as in some existing views; for instance,

• In terms of the division between procedural and declarative processes, can fast procedural versus slow declarative processes be posited?

• In terms of the division between implicit and explicit procedural processes, can fast implicit versus slow explicit processes be posited?

• In terms of the division between implicit and explicit declarative processes, can fast implicit versus slow explicit processes be likewise posited?

• What about relative speeds if we consider the four-way division together?

And so on. The conjectures as implied by the questions above may not be exactly accurate. The whole picture is not so simple; it is a lot more complex according to the framework discussed thus far.

In this regard, we may view existing models and simulations of these types of processes as a form of theoretical interpretation. In that case, we have so far the following tentative conclusions:

• Fast procedural versus slow declarative processes (as hypothesized earlier): This hypothesis of the speed difference is generally likely to be true if we examine many existing models and simulations viewing them as theoretical interpretations (Sun, 2003, 2014; see also Anderson & Lebiere, 1998).

• Fast implicit versus slow explicit procedural processes: This hypothesized speed difference is, again, generally likely to be true, using theoretical interpretations through modeling and simulation (Sun et al., 2001, 2005).

• Fast implicit versus slow explicit declarative processes: This hypothesis of the speed difference, however, is generally not true. Intuition (implicit declarative processes) may (or may not) take a long time, compared with explicit declarative processes. See, for example, Helie and Sun (2010) and Bowers et al. (1990) for possible interpretations that contradicted this hypothesis.

Thus, we need to be careful in making sweeping generalizations. We may need to characterize different types of processes in a more fine-grained fashion than the conventional wisdom would have it. Characteristics of different processes may also vary in relation to contexts (such as task demands).

Many empirical and simulation studies have been conducted within the CLARION framework that shed light on these issues, and substantiate the interpretations made above. See Helie and Sun (2010), Sun and Zhang (2006) and Sun, Zhang, and Mathews (2009), and many other prior publications for details (note that work on other existing cognitive architectures may also be relevant, at least to some of these points highlighted above).

6. Social and organizational implications

The social and organizational implications of the implicit-explicit distinction within the CLARION framework have also been explored. Since it is often impossible to run laboratory experiments on large-scale social phenomena, they need to be investigated through alternative means, including through multiagent social simulation. In particular, cognitive social simulation -social simulation that is based on detailed cognitive models - can be very helpful here, which can accommodate well two-system views (Sun, 2006).

There are relatively few detailed computational models of the cognitive processes associated with social and organizational decision-making, especially the implicit processes involved. However, studies have shown that not only reflective, rational analyses (explicit processes) but also implicit processes play important roles in social and organizational decision-making. Can cognitive architectures, connectionist models, and other computational psychological theories aid social and organizational research, while taking into account the implicit-explicit dichotomy? The answer is yes. For example, a number of cognitive social simulations have been carried out on the basis of the CLARION cognitive architecture, taking into consideration the implicit-explicit dichotomy. They include, for instance:

• Certain aspects of organizational decision-making

• Patterns of growth of academic science

• Survival of tribal society

• Moral judgment and ethical norms

Of course, they did not cover the whole range of organizational decision-making or other social science topics. But we have shown the relevance of the CLARION framework to social and organizational research (Naveh & Sun, 2006; Sun & Naveh, 2004, 2007; Sun, 2006). In this regard, we have not reached the limit yet, so we do not yet know where the ultimate limit might be. More explorations are needed.

As an extremely simple example, let us look into a commonly used organizational decision task (from Carley, Prietula, & Lin, 1998). In this task, there is an object in the airspace. An organization must determine its status: whether it is friendly, neutral or hostile. No one single agent has access to all the information necessary to make a correct decision. Organizational decisions are made by integrating separate decisions made by different agents.

In terms of organizational structures, there are two types: teams (in which individual decisions are treated as votes, and the organizational decision is chosen by the majority), and hierarchies (in which recommendations are passed from subordinates to superiors and the decision of a superior is based solely on the recommendations of its subordinates). In terms of the structure of information accessible by each agent, there are two varieties: distributed access, in which each agent sees a different subset of three attributes, and blocked access, in which multiple agents see exactly the same subset of attributes. In Carley et al. (1998), limited human experiments were done in a 2 x 2 fashion (organization x information access). The human data showed that humans generally performed better in team situations. Moreover, distributed information access was generally better than blocked information access.

It seemed worthwhile to undertake a simulation that involved a more comprehensive, more psychologically realistic agent model (i.e., CLARION) that took into account a two-system view. Moreover, with the use of a more psychologically realistic agent model, the importance of different cognitive capacities and parameters in affecting organizational performance might be individually investigated.

' '"ll^PH^M ARTICLE IN PRESS

R. Sun / Journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

Table 1

Simulation data (in terms of accuracy rate, i.e., percentage of correct classifications) compared with the human data from Carley et al. (1998). B stands for Blocked. D stands for Distributed.

Agent/Org. Team (B) Team (D) Hierarchy (B) Hierarchy (D)

Human 50.0 56.7 46.7 55.0

CLARION 53.2 59.3 45.0 49.4

The results of the CLARION simulation (with 4000 cycles of training for each agent) were as shown in Table 1 (Sun & Naveh, 2004). The results closely accorded with the patterns of the human data, with teams outperforming hierarchies, and distributed access outperforming blocked access.

Furthermore, it was shown in Sun and Naveh (2004) that effects of cognitive parameters, such as proportions of explicit versus implicit processing, were often significant on organizational performance. For example, the effect of explicitness of processing (probability of using the top versus the bottom level of CLARION) was significant, and a certain proportion of explicit versus implicit processing helped to improve performance. Moreover, its interaction with length of training was significant as well. As indicated by Fig. 2, explicit processing was very useful at the early stages of learning, when increased reliance on it tended to boost performance. However, by the 20,000th cycle, this effect disappeared. This was because explicit knowledge was crisp guidelines and they provided a useful anchor at the uncertain early stages of learning. However, after a long learning process, they were too coarse-grained to cover all possible contingencies, and were no more reliable than highly trained neural networks (embodying implicit processing). Many similar points were made. Such results may be outdated by now, but they are still of historical interest because they first pointed to the possibilities of combining realistic human psychology, social and organizational research, and computational modeling.

Beyond this simple example, other work has been carried out within the CLARION framework that investigated many other factors related to the implicit-explicit dichotomy, as well as other aspects of human cognition-psychology, and their social and organizational implications. See, for example, Naveh and Sun (2006), Sun and Naveh (2007), and Sun and Fleischer (2012). Many of the

Fig. 2. The effect of probability of using the bottom level (implicit processing) on performance over time. The X axis indicates the probability of using implicit processes. The Y axis indicates the accuracy of performance. The two curves indicate the performance after 3000 and 20,000 training cycles, respectively.

folk psychological notions that were touched upon earlier can also have interesting relevance to social and organizational research.

More generally speaking, in relation to economics and organization theory, Herbert Simon proposed his theory of bounded rationality that tried to reflect real human abilities to reason and to make decisions. A limited kind of rationality might indeed enable the social sciences to move beyond existing theories in some way. But does it go far enough in "respecting" human reality? What are the true human conditions in this regard?

For instance, Herbert Simon claimed that "Anything that gives us new knowledge gives us an opportunity to be more rational" (see Spice, 2000, p. A-11). Maybe, but how "rational" should one be? What is the proper mixture of reflective and intuitive thinking (implicit and explicit processes; Sun, 2002)? For example, for creative problem solving, we need to rely on intuition (as opposed to reflective, rational analysis) to a significant degree (Helie & Sun, 2010; Kahneman & Klein, 2009). After all, to be somewhat "irrational" is to be human. The human mind necessarily involves a lot of implicit, intuitive, instinctual, emotional, seemingly irrational processes. Herbert Simon also pointed out that "Technology may create a condition, but the questions are what do we do about ourselves. We better understand ourselves pretty clearly and we better find ways to like ourselves" (Spice, 2000, p. A-11). In my opinion, this point applies as well to understanding implicit, "irrational" processes, including in social and organizational research. It is of fundamental importance to appreciate and harness these "irrational" processes in the individual mind and in social and organizational thinking.

Conflict of Interest

The author has no conflict of interest to report.

Acknowledgements

The work overviewed here has been supported in part by the ONR grants N00014-08-1-0068 and N00014-13-1-0342. Thanks are due to my students and collaborators, past and present, whose work has been reviewed here. The CLARION software, along with simulation examples, may be found at: www. clarioncognitivearchitecture.com (courtesy of Nick Wilson and Mike Lynch). Thanks are also due to the three reviewers for their detailed comments.

References

Anderson, J. R., & Lebiere, C. (1998). The atomic components of thought. Mahwah, NJ:

Lawrence Erlbaum Associates. Berry, D. (1991). The role of action in implicit learning. Quarterly Journal of Experimental Psychology, 43A, 881-906. Bowers, K., Regehr, G., Balthazard, C., & Parker, J. (1990). Intuition in the context of

discovery. Cognitive Psychology, 22, 72-110. Carley, K. M., Prietula, M.J., & Lin, Z. (1998). Design versus cognition: The interaction of agent cognition and organizational design on organizational performance. Journal of Artificial Societies and Social Simulation, 1 (3) Curran, T., & Keele, S. (1993). Attention and structure in sequence learning. Journal

of Experimental Psychology: Learning, Memory, and Cognition, 19,189-202. Evans, J. St. B. T. (2003). In two minds: Dual-process accounts of reasoning. Trends

in Cognitive Sciences, 7(10), 454-459. Evans, J. St. B. T., & Frankish, K. (Eds.). (2009). In two minds: Dual processes and beyond.

Oxford, UK: Oxford University Press. Evans, J. St. B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8, 223-241, 263-271.

Gathercole, S. (2003). Short-term and working memory. London: Taylor and Francis. Helie, S., & Sun, R. (2010). Incubation, insight, and creative problem solving: Aunified

theory and a connectionist model. Psychological Review, 117,994-1024. Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded

rationality. American Psychologist, 58(9), 697-720. Kahneman, D. (2011). Thinking, fast and slow. New York: Farrar, Straus and Giroux. Kahneman, D., & Klein, G. (2009). Conditions for intuitive expertise: A failure to disagree. American Psychologist, 64, 515-526.

llfflBBI^M ARTICLE IN PRESS

R. Sun / Journal of Applied Research in Memory and Cognition xxx (2014) xxx-xxx

Kruglanski, A. W., & Gigerenzer, G. (2011). Intuitive and deliberate judgments are based on common principles. Psychological Review, 118, 97-109.

Libet, B. (1985). Unconscious cerebral initiative and the role of conscious will in voluntary action. Behavioral and Brain Sciences, 8,529-566.

Naveh, I., & Sun, R. (2006). A cognitively based simulation of academic science. Computational and Mathematical Organization Theory, 12(4), 313-337.

Reber, A. (1989). Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118(3), 219-235.

Rumelhart, D., McClelland, J., & the PDP Research Group. (1986). Parallel distributed processing: Explorations in the microstructures of cognition. Cambridge, MA: MIT Press.

Sloman, S. (1996). The empirical case for two systems of reasoning. Psychological Bulletin, 119,3-22.

Spice, B. (October 16, 2000). CMU's Simon reflects on how computers will continue to shape the world. Pittsburgh Post-Gazette, A-10-A-11.

Stadler, M. A. (1995). Rote of attention in implicit learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21,674-685.

Sun, R. (1994). Integrating rules and connectionism for robust commonsense reasoning. New York: John Wiley and Sons.

Sun, R. (1995). Robust reasoning: Integrating rule-based and similarity-based reasoning. Artificial Intelligence (AIJ), 75(2), 241-296.

Sun, R. (2002). Duality of the mind: A bottom-up approach toward cognition. Mahwah, NJ: Lawrence Erlbaum Associates.

Sun, R. (2003). A Tutorial on CLARION 5.0. Technical Report. Cognitive Science Department, Rensselaer Polytechnic Institute. http://www.cogsci.rpi.edu/rsun/ sun.tutorial.pdf

Sun, R. (2006). Cognition and multi-agent interaction: From cognitive modeling to social simulation. New York: Cambridge University Press.

Sun, R. (2009). Theoretical status of computational cognitive modeling. Cognitive Systems Research, 10(2), 124-140.

Sun, R. (2012). Memory systems within a cognitive architecture. New Ideas in Psychology, 30, 227-240.

Sun, R. (2014). Anatomy of the mind. New York: Oxford University Press (in press).

Sun, R., & Fleischer, P. (2012). A cognitive social simulation of tribal survival strategies: The importance of cognitive and motivational factors. Journal of Cognition and Culture, 12(3-4), 287-321.

Sun, R., & Naveh, I. (2004). Simulating organizational decision-making using a cog-nitively realistic agent model. Journal of Artificial Societies and Social Simulation, 7(3). http://jasss.soc.surrey.ac.uk/7/3/5.html

Sun, R., & Naveh, I. (2007). Social institution, cognition, and survival: A cognitive-social simulation. Mind and Society, 6(2), 115-142.

Sun, R., & Zhang, X. (2006). Accounting for a variety of reasoning data within a cognitive architecture. Journal of Experimental and Theoretical Artificial Intelligence, 18(2), 169-191.

Sun, R., Merrill, E., & Peterson, T. (2001). From implicit skills to explicit knowledge: A bottom-up model of skill learning. Cognitive Science, 25, 203-244.

Sun, R., Slusarz, P., & Terry, C. (2005). The interaction of the explicit and the implicit in skill learning: A dual-process approach. Psychological Review, 112, 159-192.

Sun, R., Zhang, X., & Mathews, R. (2009). Capturing human data in a letter counting task: Accessibility and action-centeredness in representing cognitive skills. Neural Networks, 22, 15-29.

Tulving, E. (1985). How many memory systems are there? American Psychologist, 40, 385-398.