Smarter Decisions in the Information Age
My research focuses on the theory and practice of helping people make better decisions when they confront several sources of information. Analogous to the three legs of a stool, the three elements of the basis for any decision are (1) The alternatives, or what we can do; (2) The information, or what we know; and (3) The preferences, or what we like. These elements, together with the normative logic, the committed decision-maker, and the proper framework for the problem form the six elements of decision quality. When faced with uncertainty, people may choose different alternatives based on their taste for risk and the information they have about that uncertain deal. A probability distribution captures the information element. It is a representation of their belief on the likelihood of occurrence of an uncertain event. When several sources of uncertainty are present, however, eliciting a joint probability distribution on all the variables becomes a complicated task for the decision-maker. In the early phases of my research, I have developed methods to help the decision-maker elicit single and joint probability distributions and to fill in missing information she was unable to provide to the decision analyst.
With the advent of the information age, several sources of information are available through data sets, data warehouses, the Internet, e-monitoring and information classifiers such as neural networks and genetic networks. Because information is one of the three main elements of the decision basis, it is important to incorporate these sources of information into our analysis of any decision situation.
Using Information Theory, I have also developed a method to aggregate the decision-maker's prior knowledge with the information captured from data sets and from information classifiers to give a clear method of updating the decision-maker's belief. I applied my algorithms to several real world examples including a complicated business assessment with several uncertainties. The decision-maker was president of a semi-conductor company who had uncertainty on price, market demand, cost and technical success of the product. The marketing department had data on marketing success, price, cost and market demand. The challenge was to aggregate the decision-maker's belief with the marketing data set and use the aggregated result for decision analysis. I also applied the method to a fine needle aspiration biopsy for breast cancer to help the physician update her probability assessment on whether the tumor is benign or malignant given the test indications and records of previous patients.
At the present, I am also using the same algorithms in bio-medical informatics
to infer the structure of proteins using prior information from expert
knowledge of protein sequence evolution and the output of DNA multiple
sequence alignment. In addition to a practical algorithm to make this
alignment with fewer calculations, this last application can help in the
treatment and prevention of diseases by understanding the structures in
the protein that lead to the phenotype. Instead of treating a disease
by diagnosing the symptoms and giving a cure, I am working on tracing
back to the genetic code and looking for structural deficiencies in the
patient's DNA that could have led to that disease. Although this is still
in its early research stage, applications of information theory to bio-medical
informatics are on the verge of revolutionizing the medical decision making
|Modified 15 January 2003 * Contact Us|