Bazzanella, Barbara (2010) Uniqueness in Cognition. PhD thesis, University of Trento.
|PDF - Doctoral Thesis|
A fundamental aspect of human cognition is that we construe the environment as including unique individuals that belong to various categories. An individual dog, for example, could simultaneously be a living being, a mammal or a poodle, but when it comes to things that are important to us - our dog Fido, our favorite restaurant, our spouse - we also represent the individuals themselves, not just the categories they belong to. Cognitive psychologists have made an extensive study of categories of objects but have had less to say concerning conceptions of individuals, i.e. singular concepts, and how they support our ability to uniquely identify individual entities in different situations. The primary goal of this work is to investigate the nature and the functioning dynamics of singular concepts and explore how these concepts underlie singular cognition, i.e. the ability to identify a known entity, through perceptual or epistemic access to its memorial representation, and trace it as the same unique entity over time and change. To perform such a process the cognitive system is confronted with a uniqueness problem. It needs to pick an individual entity out, secure a unique mental referential link with the entity and maintain that link over time and change. We argue that singular concepts are the cognitive devices that are specialized for this function and we propose a model of singular cognition that has the notion of singular concept at its core. The main assumption of this model is that conceptual representations about individuals (i.e. singular concepts) represent a networks of unique files in memory which mediate the direct access to individual-specific knowledge and provide a unique mechanism of identification and reference for unique individuals. According to our model, the access to this system is not mediated by higher level representations (i.e. general concepts), neither is internally organized by these representations. On the contrary, it is subjected to its own functioning dynamics and it is organized through associative links which connect different individual concepts and causal links which maintain the conceptual history of an entity, by linking different states of the same singular concept, across time and change. We can distinguish four main phases of our investigation about singular concepts which led to the proposed model of singular cognition. 1) In the first phase we investigated what is the preferential level of abstraction at which an individual entity is first identified (i.e. the entry point of recognition). Since any individual object can be identified at multiple levels of abstraction (e.g. a dog can be Identified as a “dog”, more generally as “animal” or more specifically as “poodle” or “Fido”), the aim was to test the hypothesis that the singular concept of an object acts as the access node to the knowledge that the agent has about the object and this access is direct and not mediated by higher level concepts. Results from three experiments on visual recognition provided evidences in favor of this hypothesis, indicating that the entry level of identification of unique individuals is shifted to the most subordinate level of abstraction, i.e. the level of unique identity. 2) The second phase of this work explored how our semantic representations of individual things are accessed and how these representations are inter-linked with those of other individual things. This issue has been investigated through a priming experiment which provided evidence in favor of a model in which singular concepts are organized by means of horizontal associative links instead of by vertical links with higher level representations. 3) In the third phase of our investigation we looked inside a singular concept and we explored which attributes people consider more relevant to uniquely identify entities belonging to different categories and determine the cognitive importance that individual attributes have in identifying these entities. We also explored which are the most relevant attributes that people use to identify entities in a specific task, i.e. the search for information about individual entities by means of keyword queries on the Web. 4) The last phase of the investigation concerned with the problem of how people judge the identity of entities over time and change. An experiment was conducted which explored how people evaluate the identity of entities over changes in their descriptions. The results of the study have been interpreted in the light of a causal model of the functioning of singular concepts in keeping the unique referential link with the entity across change. Beyond the cognitive issues, this work is also motivated by the recent development of technological approaches to the problem of entity identification. Since many identification problems which are addressed by a cognitive system have a counterpart in information systems which manage information about individual entities (e.g. to represent or extract information about unique individuals and manage individual-specific knowledge across time and change), the last goal of our work is to make an investigation of possible contributions that a cognitive study on the problem of individual identification can provide to technological applications. In particular we focused on the problem of entity identification in search systems. A model and an application for a specific technological problem, i.e. entity type disambiguation in Web-search queries, is described and its beneficial impact is evaluated. In summary, the contribution of this work is twofold. On one hand, we provide new evidence on the nature of high-level cognitive mechanisms involved in entity representation and identification, revealing new research issues on this topic in cognitive psychology. On the other hand, we show how a better understanding of these processes at a cognitive level can improve the development of entity identification approaches in information systems, suggesting a middle ground where cognitive models and technological models developed in other research fields can find the opportunity for communication and integration.
|Item Type:||Doctoral Thesis (PhD)|
|Doctoral School:||Psychological Sciences and Education|
|Subjects:||Area 11 - Scienze storiche, filosofiche, pedagogiche e psicologiche|
|Repository Staff approval on:||29 Nov 2010 13:06|
Repository Staff Only: item control page