+ Context as a Cognitive Process: An Integrative Framework for Supporting Decision Making Wayne Zachary, Ph.D., Andrew Rosoff Lynn Miller, Ph.D., Stephen Read, Ph.D. CHI Systems, Inc. University of Southern California 2250 Hickory Road, Suite 150 Los Angeles, CA, 90089, USA Plymouth Meeting, PA, 19462, USA Read@usc.edu lmiller@usc.edu Wzachary@chisystems.com ARosoff@chisystems.com Abstract— Multiple lines of research in cognitive science have found that the content and organization of an internal brought insight on the role that internal (cognitive) representation of the problem instance differentiated the representations of situational context play in framing decision performance of skilled decision makers (DMs) from less- making and in differentiating expert versus novice decision skilled ones. Specifically, these separate lines of research performance. However, no single framework has emerged to pointed to the fact that expert DMs – across domains – use integrate these lines of research, particularly the views from internal representations of the problem instance in its narrative reasoning research and those from situation awareness environmental setting that are richer and more stylized, and recognition-primed decision research. The integrative incorporate multiple levels of abstraction, and take on a struc- framework presented here focuses on the cognitive processes ture that enables rapid retrieval of relevant decision-making involved in developing and maintaining context understanding, rather than on the content of the context representation at any heuristics and procedures. This latter feature became widely given moment. The Narratively-Integrated Multilevel (NIM) known as recognition-primed decision-making or RPD [14]. framework views context development as an on-going and self- In the 1990s, research on the structure of mental models of organizing process in which a set of knowledge elements, rooted context across domains began to suggest that there is in individual experience and expertise, construct and maintain a consistent, hierarchical structure to (expert) mental models. In declarative, hierarchical representation of the situational context. particular, the work of Endlsey [5,6] developed a theory of the The context representation that arises from this process is then general structure of expert-level context mental models across shown to be the central point of both situational interpretation dynamic, real-time domains. Terming the understanding of the and decision-making processes at multiple levels, from achieving specific local goals to pursuing broad motives in a domain or changing external context as Situation Awareness (SA), theater of action. Endsley identified three increasingly abstract levels: 1. Perception, in which the person perceives the status, Keywords— situational awareness; recognition-primed decision attributes, and dynamics of relevant elements in the making; narrative reasoning; self-organizing architecture; decision support systems situation and their current states, 2. Comprehension, in which the person understands how the I. INTRODUCTION perceived elements can impact situational goals; and, The current scientific understanding of the role of context 3. Projection, in which the person can project the future in decision-making has evolved in multiple steps over the last actions of the elements in the environment forward in time. forty years. Cognitive science research has long shown that There is an explicitly constructive assumption about these while human actions and decisions are based on the person’s levels, in that level 1 information is represented from environmental context, the decision-making process relies on information directly perceived from the environment, level 2 an internal (cognitive) representation of the context, not information is constructed mentally from level 1, and level 3 directly on the context as sensed (see [3] for a succinct review information is mentally constructed from Level 1 and Level 2 of this literature]. In the 1980s, convergent research on: information. SA and RPD theory have led to the development • the study of decision making in its naturalistic setting of various decision support applications [9,11,18,20]. rather than in laboratory experiments [12,13]; While this thread of cognitive research was building an understanding of the role of context from the bottom-up (i.e., • cognitive skill acquisition theory [31,34]; and building from fundamental insights on human information • mental models in cognition, e.g.,[36] processing mechanisms), a separate thread of ‘top-down’ cognitive research unfolded from the 1980s forward. This thread explored how people understand and reason about Parts of this research were supported by Contract F33615-02-C-6033 from the US Navy. The remainder was sponsored by CHI Systems, Inc. internal research and development. STIDS 2013 Proceedings Page 48 sequences of action and interaction in which the main source of to represent or reproduce the higher-level complexities of variability is human behavior. (This aspect is particularly human social behavior and social intelligence. More recently, germane to military decision-making, in that it typically the research team focused on developing a cognitive involves situations with both adversaries and non-combatants). architecture called PAC, based on narrative reasoning and This research focused on narrative reasoning processes in cognitive theories of personality [24,25,33]. While PAC which the observer/participant constructs, analyzes, and proved able to model and predict complex interpersonal explains complex situations through a narrative (story-telling) behavior in off-line simulations, the translation of this to real- process. Specifically, it found that people almost universally time situations proved daunting. Specifically, it became clear use story narratives to represent, reason about, and make sense that to carry out narrative reasoning in real-time, the narrative of contexts involving multiple interacting agents, using reasoning knowledge elements required access to a dynamic, (general) motivations and (local) goals to explain both and more detailed, representation of the changing observed and possible future actions. In other words, people understanding of the problem context at lower levels of were found to generally make sense of their human contexts by abstraction. This required, in the end, adding much of the either integrating them into a novel narrative or, more SA/RPD mechanisms for building context from COGNET into commonly, by recounting them as an instance of a commonly- the narrative-based mechanisms in PAC. The addition of these known or culturally based narrative [4,10,26,28]. There is also mechanisms fell far short of true integration, however, in that a evidence that people maintain narrative structures mentally and common theoretical framework for such an integration was use them to identify, assess, and select behavioral options – lacking. The framework described below was developed to that is, to support decision-making [27,28]. These ideas have meet this need. been widely applied, for example in criminal investigations [1], legal decision-making [21,22], policy analysis and formation A. Framework for Integration [37], and in social interactions [17]. The main idea underlying this integration is that what SA/RPD and narrative reasoning theories implicitly or Despite their convergent directions, the bottom-up SA/RPD explicitly refer to as the understanding or awareness of context theories and the top down narrative reasoning theories have not is really a momentary “snapshot” of fundamental processes yet met. This paper presents a framework in which such an integrating multiple sources of information about the natural integration can occur, and explores its benefits for decision and human (i.e., social) aspects of the environment. This support and human-machine integration. process of context development is constructive, self-organizing, II. CONTEXT AS INTEGRATED PROCESS operates at multiple discrete levels of abstraction which generally involves increasing time-scales across levels. These This failure of the two theories to integrate immediately four key features are defined as follows: points out several unmet challenges for decision support. For example, changing patterns within SA do not, by themselves, • Constructive -- consists of constituent elements that, present the DM with any easy way to see alternative narrative through their interaction, build a symbolic representation, interpretations for the context dynamics (making DMs more the momentary content of which we may consciously vulnerable to deception). SA theory and RPD theory have recognize as the current context. worked best in contexts that involve well-defined problem- • Self-organizing -- the constituent elements operate solving in bounded problem domains, such as putting out fires independently but follow principles or rules of operations [15], piloting aircraft [7], and controlling complex mechanical that are organic to the human information processing systems [8]. Even though they have successfully been design, such that a consistent and self-regulating process automated as cognitive models and used for training and (of context development) emerges. advisory purposes, the upper levels of context in SA theory do not yet articulate with the narrative level of context • Operates at multiple-discrete levels of abstraction -- the representation (and the reasoning processes associated with symbolic representation which is built and maintained has that level). At the same time, decisions made at a narrative distinct layers of structure which reflect levels of level are not easily instantiated into action specifics without understanding that each incorporate a broader scope of direct access to the more detailed understanding of situational information about the environment but in correspondingly details available at the lower levels of the framework. For this increasingly abstract terms that include salient and reason, narrative reasoning has proven most useful in diagnostic attributes, with links to lower levels of applications that involve non-real-time sense-making (e.g., [1, abstraction where more detailed (but less integrated) 21,22]). information is maintained. These levels equally organize the constituent processing elements that build the context The authors and colleagues have conducted a line of representation as much as they organize the representation research to develop and apply computational models of expert itself. In this initial formulation of the framework, there cognition in various domains, both to test and refine cognitive are four levels corresponding to the three hierarchical theory and to develop support for decision making and decision levels of Situation Awareness (Perception, Comprehen- training. That research initially focused on operationalizing the sion, Projection) and one higher level of Narrative SA/RPD body of theory, and resulted in a computational Understanding which integrates the other three. We thus architecture called COGNET [35]. While this architecture call the framework the NIM (Narratively-Integrated proved successful in modeling human performance in work- Multilevel framework. tasks, it became clear that the model and behavior were unable STIDS 2013 Proceedings Page 49 • Involves increasing time-scales across levels -- each individual a constantly available basis for interacting with the increasing level of abstraction deals with a broader scope environment. The representation-building knowledge elements of events (from perceptual events at the lowest level to that construct the context representation reflect both narrative units at the highest level). As that scope individually acquired expertise and culturally-transmitted increases, the general time-scale of events similarly understanding of the local or domain-specific environment, so increases. For example, perceptual events, such as those the context representation is not only always available, but also tracking locations of a (single) moving object, are very encodes information that experience (individual and collective) dense in time and result in repeated updates to perceptual has shown to be useful in those environmental interactions. level information in the context representation. At higher Ultimately, it is through its ability to support effective actions levels, updates typically occur less frequently, as many and interactions in the natural and social environment that the lower level changes are needed to create a significant or value of the context process is realized. meaningful update. Narrative pacing, the highest level, typically is the slowest, as a great deal of action in the environment is typically aggregated into a single narrative Representation-building Declarative Context Knowledge Elements Representation unit. This relationship of increasing time scale and increasing scope is very similar to the concepts presented Narrative level Build Understanding -main (presumed) narrative in Newell’s timescale of human action [19: Figure 3-3]. downward -alternative (plausible) narrative Thus, the amount of processing would tend to be much (deduction) Build -current story unit, future options, implications for each plausible greater at lower levels, though the scope and usefulness of Understanding narrative upward the information in the representation would tend to be (induction) Projection Level much broader at higher levels. However, because of the -object near-future dynamics -object change indicators constructiveness feature, the highest level cannot be -time-base possible futures constructed without all the processing involved in building Build Comprehension Level -object types, meanings and maintaining the lower levels. Understanding -object groupings, associations within Levels -object capabilities The dynamics of the process are moved forward both by Perception Level sensory information (on the external world), physical actions Bring Information -objects, tracks into Context -locations (taken in the external world), and internal sources of Representation -the attributes information that can be termed knowledge elements. In the NIM framework, the context representation is constantly being Situational data Stream manipulated in different ways by knowledge elements (KEs) that themselves are activated by externalities (in the form of Figure 1. Context Development Process sensations and/or physical actions), or by internalities (in the form of patterns of information within the declarative III. CONTEXT AND DECISION-MAKING representation or associations to past experiences). Thus, the Research into decision-making has explored some of the various knowledge elements construct and maintain the context ways in which the context representation supports decision- representation in a self-organizing way, without any explicit making. The RPD model, most specifically, has demonstrated starting or stopping (or other control) mechanism. that expert DMs are in many cases able to select an action or B. Computational View of the NIM Framework adapt a pre-existing action plan to a specific situation based on the patterns of information in the context model. The patterns As a process, context development is an example of, and of information prime a specific decision (course of action) can be computationally modeled using, Selfridge’s without requiring intervening deliberative processes. More Pandemonium architecture [29], which has been highly analytical decision processes, in contrast, involved multiple influential in many branches of cognitive science and artificial reasoning steps that manipulate the context representation to intelligence over the last half century. In a Pandemonium-style construct, rather than derive, a plan or specific action. Across model of the context process, a hierarchical declarative this full continuum of analytical to automatized decision representation of context is the central feature, and elements making, (often called the Cognitive Continuum, see [38]) the (chunks) of knowledge are spontaneously activated (and same process is occurring. Knowledge elements derive or compete for attention) by patterns of information and dynamic construct decision options and courses of action by changes to this declarative representation. Each element of manipulating and operating on the information in the context knowledge changes the declarative context representation representation. These can be called decision-development KEs. (making it a representation-building knowledge element), either by creating new information, or by adding, replacing or In light of the above discussion on context development, deleting information, At any point in time, the DMs the decision-development KEs can be seen as are analogous, to understanding of the context is the current content of the representation building KEs. Both use the information in the declarative context knowledge structure. The context context representation, but the representation-building KEs use development process is pictured in Figure 1. it to create changes to the context representation, while the decision development KEs instead use it to reason toward It can be argued that a background process that develops actions to be taken in the external environment. and maintains an understanding of context is a highly adaptive characteristic of human beings, because it provides the To some degree, the preceding begs the question “what is STIDS 2013 Proceedings Page 50 decision-making?” For purposes here, decision is used broadly complexity of events greatly outstrip human cognitive abilities. to refer to the processes by which purposive actions are Real-time decision-making domains such as military command selected or constructed, whether or not there is a conscious and control or management of large-scale industrial processes awareness at the time that a decision is being made. This is bring environments in which it is essentially impossible for an broadly in line with RPD theory which notes that the RPD unaided DM to fully understand the context in which actions process typically renders what appears, to a novice or outsider, must be taken. to be a difficult decision, as simply an obvious or automatic The preceding half century has seen increasingly action to the expert. sophisticated efforts to support and augment human decision- One additional feature needs to be added to the NIM making. Research to understand human cognition has been framework to describe or model the relationship of the context- stimulated by the need for more effective decision-support, and development process to the decision-making process. That is has driven the evolution of decision support. In particular, it the notion of hypothesizing – constructing and manipulating has resulted in an approach (termed cognitive engineering) to alternative descriptions or relationship sets for part or all of a designing decision support systems, based on designing the context representation, typically by creating hypothesized systems to integrate well with the ways in which humans representations of future contexts that might result from perceive, think, and act. contemplated decisions or actions. For context to support decision making, there needs to be proxy representations of The NIM context-development view offers a new basis for cognitive engineering of decision support systems. The context, in which decision-development KEs can use to construct and assess potential decisions and actions. This framework shows how multiple levels of context understanding are simultaneously developed and maintained, space, unlike the context representation, is not an internal model of the external situation, but is rather a hypothesized and are also simultaneously used to identify opportunities for action and for action options. This suggests a way to design representation of it as it might be, if potential decisions and actions were taken. This allows such decision-development decision support, in which the support system develops its own context representation (based on a model of the human context- KEs to maintain alternative multi-level representations of an evolving situation, or project forward possible decisions or development process), and applies this model to develop decision/action information at multiple levels of abstraction. actions based on a narrative interpretation or course of action being considered. Figure 2 expands Figure 1 to show how Further, such a system can both provide its context representation to the DM as representational support, and decision-development KEs and hypothetical context representations extend the context development process to provide its decision/action information to the DM as decision support. Because it is expressed in fundamentally computa- support dynamic decision-making of all kinds. tional terms, the NIM framework suggests a way to develop the Maintain Parallel context and decision models that such a support system would interpretations require. Build Understanding Narrative level -main (presumed) narrative Before providing a brief example of how this might work, downward -alternative (plausible) narrative we note two other interesting characteristics of the NIM Hypothetical Context Content (deduction) Build -current story unit, future options, Understanding implications for each plausible Identify Possible framework with regard to the application areas of interest to narrative upward (induction) Projection Level Future evolutions this conference. The first is in the area of human-machine from Plausible -object near-future dynamics -object change indicators Narrative integration. Substantial research and engineering effort has -time-base possible futures Comprehension Level been devoted to automating the process by which a human Build Understanding -object types, meanings Identify Possible operates a continuous-control system, such as a vehicle or -object groupings, associations within Levels -object capabilities Action/Plan power plant. In between manual control and full automation, Instances from Bring Information Perception Level -objects, tracks Specific Narrative however, are many approaches to partial automation that into Context Representation -locations -the attributes structure the engineering space. All generally fall under the concept of supervisory control (originated by Sheridan and Situational data Stream Johannsen, [30]). In supervisory control, many or all the functions of manual control are automated within a space of Fiure 2. Context Processes Supporting Decision Processes options or assumptions. The human may turn over control to IV. CONTEXT AND DECISION SUPPORT those automated functions to free time and attention for other activities, but only while supervising the automation for The cognitive process of context development and changes in the underlying options or controls. When such maintenance is common to all human adults, just as is the changes occur, the operator will need to either resume manual process by which context understanding is used to make control and/or modify the settings on the automation. The decisions and construct actions in the external environment. autopilot on a manned aircraft is an example of this process. The environments in which these human capacities evolved Supervisory control is a human-machine integration concept, were relatively bounded and unfolded in time scales generally because it frames how the interconnection between human and in line with human information processing. However, this automated system components is engineered. If a system began to change in historical times, as social and technological allows only supervisory control, then it can be labeled as complexity rapidly increased. Since the start of the electronics having pure supervisory control. If, however, the human can and computer age, human DMs find themselves increasingly assume direct control as well as supervisory control then the embedded in complex environments in which the speed and system can be said to have mixed mode control. NIM context STIDS 2013 Proceedings Page 51 development allows control processes to be framed and The figure has a left-to-right structure as well. In the center of embedded within it. This can be done by considering control the figure is a box that represents the dynamic context to be a continuous analog of (discrete) decision-making, and development process, as it would be performed by a mapping the forms of control to the level of abstraction on computational model. That box is divided into two columns, which they rely in the context representation. Manual control, with the left depicting the various levels of context for example, involves context understanding largely at the representation, and the right representing the corresponding perceptual level and significance levels. Supervisory control, in representations constructed to develop decision and action contrast, involves context understanding at the significance and plans from the context representation. These two columns projection level. Control at the highest levels of abstraction are correspond to the Observe/Orient and Decide/Act phases of the not widely discussed in the human-machine integration OODA loop. literature, but they could be described as situational control or On the immediate left of the context development box is a narrative control, in which control is only applied to choice of narrative interpretation and choice of narrative units, with all column that represents the representation-building KEs. These KEs both dynamically build/maintain the context lower level control being automated. This relationship is pictured in Figure 3, discussed below. representation, and push information to the next (on the left) column as support for the human DM’s understanding of the The second is an interesting correspondence between the context. On the immediate right of the context-development context development NIM view of context development and box is a column that represents decision-development KEs that military models of decision making, particularly the military dynamically build/maintain representation of decisions and decision making model known as the Observe-Orient-Decide- actions based on current context dynamics, and that push Act or OODA Loop, first created by Boyd in the 1980s [2,23]. information to the next (on the right) column as support for the It teaches military DMs to view decision-making as an on- human DM’s selection and instantiation of action options. going process, in which situational understanding, achieved by Thus, the entire left side of the figure represents support for the careful observation (Observe) and interpretation (Orient), lead OO parts of the OODA loop, while the entire right side to courses of action (Decide) that are implemented and have represents the support for the DA parts. effects on the situation (Act). These effects then change the Below the lowest level of context is a black bar that situation (as do actions of the opponent and other non-combat processes), requiring a new or ongoing process of observation represents the environmental interfaces decision system (human augmented by context-driven support). In the case of and interpretation. In addition to it being widely used in military education and doctrine development, the four multi-UxV command and control, these environmental interfaces would be with various sensors and information components of the loop map very closely to the ways in which streams from the UxVs being controlled. context information is used in the NIM framework. That is, the activities of the: In Figure 3, the context-development process builds upward from perceiving basic situational information (Level 1) • representation-building KEs that effectively import sensed through identifying the significance of the elements (Level 2) information into the context representation corresponds to and projecting the capabilities of key elements forward into the the Observe stage; future (Level 3). From that, the lower level information is fit • representation-building KEs that integrate context into stories and understood in the context of the narrative of the information and build context understanding through and current mission (Level 4). The right-most column of Figure 3 across levels corresponds to the Orient stage; then depicts the reasoning activities that the context-based decision support model is performing to take action in the • decision-development KEs that identify potential courses environment and accomplish the mission. At the highest level, of action corresponds to the Decide stage; and the model may revise or refine the current story narrative, and • decision-development KEs that construct the details of update it in terms of his/her evolving lower level context action plans and physically implement those plans maps to understanding. As the action proceeds to the point that a the Act stage. choice must be made between possible ‘next’ narrative units, the model makes use of the current context to choose a possible V. A CONCEPTUAL EXAMPLE path forward (through the current narrative space), and conveys it to the human DM. If the DM concurs, the model could A notional example is provided below to demonstrate the translate that general narrative step into specific local action potential application of the NIM framework. The example plans (e.g., creating new waypoints, altitude, sensor-settings, focuses on the management and control of multiple etc.). uninhabited vehicles (UxVs). Such groups of vehicles can be used in diverse missions ranging from post-disaster search and Additional detail can be seen by more closely examining rescue, to battlefield intelligence collection and tactical the two columns labeled “Observe/Orient” and “Representation interdiction. The framework was used to map out the context Building KEs” from bottom to top. Figure 3 shows that the: process in this domain, and to link it to support for both the Observe/Orient stages of the OODA loop and the Decide/Act • Object representations of information from sensors and/or stages. The result is pictured in Figure 3. data streams are created as the lowest levels of context information, using sensory KEs (e.g., monitoring sensor The figure is organized top-to-bottom into four stacked feeds looking for new data, which are then processed to bands that represent the four levels of context representation. create a new track object or update an existing one). STIDS 2013 Proceedings Page 52 STIDS 2013 Proceedings Page 53 • Declarative context representation is built and updated decision. When the decision maker is operating within a class from the primitive object representations by perceptual of situations whose structure he/she understands very well, her KEs that construct a multi-level structure with built-in or his internal context model will be rich and organized at semantic significance regarding the levels; information is multiple interconnected levels of abstraction. Such a NIM created and modified as elements of meaning are inferred context representation provides insights at each level of or created for them. Initially, the perceptual KEs look for abstraction – from low-level immediate details to long-term information with specific kinds of meaning (e.g. high-level story-structures – and enables mechanisms that waypoints, vehicle locations) and place them in the allow situational interpretations and decision options to be context structure. considered at each level in an integrated way. A key implication of this research is that any externally provided (i.e., • Context representation updates happen continuously as computational) decision support information will be evaluated situational KEs combine information from multiple places and considered by the decision maker in terms of his/her own in the context representation. For example the appearance internal context understanding. Thus, from a cognitive of a hostile radar emission (created by a perceptual KE) engineering perspective, any and all decision support might trigger a situational KE to examine all UxV tracks components, algorithms, etc., should present their results in and infer which one(s) may have been detected, and to add terms of the decisions makers’ context model, and should a ‘likely detection by hostile’ annotation to that UAVs ideally be designed to be presented in such terms from the information on the context representation. That start. As implied here, one way in which this can be done is changemay, in turn, trigger another KE to add a ‘need to for the computational decision support system to build and evaluate’ annotation on the track to stimulate examination maintain its own context representation, strongly modeled to of its altitude or flight path. mimic the context representations created and maintained by • Narrative updates happen as changes in the dynamic expert decision-makers in the domain. content trigger Dynamic Narrative KEs to offer evidence In conclusion, we offer thoughts on the validation of the on whether the narrative may have changed from one NIM framework, and the ways in which semantic technologies narrative state to another. For example, the preceding can be used to implement the NIM framework. hostile radar may, if expected, activate a Dynamic KE to post evidence that the narrative may have moved from an Validation. The difficulties of validating models of ‘ingress’ phase to an ‘in hostile airspace’ narrative state. cognitive processes, which are inherently unobservable, are well discussed in the literature. Validation in cognitive science • Finally, Narrative Space Update and Narrative is, in philosophy of science terms, typically limited to Interpretation are made as Narrative Update KEs weigh standards of sufficiency (i.e., can a model explain all the data) evidence for and against a transition across narrative units. rather than necessity (i.e., only that model do so). Prolonged If posted evidence outweighs posted counter-evidence validation studies for very fundamental constructs (such as above a threshold, then a Narrative Update KE may be short term and working memory, see [3]) have be approached triggered to update the story narrative to reflect that with experimental studies, but, even there, competing models narrative-state transition. Other Narrative Update KEs can remain even after decades of experimentation. For higher level be triggered by very anomalous information that may models of cognitive processes that are not biological but rather activate narrative re-examination. For example, if the which emerge from embodied experience in the world (such as story narrative were about a reconnaissance in a the NIM framework for context), the validation problem is that demilitarized area, the presence of the sudden hostile radar much more difficult. Ultimately, we believe that validity can detection may trigger a Narrative Update KE that would be locally approached with specific domains and specific look for other narratives that might incorporate this fact populations of decision makers, using established cognitive which does not ‘make sense’ in the baseline narrative. science data collection methods such as thinking aloud data That KE might suggest re-examination of the data against collection, situationally-adapted verbal probes, and stories of outbreak of hostilities or new insurgent activity retrospective interviews. Through such domain-based as alternative stories. explorations, incremental local validation may be achieved, This example is intended to point out how the NIM view of which may lead to broader acceptance over time. context development as an ongoing and core cognitive process Semantic Technologies and NIM Implementation. can act as an integrating element for advanced decision support Semantic technologies (the topic of this conference) can form systems. Moreover, the example suggests how the framework the core of a computational system that implements a domain- can be further applied to integrate the design of human-systems specific model using the NIM framework. In fact, initial integration and to translate the cognitive and technological efforts to date have made increasing use of these, particularly issues into widely accepted military concepts such as OODA the Resource Description Framework (RDF) semantic that can support the transition of such advanced decision representation. While the earlier COGNET software used a support systems into operational use. custom-coded blackboard representation to create the lower three levels of the NIM declarative context representation, the VI. DISCUSSION most recent versions of the PAC software have moved toward This NIM framework presented here is built on the premise a implementing the declarative context representation fully in that human decision makers approach and resolve a decision RDF. Current research to integrate these two computational based on their understanding of the situational context of that models is also focusing on RDF for all levels of context STIDS 2013 Proceedings Page 54 representation. The semantic RDF representation is then [20] Niu, L., & Zhang, G. (2008, December). A model of cognition-driven manipulated by KEs implemented as production mechanisms, decision process for business intelligence. In Web Intelligence and Intelligent Agent Technology, 2008. WI-IAT'08. IEEE/WIC/ACM sometime gathered into more complex require structures that International Conference on(Vol. 1, pp. 876-879). IEEE. chunk multiple reasoning elements into a unitary NIM KEs. [21] Pennington, N., & Hastie, R. (1993). The story model for juror decision- making. In R. Hastie, (Ed). Inside the juror: The psychology of juror REFERENCES decision-making. Cambridge series on judgment and decision-making. [1] Bex, F., Van den Braak, S., Van Oostendorp, H., Prakken, H., Verheij, (pp. 192-221). New York, NY, US: Cambridge University Press, B., & Vreeswijk, G. (2007). Sense-making software for crime [22] Pennington, N., & Hastie, R. (1993). The story model for juror decision investigation: how to combine stories and arguments?. Law, Probability making(pp. 192-221). Cambridge University Press. and Risk, 6(1-4), 145-168. [23] Polk, R. B. (2000). A Critique of the Boyd Theory-Is it Relevant to the [2] Boyd, J. R. (1987). Organic design for command and control. A Army?.Defense Analysis, 16(3), 257-276. Discourse on Winning and Losing. [24] Read, S. J., Miller, L. C., Kostygina, A., Chopra, G., Christensen, J. L., [3] Card, S., Moran, T., and Newell, A. (1983) The Psychology of Human- Corsbie-Massay, C., Zachary, W., LeMentec, J. C., Iordanov, V., & Computer Interaction. Mahwah, NJ: Erlbaum. Rosoff, A. (2007). The Personality-enabled architecture for Cognition. [4] Dyer, M. G. (1983). In-depth understanding: A computer model of In A. Paiva & R. Picard (Eds.). Affective Computing and Intelligent integrated processing for narrative comprehension. MIT press. Interaction 2007. Springer-Verlag. [5] Endsley, M. R. (1995). Toward a theory of situation awareness in [25] Read, S.J., Miller, L. C., Monroe, B., Brownstein, A., Zachary, W., dynamic systems. Human Factors: The Journal of the Human Factors LeMentec, J. C., & Iordanov, V. (2006) A Neurobiologically Inspired and Ergonomics Society, 37(1), 32-64. Model of Personality. In J. Gratch et al. (Eds.), Intelligent Virtual Agents 2006 (IVA 2006) (pp. 316 – 328), Lecture Notes in Artificial [6] Endsley, M. R. (1997). The role of situation awareness in naturalistic Intelligence, 4133. Berlin: Springer-Verlag. decision making. Naturalistic decision making(A 97-39376 10-53), Mahwah, NJ, Lawrence Erlbaum Associates, Inc., , 1997,, 269-283. [26] Riessman, C. K. (1993). Narrative analysis (Vol. 30). SAGE Publications, Incorporated. [7] Endsley, M. R. (2000). Flight crews and modern aircraft: In search of SA.Royal Aeronautical Society, 12. [27] Schank, R. C., & Abelson, R. P. (1995). Knowledge and memory: The real story. In R. S. Wyer, Jr. (Ed.), Knowledge and memory: The real [8] Endsley, M. R., & Connors, E. S. (2008, July). Situation awareness: story. Advances in social cognition, vol. 8 (pp. 1-85). Hillsdale, NJ: State of the art. In Power and Energy Society General Meeting- Lawrence Erlbaum Associates, Inc. Conversion and Delivery of Electrical Energy in the 21st Century, 2008 IEEE (pp. 1-4). IEEE. [28] Schank, R. C., Abelson, R. P. (1977). Scripts, plans, goals, and understanding: An inquiry into human knowledge structures. Hillsdale, [9] Feng, Y. H., Teng, T. H., & Tan, A. H. (2009). Modelling situation NJ: Erlbaum. awareness for Context-aware Decision Support. Expert Systems with Applications, 36(1), 455-463. [29] Selfridge, O. G. (1958). Pandemonium: a paradigm for learning in mechanisation of thought processes. [10] Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text comprehension. Psych. Rev, 101(3), 371. [30] Sheridan, T. B., & Johannsen, G. (Eds.). (1976). Monitoring behavior and supervisory control. New York: Plenum Press. [11] Hair, D. C., & Pickslay, K. (1993). Explanation-based reasoning in decision support systems. Technical Report. NAVAL COMMAND [31] VanLehn, K. (1996). Cognitive skill acquisition. Annual review of CONTROL AND OCEAN SURVEILLANCE CENTER RDT & E DIV. psychology, 47(1), 513-539. SAN DIEGO CA. [32] Zachary, W., Le Mentec, J.-C.. Miller, L.C., Read, S. J., & Thomas- [12] Hutchins, E. (1995). Cognition in the Wild (Vol. 262082314). Meyers, G. (2005a). Human behavioral representations with realistic Cambridge, MA: MIT press. personality and cultural characteristics. Proceedings of the Tenth International Command and Control Research and Technology [13] Klein, G. (2008). Naturalistic decision making. Human Factors: The Symposium, McLean, VA. Journal of the Human Factors and Ergonomics Society, 50(3), 456-460. [33] Zachary, W., LeMentec, J-C., Miller, L. C., Read, S. J., & Thomas- [14] Klein, G. A. (1993). A recognition-primed decision (RPD) model of Meyers, G. (2005b). Steps toward a Personality-based Architecture for rapid decision making. Decision making in action: Models and Cognition. Proceedings of the 2005 Conference on Behavioral methods, 5(4), 138-147. Representation in Modeling and Simulation, Los Angeles, CA. [15] Klein, G. A., Calderwood, R., & Clinton-Cirocco, A. (1986, September). [34] Zachary, W. & Ryder, J. (1997) Decision Support: Integrating Training Rapid decision making on the fire ground. In Proceedings of the Human and Aiding. In M. Helander, T. Landauer, and P. Prasad, Eds., Factors and Ergonomics Society annual meeting (Vol. 30, No. 6, pp. Handbook of Human Computer Interaction, 2nd Edition. Amsterdam: 576-580). SAGE Publications. North Holland, pp. 1235-1258. [16] Klein, G., A., Calderwood, R., & MacGregor, D. (1989). Critical [35] Zachary, W., Ryder, J., Stokes, J., Le Mentec, J.-C., Santarelli, T. (2005) decision method for eliciting knowledge. IEEE Transactions on Systems, A COGNET/iGEN Cognitive Model that Mimics Human Performance Man, and Cybernetics, 19, 462-472. and Learning in a Simulated Work Environment. In K. Gluck & R. Pew [17] Miller, L. C., & Read, S. J. (1991). On the coherence of mental models (Eds.) Modeling Human Behavior with Integrated Cognitive of persons and relationships: A knowledge structure approach. In G. J. Architectures. Mahwah, NJ: LEA. O. Fletcher & F. Fincham (Eds.), Cognition in Close Relationships. (pp. [36] Gentner, D., & Stevens, A. L. (Eds.). (1983). Mental models. 69-99). Hillsdale, NJ: Erlbaum. Psychology Press. [18] Morrison, J. G., Kelly, R. T., & Hutchins, S. G. (1996, October). Impact [37] Hampton, G. (2009). Narrative policy analysis and the integration of of naturalistic decision support on tactical situation awareness. In public involvement in decision making. Policy sciences, 42(3), 227-242 Proceedings of the Human Factors and Ergonomics Society annual meeting (Vol. 40, No. 4, pp. 199-203). SAGE Publications. [38] Hammond, K. R, Hamm, R. M., Grassia, J., & Pearson, T. (1987). Direct compar- ison of the efficacy of intuitive and ana- lytical cognition [19] Newell, A. (1990) Unified Theories of Cognition. Cambridge: Harvard in expert judgment. IEEE Transactions on Systems, Man, and Univ. Press Cybernetics, 17(5), 753–770. STIDS 2013 Proceedings Page 55