A major part of the training that you will receive as part of this course will result from working on a group project involving solving a real (or realistic) decision problem and building a decision support system that will support the problem. In the project, you will have an opportunity to apply the techniques that you will have learned in class.

We will identify a small number of project topics during the first few classroom meetings. The project topics will be identified and selected based on the expertise and the interest of the class members, significance of the problem, and feasibility of finding an acceptable solution within the scope of a class project (I will help with the feasibility judgment). The decision problems that we will consider will be real (or very realistic). They will be challenging in the sense that the choice from among the possible alternatives will not be obvious and the problem will be sufficiently complex to strain the limits of intuitive judgment. A challenging decision problem will have at least two of the following characteristics: uncertainty, multiple conflicting objectives, a large number of decision options, outcomes that extend over several time periods, or two or more decision makers with conflicting preferences.

Examples of challenging project topics are: evaluation of credit-worthiness of a loan applicant from the point of view of a bank, evaluation of risk related to undersigning an insurance policy from the point of view of an insurance company or determining the premium amount, evaluation of the quality of a job applicant from the point of view of a hiring company, identification of prospects for gain in a real estate deal, diagnosis of machine failures, diagnosis or therapy choice in a small area of medicine, etc. It is important that the decision problems that we identify are real, as this will allow you to appreciate the power and the limitations of the methods that you will have learned.

We will spend some time during our classroom meetings talking about the projects and the relevance of the material to your project. The real work, however, will happen outside the classroom. You will form teams of 5-10 people that will split the work and carry the project to a successful completion. The teams will be formed during the first class meetings. I will gladly assist you in case of difficulty in forming a team. Each member of the team carries the responsibility for the success or failure of the entire team. Individual team members will have the opportunity to evaluate the contribution of each of their colleagues in the team to the final result. While any conflict within a group is that group’s internal matter, I reserve the right to lower the grade of those students who fail to make a significant contribution to the project (as indicated by other team members).

The second type of alternative project involves solving a research problem of your choice. Most research fields have something to do with decision making and you can carve out an aspect of your current research that will tie nicely to this class. I will be also glad to suggest small and manageable research-oriented projects within the Decision Systems Laboratory. There have been at least two projects in the past that led to publications.

The choice and framing of a decision problem will be a major factor in the success of your project. In order to prevent you from wasting your time and to make sure that your topic is challenging enough, I want every team to submit a written statement of the problem (project proposal) for my approval. The proposal, due in about a month from the starting date of the course (the deadline is marked on the syllabus), should clearly state:

Details of project proposal, progress report, and final report concerning the project of an alternative type (software development and research) are quite flexible and will be negotiated on an individual basis in the beginning of the semester. Typically, the deliverable of a research-oriented project is a draft of a paper and the deliverable of a software implementation project is C++/Java/Visual Basic code with clear comments.

The proposal can be fairly short — a few pages usually suffice. I will let you know whether your formulation of the problem is acceptable as it stands or what you can do to improve it. If your proposal is not approved, you will have to submit another within a week. Once your proposal has been approved, you can proceed with your project.

As an indication, your report might contain the following sections:

It is often a good idea to start with a simplified model and a rough sketch of uncertainties and preferences. I expect that for each of the project teams we will start on such a sketch in the classroom. This simple model can then be refined by elaborating on those of its elements that are important for the decision. One way to identify these elements is to use simple methods for sensitivity analysis, such as the tornado diagrams. This approach will protect you from building a huge model that will imply numerous elicitations, many of which could be avoided.

The criteria for grading your project reports are: organization and planning of your work (as expressed by your proposal and mid-semester progress report), soundness, creativity, and, finally, clarity of your writing and expressing your ideas.

Aim at excelling in your project. If you work on a practical problem that is of interest to somebody, you may share your report with the decision-makers. You can also try to publish your report as an article describing a successful application of decision-analytic methods in decision support systems. Research-oriented projects should quite naturally lead to a publication — two projects in the past did. Software implementation will get you acquainted with the internals of Bayesian networks and influence diagrams and will earn you DSL's gratitude and an acknowledgment inside the package. A winning project may help you to advance your career. It is easier and more rewarding to excel as a student than as a "mature" professional.

All projects will be demonstrated during the last class meeting (see the course schedule for the date). The best project in class (as judged by the founder of the award) will be awarded a doughnut and a drink of the team’s collective choice along with an accompanying certificate.