An introduction to decision making under uncertainty in terms of computational science, covering applications and theoretical areas ranging from self-driving cars to aerial collision avoidance. This book is the first of a two-book series on Artificial Intelligence (AI). In this book, I cover the history of AI, describe what the current state of the art is, and briefly survey some of the challenges that researchers are currently facing. Additionally, I will discuss the potential for AI to provide humans with new capabilities.
Since the dawn of computing, artificial intelligence has had to deal with decision making under uncertainty. Typically, artificial intelligence systems have performed poorly when presented with multiple options, many of which are uncertain. Arguably, the most frustrating example of artificial intelligence decision making under uncertainty is the game called Prisoner. Players take turns being a prisoner and being able to think ahead and make decisions. The computer keeps track of all actions played and calculates the optimal strategy based on the information it has.
Decision Making Under Uncertainty
Max Tegrist and Bruce Eckel are the founders of the Max Tegrist Center for Computational Thinking at the University of Maryland at College Park. Max and Bruce are currently funded by the National Science Foundation and the National Institutes of Health. This book is a short introduction to some current research in the field of decision making under uncertainty. It provides an overview of their work and conclusions.
Artificial intelligence researchers face decision making under uncertainty in many of their areas of research. Two of the biggest problems facing these researchers are the lack of a clear cut definition of what intelligence is and the lack of a framework by which to evaluate artificial intelligent software. The lack of a clear cut definition has given rise to a large number of different human styled languages and a great deal of technical jargon. These terms often lead to disagreements within the artificial intelligence community. The second major problem faced by artificial intelligence researchers is that of decision outcomes that are sensitive to various unknown external factors.
In many decision-making situations uncertainty is multiplied by the number of agents who are involved in the decision. For example, if there are five agents involved in a decision-making process and one of those agents has uncertainty about the order the stock will be picked, then the uncertainty factor is multiplied five times. One of the main approaches used to reduce this uncertainty is to perform an uncertainty analysis. The uncertainty analysis reduces the standard deviation of the outcome by removing the known factors that cause uncertainty and then incorporating the new variables into the original decision making problem.
A Much Ado
The uncertainty analysis presented in this text is presented in a graphical format, with a range of confidence intervals representing the range in which the decision lies. It is hoped that by understanding the uncertainty factors better the researcher can better understand the problems they are faced with. This text provides a useful introduction to the various difficulties encountered in decision making under uncertainty.
The uncertainty analysis presents a range of recommendations on how to make decisions under uncertainty. One of the recommendations is to use a range rather than a single value or parameter. Another recommendation is that estimates are performed over the interval instead of just the single value or interval. Also suggested is that researchers consider using a bootstrap procedure, which is a more rigorous way of estimating uncertainty, but not as demanding as the traditional procedures. The text also concludes that it is important for decision-makers to understand the uncertainty of the estimates so that they can make good estimates.
In conclusion, I have presented a brief review of the uncertainty analysis problems associated with decision making under uncertainty. There are a number of difficulties encountered when making estimates of uncertainty. The review has concluded that it is important to understand and determine the uncertainty of your estimates. This should be done not only prior to making an investment but also after an investment has been made. The research should include a detailed description of the uncertainty issues, an analysis of the risks associated with the uncertainty, and recommendations on how to make better estimates.