Cognitive Decision Models - Essay Example

Cognitive Science Introduction The term cognitive science has been described as a rubric for a multiplicity of associated disciplines, subdisciplines, and intradiscipline lines of inquiry. Flanagan (Press) stated in his treatise on the topic,” The Science of Mind,”: Cognitive science is not a discipline so much as it is an increasingly well-organized committee of disciplines and sub disciplines, all of which claim to have something to contribute to our understanding of mentality. More specifically, cognitive science is a confederation of philosophy (especially philosophy of mind, philosophy of language, epistemology, and logic), cognitive psychology, neuroscience, linguistics, and computer science. (p. 177) The current aptness of this summary is corroborated by an examination of advertisements for graduate cognitive science programs. Affiliated faculty represent computer science, social psychology, cognitive psychology, linguistics, philosophy, human-computer interaction, cognitive development of mathematical concepts, emotion (its cognitive representation), neuroscience, statistics (applied probability), and even nonlinear dynamical systems analysis and sociobiology. Such a gathering of contemporary scientific pursuits is bound to lodge a treasure trove of nascent technology for cognitive assessment in the clinical arena. The rigorous theoretical platform for psychological assessment generally, which historically has been supplied by psychometric theory, stands to be paralleled by quantitative cognitive science when it comes to cognitive assessment (Cecile & Ruder, 432). Such application has been variously depicted as “cognitive psychometrics” (Batchelder & Riefer, 60),“cognitive modeling” (Busemeyer & Stout, 262), and “quantitative cognitive clinical-science” (McFall & Townsend, 320). The corpus of contemporary developments in cognitive-science-based assessment technology is variously represented throughout the present set of articles. Concepts and Themes From a cognitive-science perspective, understanding cognitive-task performance necessitates theoretical modeling of constituent transactions. In clinical applications, models or modeling approaches from mainstream cognitive science are imported and adapted. Properties of performance include response accuracy (e.g., correctness in identifying specified items as members of visual arrays or previously memorized item sets), response speed (e.g., latency from trial commencement to response registration), and response type (e.g., item classification as new vs. previously encountered, or more vs. less desirable; rating of similarity between items; and so on). One or some combination of these response properties become the empirical anchors of prediction and explanation. The most productive models arguably are those whose formats comprise formal deductive systems (see, e.g., Braithwaite, Press). Reasoning in such systems is conveyed by specified rules, as exemplified in computer languages, symbolic logic, and mathematical derivations. The latter tack represents the chief formal modus operandi brought to bear on clinical cognitive assessment. Quantitative formal logical- deductive systems are deemed to lodge explanatory constructs that can be recruited to the service of the desired explanation, measurement and prediction. By the very nature of the aims of the clinical- assessment enterprise, performance models of most immediate relevance justifiably are quantitative. If individual client assessment is a desideratum of clinical cognitive science, quantitative models are indicated all the more. Model Architectures and Parameters A major component of almost any formal performance model is its architecture, or structure. This property comprises the arrangement of theoretical cognitive processing operations. To illustrate, the encoding of physical item features into a format facilitating collateral cognitive activities (e.g., memorial manipulations of item associations) may proceed in serial (successive feature encoding), parallel (“simultaneous” feature encoding), or possibly as a serial-parallel hybrid, comprising some combination of the parallel-serial constructions. Relations among constituents of any number of “processing composites,” such as the visual, memory, and response functions ostensibly mediating item classification, may be fashioned in like manner. Another prominent feature of many formal models involves their parameters. 1 A parameter is “an arbitrary constant whose value affects the specific nature but not the formal properties of a mathematical expression” (Borowski & Borwein, p. 435). It may be said that in modeling of cognitive performance, a parameters value affects theoretical predictions but not model architecture. Most readers are well versed in the role of parameters in statistical analyses, such as those making use of the Fisherian analysis- of-variance linear model, the general linear model, or the common-factor model. Parameters in these cases depict sources of variance and/or covariance in empirical data. These parameters facilitate the partitioning and analysis of data sets, making for tractable interpretation of research results. Such parameters are embodied in “data theory,” and can be deemed “generic,” in that they cut across the specific content of investigation. Parameters of cognitive theoretical models, on the other hand, are substantively significant, expressing designated aspects of mentation. Examples from the present articles include attention (“perceptual augmentation-diminution” of stimulus dimensions), response bias, and sensitivity to interstimulus separation during classificatory judgments (Treat et al., 240); motivational, response, and information-integration processes (Busemeyer & Stout, 265); strength of activating connections among neuronal units, or psychologically significant unit modules (Siegle & Hasselmo, 264); capacity, or rate of process transaction, and process complexity or magnitude (Neufeld, Vollick, Carter, Boksman, & Jetté, 279). As noted by Chechile and Roder (p. 435; see also Busemeyer & Stout, 267), formal modeling of cognitive performance parallels the well-known application of signal- detectability theory (SDT) in perception. The parameters d′ and β of the performance model delineate two determinants of observed responses: sensitivity to the occurrence of the designated signal and liberalness-conservativeness in inferring its presence on given trials. These influences on performance are conflated in raw responses; however, separate measures of their values can be realized through model application. The parameters have differing psychological meaning, and are disparately predictive of other variables. Apropos of clinical cognitive dysfunction, performance associated with clinical disorder is put under the scrutiny of formal cognitive models. Altering model architecture may be required to capture performance deviations, but it is typically more parsimonious to perturb model parameters. Either way, the nature of model modification discloses changes in cognitive workings associated with the presenting disturbance. As in the case of SDT, cognitive functions to which the parameters pertain are set within their context of operation, which includes the accompanying functions collectively involved in task execution (Busemeyer & Stout, 269). Function assessment is achieved through the model-driven methods that validly dissect performance into its component parts. In this way, the intrinsic context of function-operation is retained, rather than being dismantled through isolating, and potentially distorting, interventions. This is not to say that experimental manipulations per se are incompatible with modeling strategies. Quite the opposite: A key aspect of verifying the nature of model parameters, for example, includes selective responsiveness of their values to variables they are supposed to express (see, e.g., Chechile, 222). The purpose of parameters and architectures in any particular theory and its application can best be understood by examining their roles within the formal deductive system of which they are a part (Braithwaite, Press). This observation, of course, holds for each of the articles in this special section. Finally, parameters ideally are linked to significant clinical issues or to variables aligned with issues. The latter notably have to do with treatment and prognosis (Maher, 306). Cognitive Efficiency and Cognitive Content An additional division pervading cognitive-clinical science and assessment entails the effectiveness of cognitive functions (e.g., speed and completeness of executed operations) and the semantic aspects of processed material. In information-processing terms, a distinction roughly might be made between processing efficiency and the subject matter of processing (a distinction that parallels what colloquially has been called “cold” vs. “hot” cognition). Applications of stochastic multidimensional scaling by Treat et al. (p. 242) entail quantification of perceptual salience of dimensions of interstimulus distance that underlie clients cognitive mapping of problem-significant items. They entail, as well, “efficiency” in implementing the above distances in classificatory judgments, with respect to acuity of interstimulus separation. Busemeyer and Stouts (p. 272) decision-field-theoretic analysis of debilities in choice-performance provides for estimation of motivational bias in selection responses, meaning the comparative influence of positive versus negative outcomes attending test-trial selections. It provides, as well, for estimation of the way in which choice-relevant information conveyed by past trials is brought to bear on current selections. Third, it assesses the extent that choice performance is dictated by (as opposed to dislodged from) ones bank of response-pertinent information. Conclusion These developments illustrate an admixture of model features differentially emphasizing processing efficiency and content. In some accounts, content emanates from efficiency; such is the case, for example, in Siegle and Hasselmos (p. 266) approach to the assessment of affect and associated responses. Within the connectionist architecture, efficiency of inter-unit communication roughly comprises the network of weight- settings governing unit activation. The pattern of weight strengths, in turn, embodies the tone and intensity of affect imparted by task items. In like fashion, our own developments on task-facilitative stimulus encoding in schizophrenia (Neufeld et al., 281) converge with thought-content symptomatology (delusions and thematic hallucinations). Elongation of the encoding process risks the loss of contextual cues conveying the objective significance of events, persons, or objects. Owing to diminished encoding efficiency, inferences may be less than veridical, if mutually coherent. References Batchelder, W. H., & Riefer, D. M. (1999). Theoretical and empirical review of multinomial process tree modeling. Psychonomic Bulletin and Review, 6, 57–87. Borowski, E. J., & Borwein, J. M. (1999). Collins dictionary of mathematics. London: HarperCollins. Braithwaite, R. B. (1968). Scientific explanation. London: Cambridge University Press. Busemeyer, J. R., & Stout, J. C. (2002). A contribution of cognitive decision models to clinical assessment: Decomposing performance on the Bechara gambling task. Psychological Assessment, 14, 253–272. Chechile, R. A. (1987). Trace susceptibility theory. Journal of Experimental Psychology, 116, 203–222. Chechile, R. A. (1998). A new method for estimating model parameters for multinomial data. Journal of Mathematical Psychology, 42, 432–471. Flanagan, O. (1991). Science of the mind (2nd ed.). Cambridge, MA: MIT Press. Maher, B. A. (2002). An afterword: The utility of cognitive models for the field of psychopathology. Psychological Assessment, 14, 304–310. McFall, R. M., & Townsend, J. T. (1998). Foundations of psychological assessment: Implications for cognitive assessment in clinical science. Psychological Assessment, 10, 316–330. Neufeld, R. W. J., Vollick, D., Carter, J. R., Boksman, K., & Jetté, J. (2002). Application of stochastic modeling to the assessment of group and individual differences in cognitive functioning. Psychological Assessment, 14, 279–298. Siegle, G. J., & Hasselmo, M. E. (2002). Using connectionist models to guide assessment of psychological disorder. Psychological Assessment, 14, 263–278. Treat, T. A., McFall, R. M., Viken, R. J., Nosofsky, R. M., MacKay, D. B., & Kruschke, J. K. (2002). Assessing clinically relevant perceptual organization with multidimensional scaling techniques. Psychological Assessment, 14, 239–252. Read More