Manas Hardas

980 Morris Road Apt #4, Kent, Ohio, 44240.  E-mail: mhardas (at the rate ) cs (dot) kent (dot)edu  Phone: 330-389-0464

 

 Home

 Research

 Resume

 Publications

 Contact

Research interests:

Manas Hardas – Research Statement 2008-09

In the academic year of 2007-08 I was able to publish two papers and present them at conference such as 5th Atlantic Web Intelligence Conference (AWIC 2007 Fontainebleau, France) and 3rd Indian International Conference on Artificial Intelligence (IICAI 2007 Pune, India). I also published a poster at K-CAP 2007 (The 4th International Conference on Knowledge Capture Whistler, BC). My primary area of research is learning theory, machine learning, knowledge representation, educational psychology and cognition. I was also able to co-author three more published papers with a fellow graduate student in the research area of “personalized negotiation in federated systems”. In this academic year of 2008-09 I would like to extend my research with a more specific area concentrating on “mathematically modeling learning theories”.

Traditional learning theories can be broadly divided into behavioral, connectionist or cognitive theories. I am particularly interested in the cognitive explanations for learning because they fit well with the traditional graph theoretic approaches of computer science. We envision the learner’s mental map as a complex graph of concepts linked with relationships labeled with different kinds of semantics. I our representation we symbolize the mental map as a constructively developed map with newer concepts formed over the knowledge of previous concepts. Constructivism is a cognitive theory of learning given by the acclaimed psychologist Jean Piaget which explains how knowledge is acquired by a learner by the means of assimilating and accommodating newer knowledge into already existing structures of knowledge [1, 2]. It says that newer knowledge is built on existing knowledge with the advantage of experience. Of course educational psychology fails to address their specific explanations in the context of graph theories which are firmly grounded in computer science with sound mathematical basis. I plan to model the phenomenon of constructivism as seen happening in everyday learning and give it a mathematical explanation. In the previous published work, we have been able to get our representation of knowledge framework critically evaluated in various conferences [2, 3]. The basis of this representation again is engrained in the theory of constructivism. In our representation the knowledge required to understand a particular concept is given by the children concepts. Thus a constructively developed bottom-up map is formed where in the leaf concepts constitute the knowledge of their parents, which constitute the knowledge of their parents and so on. Using this as the basis of our knowledge structure I plan to define the exact operations which contribute to “learning”. We define learning as a “demonstrable acquisition of the expert knowledge map by the learner”. Using this definition I will conduct experiments where I can test different cognitive learning theories with a mathematical basis.

 The research project can be broadly divided into three phases:

1. The very first phase in the project is to ready the expert knowledge map for hypothesis testing and experimentation. We have decided to compose the expert knowledge map for the field of Computer Science, with 7 basic topics namely, Operating systems, Computer Architecture, Data Structures, Programming Languages, Discrete Structures, Algorithms, and Computer Networks. The granularity of the knowledge map will be refined as time progresses and finer concepts and links are added and deleted from the knowledge map. Standardized text books are used to create these knowledge maps, the granularity of concepts ranging from the high level concepts from the index to low level concepts extracted from actual lines of text. Creating a comprehensive map of Computer Science knowledge is the first phase of the project.

 2. In the first phase we develop a mathematical model for explaining the process of learning defined by a learning theory from educational psychology. The backbone representation for this model is the constructively developed graph, also called as the learner’s mental map, which adheres to the original graph representation. The act of learning is said to have occurred when there are changes made to this graph by the learner as a response to a particular stimulus. We define an array of plausible operations like add link, delete link, add node, delete node, merge node, split node, reinforce link etc. that are used to make amendments to the learner’s map. Once we have all the plausible operations, we go further and try to analyze in a step by step process the actual metamorphosis of the learner’s mental map from M1 to Mn as the learner acquires the expert map through the medium of instruction. For the time being we restrict the medium of instruction to text excerpts from standardized text books. The next step in this phase is to mathematically explain the processes of assimilation and accommodation as described by Jean Piaget. By the theory of constructivism, assimilation of knowledge occurs when a student acquires new knowledge and it does not conflict with previous conception. However if it does conflict then a learner moves into a state of disequilibrium (or confusion) and to recover from this state the learner has to change this preconceptions to accommodate the new knowledge. These two theories fit very well with the graph theoretic point of view.

3. In the second phase, we develop mechanisms to understand individual student learning process and consequently guiding the process of learning based on the mathematical model from phase 1. This phase is loosely divided into two experiments: · In the first experiment I plan to map the learner’s mental maps before, during and after the process of learning. The mapping of the gradual metamorphosis of the mental map of a student can give great insights into exactly how human mind works while learning. We use the technique interviews-about-instances (IAI) to roughly map the student’s alternative conceptions, or naďve conceptions (learner’s initial mental map) at different times of the experiment and using the operations defined in phase 1 try to visualize the actual learning process. The IAI technique is carefully designed so as to exactly figure out which operation is the learner subconsciously executing at exactly which instance of learning activity. ·

 In the second experiment we streamline the process of learning itself by controlling the process of instruction. By algorithmic means, based on the specific learning theory, the sequence in which particular topics need to be taught can be generated. The algorithm travels through the predefined course concept space generating a sequence of concepts which need to be taught. Unless the learner demonstrates that he/she has acquired the concept knowledge, the algorithm does not proceed. Then the student can be recursively accessed for the acquired knowledge, making sure at every stage of instruction the student acquires the correct mental structures.

This research has a lot of implications to education psychology and machine learning. Until now, nobody in either of the field has looked at learning as a metamorphosis of a concept map. Educational psychologists do have theories for learning and instruction by conceptual transformation but none of them have graph theoretic approach and therefore lack the soundness of mathematical proofs. By treating the process of learning more objectively than subjectively we can tap into the vast knowledge of computer science and use it to explain to some extent the subjectivity in psychology. This research grant will help me pursue this multi disciplinary research project for the next year and hopefully I will be able to come up with some ground breaking results.

 

 References:

 [1] Piaget, J. (1971). Biology and Knowledge. Chicago: University of Chicago Press.

 [2] ATHERTON J S (2005) Learning and Teaching:  Assimilation and Accommodation   [On-line] UK: Available: http://www.learningandteaching.info/learning/assimacc.htm  Accessed: 31 January 2008

 [3] Javed Khan and Manas Hardas, “A Technique for Representing Course Knowledge Using Ontologies and Assessing Test Problems”, pp. 174-179, 5th Atlantic Web Intelligence Conference 2007, June 25-27, 2007 - Fontainebleau, France.

 [4] Javed Khan and Manas Hardas, “Hierarchical Course Knowledge Representation Using Course Ontologies”, in IICAI 2007, the 3rd Indian International Conference on Artificial Intelligence, December 17-19 2007, Pune, India.

 

 

 

   © 2005-2006 Manas Hardas, Mail to: mhardas (at the rate ) cs (dot) kent (dot)edu