Home / Introduction / Motivation / Methods / Applications / More Information

I. Cognitive Processes. The methods described in this section primarily concern themselves with modeling the knowledge and cognitive activities necessary to perform work tasks. Cognitive Task Analysis seeks to model workers’ knowledge and cognition descriptively, whereas Computational Cognitive Models are more detailed analyses that can be run on a computer. Knowledge Elicitation techniques provide input to Cognitive Task Analyses and Computational Cognitive Modeling.
   
  A. Cognitive Task Analysis: Methods to analyze and represent the knowledge and cognitive activities workers utilize to perform complex tasks in the work domain. These methods focus primarily on how workers function in cognitively-demanding domains. They are most useful in developing training programs, developing means to assess performance, and developing criteria to select people for certain jobs. They may also provide insights into creating effective decision support systems and other software systems. 

References

 
    1. Applied Cognitive Task Analysis (ACTA): A method for performing Cognitive Task Analysis that consists of a series of three structured interviews. The first interview generates the Task Diagram, which provides a broad overview of the task and highlights difficult cognitive portions of the task that should be probed further. This is followed by a Knowledge Audit, which surveys the aspects of expertise required for a specific task or subtask. Finally, in the Simulation Interview step, the cognitive processes of experts are probed within the context of a specific scenario. The output of the process is a Cognitive Demands Table, which presents the results so they can be applied to a specific project.

References
    2. The Critical Decision Method (CDM): A past non-routine incident is recalled, and a set of cognitive probe questions is used to determine the bases for situation assessment and decision making in a non-routine incident. The technique aims to get at the subtle cues experts rely on and novices may miss in assessing a situation. The probe questions are based on Klein's Recognition-Primed Decision model. 

References
    3. PARI Method (Precursor (reason for action), Action, Result, Interpretation (of result)): A method developed primarily to capture the cognitive and behavioral demands of troubleshooting complex systems. This method is particularly suited to developing training programs. The approach consists of a structured interview in which novice and expert troubleshooters diagnose a fault in a problem scenario posed by another expert. The outputs are PARI diagrams of both expert and novice solutions to a representative set of troubleshooting problems. At each step of the solution, an Action is taken because of a Precursor, the Action produces a Result, and the Result of that Action is Interpreted. Also, at each step, problem solvers are asked to draw a diagram representing their current mental model of the state of the system. A streamlined version of this method has also been developed. 

References
    4. Skill-Based CTA Framework: A framework for conducting CTA that assumes simpler cognitive skills form the basis for more advanced skills, and it attempts to identify the hierarchy of skills needed to operate in a domain. This hierarchy, starting at the most complex skill type, includes: Strategies, Decision-Making Skills, Representational Skills, Procedural Skills, and Automated Skills. They propose that existing CTA techniques can be used to get at each of the skill types they identify. Automated skills can be analyzed by the Consistent Component Method and the Verbal Report Method (aka Think Aloud), procedural skills can be analyzed by PARI, representational skills (elements of mental models that can predict required actions) can be analyzed using Diagramming and Rating/Sorting knowledge elicitation techniques, decision making skills can be analyzed with the Critical Decision Method, Error Analysis, and Verbal Report Methods. Strategies can be analyzed with the Team Communications method and the Structured Interview method. This framework has been used to develop training programs in the Air Traffic Control domain

References
    5. Decompose, Network, and Asses (DNA) Method: DNA is a computer-based tool created to aid in knowledge elicitation and organization for instruction and training purposes, primarily to develop intelligent tutoring systems. DNA helps to decompose a domain into its constituent elements, network the elements into an inheritance hierarchy, and assess the knowledge structure for validity and reliability. DNA attempts to automate the bulk of the interview process, and it represents information gathered through the interviews in a hybrid knowledge structure that combines aspects of production systems (if-then rules related to procedural knowledge, e.g. ACT-R) and conceptual graphs.

References
    6. Task-Knowledge Structures (TKS): Task Knowledge Structures were developed by Johnson and Johnson to model conceptual, declarative, and procedural knowledge in terms of the roles, goals, objects, and actions concerned with executing work tasks. A set of methods have also been developed to collect and analyze data to create TKS models, and to use those TKS models in the design of interactive systems to support work tasks. TKS has been applied successfully to the design of graph and picture editing software, word processing software, e-mail messaging software, flight deck control systems, and intelligent tutoring systems. 

References
    7. Goal-Directed Task Analysis (GDTA): A Cognitive Task Analysis technique that focuses on uncovering the Situation Awareness (SA) requirements associated with a job. According to Endsley, SA is "the perception of the elements in the environment within a volume of space and time, the comprehension of their meaning, the projection of their status into the near future, and the prediction of how various actions will affect the fulfillment of one's goals." Thus, GDTA's emphasis is on information needs and how that information must be used to support complex decision making in dynamic, complex environments. Four steps are involved in the GDTA process: 1.) identification of key decision makers, 2.) identification of major goals and associated sub-goals for each decision maker, 3.) the primary decision needed for each sub-goal, and 4.) the SA information requirements for making those decisions and carrying out each sub-goal. 

References
    8. Cognitive Function Model (CFM): A technique developed to bridge the gap between Operator Function Models and Cognitive Task Analysis. The intent is to identify nodes in an Operator Function Model that are highly challenging cognitive tasks which should be pursued more deeply with CTA. A software application developed by Klein Associates and Aptima can be used to construct CFMs and assess cognitive complexity of nodes.

References
    9. Cognitively Oriented Task Analysis (COTA): A collection of procedures developed by Dubois and Shalin to improve assessment of job expertise and performance. COTA uses a verbal protocol analysis technique to determine the standard methods used for accomplishing tasks, how these methods are selected, initiated, and completed, and how these methods are adopted to novel situations. The knowledge representation format used is a plan-goal graph, where the overall purpose of a work activity is decomposed into sub-goals and plans for accomplishing goals.

References
    10. Hierarchical Task Analysis (HTA): A popular form of Task Analysis that breaks down a task into a hierarchy of goals and supporting sub-goals, as well as the actions performed to accomplish those goals. We include this here among the Cognitive Task Analysis techniques because the emphasis in HTA is not so much recording overt behavior but instead on the systematic decomposition of goals. 

References
    11. Interacting Cognitive Subsystems (ICS): ICS is a theory of cognition comprised of a multiprocessor architecture made up of nine subsystems related to sensory, central, and effector processing activity. The theory describes how information flows between the subsystems, how information is transformed by one subsystem into a format that can be used by another, how information is stored in memory and re-used, the hierarchical and temporal structure of mental representations, and how new knowledge is acquired. These aspects of the theory form the basis for a set of methods to conduct Cognitive Task Analysis. One method allows for the construction of a Computational Cognitive Model that can be used to predict behavior. Less formal methods based on the theory guide the construction of diagrams that describe attention flow and information flow. This technique is in its infancy and has not yet been used outside of academia. 

References
    12. Knowledge Analysis and Documentation System (KADS): A knowledge elicitation methodology developed to build knowledge-based systems (e.g. expert systems) that models knowledge in terms of generic tasks and standard organizations for declarative knowledge. Generic tasks include concepts such as classification, assessment, diagnosis, design, configuration, and planning. By identifying the category of task being analyzed, the analyst is able to make use of established patterns of knowledge. The technique elicits which performance norms are prescribed and what strategies and knowledge should be used in order to accomplish the tasks' goals. This technique has been successfully applied in the construction of expert systems.

References
    13. Team CTA Techniques: The traditional focus of CTA has been on the study of the cognitive activities and knowledge of individuals as opposed to teams. While there are few CTA methods that have been developed exclusively for analyzing the activities of teams, most of the other methods described here may also be applied to the analysis of teams. Cognitive processes that are important in the analysis of teams include: 1.) control of attention (how teams engage in information management), 2.) shared situation awareness (how members have the same interpretation of ongoing events), 3.) shared mental models (how members have the same understanding of the dynamics of key processes), 4.) applications of strategies and heuristics to make decisions, solve problems, and plan, and 5.) metacognition (how a team is able to monitor itself and determine when it is running into difficulties.

References
   
  B. Knowledge Elicitation:  Methods to determine the knowledge required to perform work tasks. These techniques provide input to cognitive task analysis activities and computational cognitive models, as well as to traditional task analysis activities.

References

 
    Interviewing/Observing Methods: Direct methods of watching workers and talking with them. These methods are well suited to the initial phases of knowledge elicitation, and can provide direction on where to focus further knowledge elicitation activities. The downside to these techniques is that they can provide large amounts of data that can be difficult to interpret.
    1. Unstructured Interviews: An open dialog in which the interviewer asks open-ended questions about the expert's knowledge and reasoning. Initial unstructured interviews allow the analyst to gain an oversight of the domain. Increasing structure can be imposed on successive interviews.

References
    2. Structured Interviews: An interview in which the range and content of the questions are carefully planned prior to meeting with experts. Each expert is asked the same question in the same order. Extra information can still be added and relevant issues pursued. 

References
    3. Step Listing: In a structured interview, the expert lists the steps involved in performing a specific task in his or her domain.

References
    4. Group Interview: A structured interview of a group of experts. Typically, group interviews are conducted if experts in the domain have differing areas of expertise or reliance on more than one expert is necessary to assess the reliability or importance of particular aspects of knowledge or reasoning.

References
    5. Questionnaires: Formalized and standardized sets of questions, which can include open-ended questions, produced to elicit a wide variety of responses. Questionnaires are useful when there is value in collecting a large number of viewpoints.

References
    6. Teachback: The expert explains a concept to the analyst. The analyst then explains this concept back to the expert. This process continues until the expert is satisfied with the analyst’s understanding of the concept.

References
    7. Field Observations/Ethnographic Methods: These are techniques that stem from the anthropology and psychology communities that study how people interact with technology. Domain practitioners are observed and interviewed in their actual work environments as they perform regular work activities. There are also a host of "rapid ethnography" methods being developed by the HCI community with the goal of providing a reasonable understanding of workers and their activities given significant time pressure and limited time in the field. These methods can be useful in gathering user requirements, understanding and developing user models, and evaluating new systems and iterating their design. 

References
    8. Twenty Questions: This technique is similar to the parlor game of the same name. The expert is provided with little or no information about a particular problem to be solved and must ask the analyst yes/no questions for information needed to solve the problem. The information the expert requests, along with the order in which it is requested, can provide the analyst with insight into the expert's problem solving strategies.

References
    Process Tracing Methods: Oftentimes workers aren't able to articulate how they perform tasks when they are interviewed in isolation from the work environment. Process tracing methods typically immerse the worker in an actual task performance or problem-solving context. The methods are used to make inferences about the cognitive processes and knowledge underlying task performance. Like interviewing and observing methods, process tracing methods often result in large data sets that can be difficult to interpret, but they can also provide a rich source for understanding.

References

    9. Discourse/Conversation/Interaction Analysis: These are techniques stemming from the linguistics and anthropology communities for analyzing group interactions using recorded conversation and actions. They examine messages exchanged among group members to discover the systematic and orderly processes by which the members communicate. They also aim to uncover standard sequences of interaction. The methods require rather detailed analysis and are time consuming to use.

References
    10. Activity Sampling: A large number of observations are made of a group of machines, processes, or workers over some period. Each observation records what is happening at that instant. The percentage of observations recorded for a particular activity enables the total time during which that activity is occurring to be predicted.

References
    11. Think-Aloud Problem-Solving/Protocol Analysis: The expert thinks aloud while actually performing some task or solving a problem. The procedure generates a protocol (a recording of the expert's deliberations, possibly including actions the expert took, what the expert was looking at, etc.) that can be transcribed and analyzed to uncover information about the expert's reasoning sequences and goal structures. Research has shown that thinking aloud does not cause much interference in problem solving performance, but there are individual differences in the level of verbal expressiveness.

References
    12. Retrospective/Aided Recall: The expert performs some task or solves a problem in an uninterrupted manner. During this process, all actions of the expert may be recorded so that the task execution can be re-played. Upon completion of the task, the expert is guided through her task behavior and asked to report on her thoughts. This technique may be used when thinking aloud while concurrently performing the task would be impractical.

References
    13. Interruption Analysis: While performing a task or solving a problem, experts are periodically interrupted with probe questions, such as "What were you just doing?" or "What would you have done just then if...?", when the analyst is having difficulty inferring the expert's thoughts.

References
    14. Shadowing Another: One expert observes the performance of another expert, either live or on tape, and provides real-time commentary (e.g. what the expert is doing, why the expert is doing it, etc.). Shadowing another is useful in process-control situations in which the expert performing the task may not have time to comment.

References
    15. Shadowing Self:  In self-shadowing, the same expert performs the task and then comments on her performance, usually while observing herself on tape. 

References
    16. Simulators/Mockups and Microworld Simulation: Equipment or information that is representative of what will be used during the task is constructed. Task activity while using this equipment is then observed and recorded. In a similar vein, Microworlds that simulate key human performance aspects of a task may be constructed to understand the expert strategies used in performing the task. We include this technique here as a means to elicit knowledge regarding task performance and problem solving, but a good deal of analysis may be required beforehand to understand key aspects of a task before construction of a simulated domain for use in this method.

References
    17. Exploratory Sequential Data Analysis (ESDA): A family of observational data (i.e. think-aloud transcripts, video transcripts, etc.) analysis techniques that aim to uncover common sequences of interactions. For example, the analyst may look for repeated sequences of keystrokes that suggest poor design. MacShapa (yes, it is currently only available for Macintosh computers) is an available software tool.

References
    18. Minimal Scenario Technique: The expert is given a problem to solve, but only the minimal amount of information required to state the problem is provided. The expert describes the domain by requesting the information that they need to solve the problem.

References
    19. Critical Incident Technique (CIT): The expert is asked to recall and discuss a specific incident that was of particular importance in some context of task performance. The goal of the CIT is to obtain a complete account of the expert’s solution plan as well as the factors that influenced the design of the plan. A risk associated with using the technique is that knowledge elicited may be idiosyncratic or atypical, so it is usually used in combination with other methods.

References
    20. Cloze Technique: Experts are presented with textual material, typically a sentence or a passage of text, in which important information has been omitted. Experts are asked to fill in the missing information according to their knowledge of the task domain. By carefully conducting a series of Cloze sessions that cover a range of scenarios or problems, the analyst can gradually develop a model of the expert's decision making rules and structures.

References
    21. Critiquing: The expert discusses both positive and negative aspects of a particular problem solving strategy in comparison to alternatives that might be reasonable or preferred. In a variation of the technique developed by Miller, Patterson, and Woods, an expert is walked through a novice's execution of a task and provides commentary. 

References
    22. Crystal Ball/Stumbling Block Technique: The expert describes a challenging assessment or decision. The analyst insists that the assessment is wrong, and that there are alternative interpretations of events, missing information, or assumptions. The expert generates these.

References
    23. Table-Top Analysis: A group of experts meets to discuss a problem perspective of the task, using task scenarios to explore the problem and derive a solution. The technique seeks to aggregate expert opinion in problem-solving mode.

References
    24. Wizard of Oz Technique: In this technique, which is also called symbolic simulation, the expert is asked to simulate the behavior of a future system. One approach is to have the expert play the role of a future expert system, responding to user’s queries.

References
    Conceptual Methods: Indirect techniques that aim to produce representations of what the relevant concepts in a domain are and how they are related. These techniques focus on concepts and their relations rather than heuristics, rules, and strategies. It is generally recommended that these techniques be combined with process tracing methods to gain a more thorough understanding of a domain.
    25. Decision Analysis: A set of procedures that involve utility (i.e. the worth associated with a particular event) and probability (i.e. the likelihood of a particular event) modeling based on inputs provided by the expert. Such knowledge can be used to build mathematical models of reasoning, decision trees, inference networks, decision aids, and expert systems.

References
    26. Rating and Sorting Tasks: In rating tasks, experts are asked to rate domain elements and provide the basis for those ratings. For example, they may be asked to judge the accident rate of a particular road. Rating tasks can be used to highlight differences between expert and novice classifications of domain elements. The sorting task is a variation of the rating task in which experts sort domain elements into various categories. In the typical sorting task, participants sort cards bearing the names of domain elements. They are then asked to provide a label for the various piles according to particular features or dimensions.

References
    27. Magnitude Estimation: Magnitude estimation is a rating task in which experts are asked to assign a magnitude to a particular concept along some scale. For example, an expert may be asked to assign a severity rating to a particular crime. The technique is often combined with other techniques, such as the Repertory Grid Method or Card Sorting, which elicit attributes that are then scaled with Magnitude Estimation.

References
    28. Repertory Grid Technique: A rating task in which experts are asked to generate important aspects of a domain and then provide dimensions on which those aspects may be rated. For example, a medical expert may be asked to generate a list of different types of diseases and then rate each disease on a number of different dimensions (i.e. the symptoms and features of the disease). The technique is based on Personal Construct Theory in which concepts or elements are identified and rated along dichotomous dimensions, referred to as constructs.

References
    29. P Sort: A sorting technique in which the expert sorts domain concepts into a fixed number of categories with limitations on the number of concepts per category.

References
    30. Q Sort: A sorting technique in which the analyst presents the expert with domain concepts which are then sorted into piles base on relatedness.

References
    31. Hierarchical Sort: A variation of the sorting task in which the expert sorts concepts in an increasing number of piles on each pass.

References
    32. Cluster Analysis: Cluster analysis is a mathematical technique for forming "clusters" of related concepts. Clusters are groups of concepts in which the similarity is high for concepts within a group, and low for concepts between groups. Cluster analysis is typically combined with an elicitation technique that can be used to produce similarity measures between concepts, such as the Repertory Grid Method.

References
    33. Multidimensional Scaling (MDS): The analyst presents all combinations of pairs of concepts from the domain to an expert who is asked to rate them in terms of proximity or similarity. This data is used to create a k-dimensional space in which the concepts are located.

References
    34. Likert Scale Elicitation: Experts are presented with assertions regarding some aspect of the domain, such as "The Corps Commander does not need to know how the most probable enemy avenue of approach is determined." They are then asked to indicate their degree of agreement on a Likert scale, as well as offer comments on the assertion. Likert scales are typically 5-point rating scales ranging from "Strongly Agree" through "Neither Agree nor Disagree" to "Strongly Disagree."

References
    35. Structural Analysis Techniques: The analyst uses a mathematical algorithm to transform the expert’s concept relatedness measures (which may come from a technique such as Repertory Grid or Multidimensional Scaling) into a graphical representation of the domain. Knowledge Network Organizing Tool (KNOT) is an available software tool.

References
    36. Conceptual Graph Construction: In conceptual graph construction, which is also called concept mapping or semantic network elicitation, the expert and analyst work together to draw a graphical representation of the domain in terms of the relationships (links in the graph) between the domain elements (nodes in the graph). Conceptual graph construction can be used to provide an expert's cognitive organization of a task. Robert Hoffman and his colleagues at the Institute for Human and Machine Cognition (IHMC) are currently researching the use of concept maps to represent and share expert knowledge. A software tool to assist in concept map construction is also available for download from IHMC.

References
    37. Diagramming: The expert draws pictures of the domain's basic constructs and their interrelationships. The diagram might be a map showing spatial-relationships, a flowchart showing temporal relationships, a network or tree diagram showing hierarchical relationships, or a system state/action diagram.

References
    38. Laddering: A diagramming technique in which the analyst asks the expert questions to systematically build a hierarchy of domain concepts. The technique begins with the analyst stating the name of a "seed item" from the task domain. Different questions are used to lead the expert through the task domain/hierarchy. This technique is useful to apply when the domain constructs are known but the interrelationships between them are poorly understood.

References
    39. Influence Diagram Construction: Target events (e.g. system failures) are defined prior to using the technique, and then the analyst produces a directed graph representing the influences (e.g. procedure adequacy, etc.) that determine the outcome (success or failure) of each event, together with any dependencies between them. The technique is often used in human reliability assessment with systems that have been operational for some time, but there is no empirical data available regarding human error probability. Available software tools are Analytica and DEMOS. 

References
   
  C. Computational Cognitive Modeling: Methods that produce detailed models of how humans perform complex cognitive tasks that can be run on a computer. Such models can provide a priori performance predictions of how well a certain system will support the tasks workers perform by assessing factors such as how easy the system will be to learn and use, the workload it imposes, and the propensity for errors. Software agents that perform work tasks in the same way that humans perform work tasks can be used to evaluate proposed system designs without the need to conduct these types of evaluations with actual workers. The downside to these methods is that they can be time-consuming and, with the exception of GOMS, have not yet seen widespread use in the engineering community.

References
    GOMS Family of Models: GOMS (Goals, Operators, Methods, Selection rules) is a task modeling language intended to be used by practitioners in the HCI domain. This methodology can be used to define how tasks are organized by an expert worker in order to produce detailed models of human-computer interaction. Such models can be used to make predictions about how well humans will be able to perform a task with a proposed design. These predictions can be used as a substitute for data generated from usability studies, making it possible to iterate through design revisions and evaluations more rapidly early in the design process. Actual user testing will still be necessary to uncover usability problems not specifically addressed through GOMS modeling.

In the methodology, Goals are an end state that must be achieved to accomplish a task. Operators are the task actions that must be performed to attain a goal or sub-goal. Methods are sequences of operators used to accomplish a specific goal. Selection rules are sets of conditions or decision rules that are used to determine which specific method should be used to accomplish a goal if more than one method is applicable. 

The GOMS approach is limited in that it can only describe tasks that rely exclusively on procedural ("how-to-do-it") knowledge and routine cognitive skill. It should also be noted that a Task Analysis to determine worker goals (such as Hierarchical Task Analysis) must be conducted first so that those goals can be expressed in a GOMS model of how the worker can accomplish them with a proposed system. Variations of the GOMS approach are described below.

References
    1. Keystroke Level Model (KLM): KLM is simplified version of GOMS that only allows for a task to be modeled as a specific sequence of keystroke-level primitive operators. There are six types of operators: K to press a key or button, P to point with a mouse to a target on a display, H to home hands on the keyboard, D to draw a line segment on a grid, M to mentally prepare to do an action, and R to symbolize the system response time during which the user must wait.

References
    2. CMN-GOMS (Card Moran Newell GOMS): CMN-GOMS is a slightly more detailed version of GOMS that describes a task in terms of a hierarchical goal structure and set of methods in program form, each of which consists of a series of steps executed in a strictly sequential manner.

References
    3. NGOMSL (Natural GOMS Language): NGOMSL refines the CMN-GOMS model by connecting it to a simple cognitive architecture. The technique provides a well-defined, structured natural language to describe tasks in practical applications. The technique also defines explicit procedures for constructing GOMS models. NGOMSL models can predict the time it takes to learn procedures as well as their execution times. GLEAN3 is a computer-based tool that generates quantitative predictions of human performance from a supplied GOMS model.

References
    4. CPM-GOMS (Critical Path Method/Cognitive-Perceptual-Motor GOMS): CPM-GOMS is a task analysis technique based directly on a parallel multi-processor stage model of human information processing. CPM-GOMS uses a schedule chart (Pert chart) to represent the operators and dependencies between operators. The critical path through a schedule chart provides a simple prediction of total task time. 

References
    5. CAT (Cognitive Analysis Tool): CAT is a software application that automates a practical cognitive task analysis technique developed by Kieras for use in the human-computer interaction domain. CAT uses a structured interview process to elicit from workers descriptions of how they would perform a task, and it can structure those descriptions as production rules. These production rules form the basis of GOMS models that can be used to generate detailed predictions of task execution time using a proposed interface. CAT-HCI provides extensions to CAT that allow for the elicitation of more low-level perceptual, motor, and cognitive operations that can improve the predictive accuracy of resulting GOMS models. 

References
    Integrated Cognitive Architectures: Integrated cognitive architectures allow for the memory, cognition, motor behavior, sensation, and perception necessary to perform complex tasks to be modeled at a high level of detail. It is the analyst's job to describe a task or problem solving activity using the modeling framework provided by the architecture. 

References
    6. COGNET (COgnition as a NEtwork of Tasks): COGNET is a theoretically-based set of tools and techniques for performing Cognitive Task Analyses and using those analyses to build models of human-computer interaction in real-time, multi-tasking environments. COGNET provides a language for describing cognitive tasks in information processing terms based heavily on the GOMS notation, but includes additional features to allow for the modeling of cognitive operators. COGNET models may be implemented using the iGEN software system, available from CHI Systems. COGNET has been used to model a number of tactical decision making tasks.

References
    7. COGENT (Cognitive Objects within a Graphical EnviroNmenT): COGENT is a software environment for developing cognitive models. It provides a visual programming interface to allow users to build cognitive models using memory buffers, rule-based processes (similar to production rules), connectionist networks, and I/O sources and sinks. It is primarily intended for the cognitive science research community, and has been used to model a variety of decision making tasks in laboratory settings. The software is freely available.

References
    8. ACT-R (Adaptive Control of Thought – Rational): ACT-R is a unified theory of cognition that allows for the declarative and procedural knowledge necessary to perform a wide variety of cognitive tasks to be modeled with a high level of detail. Declarative knowledge (i.e. factual knowledge) is represented as chunks of information, and procedural knowledge (i.e. how-to-do-it knowledge) is represented as if-then like rules called productions that direct behavior. ACT-R/PM includes perceptual and motor extensions based on those implemented in EPIC. ACT-R has primarily been used by the cognitive science community to model individuals performing complex problem solving tasks in laboratory settings, but current projects have used ACT-R to model submarine commanders, air traffic controllers, and users of computer interfaces.

References
    9. Soar: Soar is a unified theory of cognition that implements goal-oriented behavior as a search through a problem space and learns the results of its problem solving. It has been used by the artificial intelligence community to build artificial agents to solve problems as efficiently as possible in real-time environments, as well as by the cognitive science community to understand human behavior.

References
    10. EPIC (Executive-Process Interactive Control): EPIC is a symbolic cognitive architecture that emphasizes the modeling of perceptual and motor processes, but also includes a simple production rule cognitive processor for modeling task performance. It is particularly suited to modeling multiple-task performance. It is intended to be used to generate predictions of human performance for use in system design, training, and personnel selection. 

References
    11. Apex: Apex is a software architecture for modeling human performance in complex, dynamic environments. It is intended to be used by people without a great deal of expertise in cognitive modeling, but it nonetheless allows for modeling behaviors ranging from perceptual motor actions to intelligent management of multiple, long-duration tasks. The language used to describe cognition is PDL, which incorporates and extends the capabilities provided by GOMS. Intended applications for Apex include time-analysis of skilled behavior, partially-automated human factors design analysis, and creation of artificial human participants in large scale simulations. It is freely available from NASA.

References
    12. MIDAS (Man Machine Integrated Design and Analysis System): MIDAS was developed by NASA to model pilot behavior. It consists of an agent-based operator model, with modules for representing perceptual, cognitive, and motor processing, as well as a model of the proximal environment (e.g. displays and controls), and the distal environment (e.g. other aircraft, terrain). The cognitive module of MIDAS includes if-then rules, daemons to monitor changes in the world and perform designated operations when such changes are detected, and generalized decision algorithms such as weighted additive and majority of confirming decisions.

References
    13. SAMPLE (Situation Awareness Model for Pilot-in-the-Loop Evaluation): SAMPLE is a stage-based, situation awareness-centered cognitive architecture that decomposes decision making into information processing, situation assessment, and procedure execution. SAMPLE assumes that the behavior of a crew (or an individual operator) is guided by highly-structured standard procedures and driven by detected events and assessed situations. It was developed to model situation assessment for combat pilots, but has also been used to evaluate landing procedures of commercial aircraft, nuclear power plant control and automation, decision aiding, air traffic alerting systems, and supervisory control of Uninhabited Combat Aerial Vehicles (UCAVs).

References
    14. OMAR (Operator Model ARchitecture): OMAR models human operators in complex systems, such as command and control systems, aircraft, and air traffic control. It has been used to evaluate operator procedures in order to evaluate system design, particularly the design of displays and controls. OMAR assumes that human behavior is goal-directed, and agents in OMAR are capable of executing goals, plans, and tasks. OMAR also has provisions for executing multiple tasks concurrently, as well as modeling multiple, communicating operators using D-OMAR (Distributed Operator Model Architecture). 

References