Simulación de la comprensión de narraciones mediante la implementación del modelo "Landscape" y el rol de las inferencias emocionales

Abstract

El propósito de este trabajo consistió en estudiar el rol de la generación de inferencias en la comprensión de textos narrativos; para ello se realizaron dos simulaciones con el Programa Landscape sobre un texto narrativo natural, una simulación con arreglo de inferencias emocionales y otra sin ellas, que se pusieron en relación con datos conductuales de sujetos, obtenidos del reconocimiento de las oraciones del texto y de su valoración o relevancia para la historial. Con este propósito, participaron 30 adultos universitarios. Leyeron la narración y posteriormente respondieron a un protocolo de reconocimiento y valoraciones de proposiciones del texto. A continuación se llevó a cabo un análisis de correlaciones y luego un análisis de regresión lineal empleando los valores de proposiciones predichas a partir de dos simulaciones, una simulación causal-referencial (sin emociones) y otra con implementación de inferencias emocionales, como variables predictoras, y los valores obtenidos de reconocimiento y valoración y/o relevancia para la historia, como variables dependientes. El análisis de correlación mostró que ambas simulaciones se asociaron con los valores de reconocimiento y valoración, aunque en mayor medida, la simulación con implementación de inferencias emocionales; pero el análisis de regresión detectó que únicamente la simulación con implementación de inferencias emocionales explicó los datos obtenidos de reconocimiento y valoración. Estos hallazgos sugieren que las inferencias emocionales juegan un rol de importancia en la comprensión de textos narrativos, ya que permiten que el lector focalice su atención hacia determinados puntos de la historia, que serán los nucleares o más importantes.

Text comprehension requires the construction of a coherent mental representation, integrating text information with previous knowledge. Several studies have suggested that the emotional states of characters need not be stated explicitly: readers can infer them as a consequence of the narrative situation, characters' goals, actions, and relations to other characters. This study investigated emotional inferences generation during reading using the simulation of a narrative text, carried out with the Landscape Program and assessing text sentence recognition and the relevance of the sentence to the story. This program is a connectionist model that represents comprehension as a changing landscape of activations of propositions along several reading cycles. This model proposes that, as the reader procedes through a text, propositions fluctuate in activation. That is, with each reading cycle, new propositions are activated, and activation values of current propositions change. In addition, the coactivation of propositions leads to the establishment of connections between them. Through these fluctuating activations, a memory representation of the text gradually and dynamically emerges. The peaks and valleys of this landscape represent the relative contribution of each proposition at any given point in the story, and are the base for the construction of a mental representation of the story. The aim of this work is to determine if comprehension of a narrative with emotional inferences is a better predictor of textual proposition recognition and its relevance to the story than comprehension of a narrative without emotional inference, using a program that simulates the comprehension process. In order to simulate the generation of emotional inferences, we used the Landscape Computational Model. For this purpose 30 participants, undergraduates (9 males -30%- and 21 women, with a mean age of 20.67 years, SD = 2.85), read a story and subsequently completed a proposition recognition and proposition relevance for the story protocol. Pearson correlation analysis and linear regression analysis were conducted. Two linear regression models were tested, both including propositional values of simulation with emotional inference and without emotional inferences as independent variables, one including proposition recognition as dependent variable, and the other including proposition relevance for the story as dependent variable. The Pearson correlation coefficients showed that the simulation of the story with emotional interference and the simulation of the story without emotional interference are related to proposition recognition and story relevance, although the relation between this values and the simulation based on emotional inference had better coefficients (r = .35, p < .01, and r = .38, p < .01 respectively) than the causal-referential simulation (r = .27, p < .05, and r = .28, p < .05 respectively). The linear regression analysis detected that only the simulation with emotional inference explained the variance of recognition data ( = .45, p < .01) and the variance of the relevance to the story values ( = .53, p < .01). These findings suggest that emotional inferences play an important role in the understanding of narratives texts; because they focus the reader's attention to certain important points of the story. That is, the realization of emotional inferences seems to intensify the attention that the reader devotes to the entire cycle (which includes the proposition that prompts the inference, and also those that are causally connected to it), facilitating its later recognition and relevance to the story. This intensified processing can be related to the role that characters' emotional reactions play in a narrative.