 Decision Management Solutions® |
||||||||||||
|
The Use of Belief Maps to Track Predictions for the 2008 Election
The Decision Expert Newsletter™ - Volume 2; Issue 3Time Magazine has begun tracking the 2008 election using an "Election Index," a plot of support for each of the leading candidates versus knowledge about them. Time's Election Index is very similar to the Belief Map developed by Robust Decisions Inc. in the 1990s and patented in "General Decision-Making Support Method and System" 6,631,362, October 2003. Belief Maps can support many decision-making activities—Time's version, from the April 9, 2007, edition, is shown below. (click to see a larger version)  Belief Maps came into being as a simple graphical method to communicate uncertainty or knowledge during the evaluation of critical technical and business decision-making information, much as Time is doing. Beyond this, Belief Maps are also used to help develop consensus for critical evaluations and as an information capture interface for a leading decision management software tool—Accord™. Belief Maps are fully described in Chapter 8 of Making Robust Decisions. Exploring Time's use of Belief Maps shows how easy it is to communicate and manage uncertainty (i.e. knowledge) in a decision process using these simple plots. Uncertainty during most decision-making implies a lack of knowledge, as knowledge grows so does certainty - sometimes (See the counter-example of the Wright Brothers on page 220 of Making Robust Decisions).The rest of this article will show the communication ease and other benefits of Belief Maps and potential ways that Time could increase the value of the election data with very little additional effort. Making business and technology decisions is not the same as a poll about the 2008 election. Each of the points made here apply even more so to decisions that require more than a one-dimensional vote.  To begin, the data for some of the candidates was read from Time's figure of a month ago and input into Accord to be evaluated against the single criteria "support for the candidate". This is shown on the right. While not as graphically pretty as Time's version, it helps us explore what is really happening. Time refers to their plot as an "Election Index". Usually the term "index" refers to a single value, but Time does not develop such a value. However, Accord finds the overall satisfaction with each of the candidates and gives the following results for the leading candidates:
Using these satisfaction values as the "Index", the Obama-Giuliani-Clinton grouping is not surprising, they are all slightly better than 50%. What is important about this grouping is that, as Time points out, Hillary Clinton has nowhere to go. Knowledge about her is high so her rating can't change much as more is learned about her (note the Wright Brother's counter-example mentioned above). Support for Barack Obama on the other hand can increase or decrease as more is learned about him.  The results also show virtually the same index for Edwards, McCain, Dodd and Romney. This warrants explaining and can best be seen on the Accord software's Belief Map with the "Show Isolines" option on. These lines show, for subjective evaluations, lines of equal satisfaction represented as probabilities. The shape of these lines is based on Bayesian Team Support (BTS)—the mathematics on which Accord is built. If a candidate's point was in the upper right corner, their support would be very high and everyone is sure of it, hence the probability would be 1.0 (100%). If in the lower right corner, their support is very low and everyone is sure of it, hence the probability would be 0.0 (0%). If anywhere along the left edge, the people know nothing about the candidate so, the evaluation is no better than a flip of a coin; the probability is 0.5 (50%). Thus, Romney and Dodd are little known and with little support, so they are at about the same level as McCain and Edwards who are better known with less than 50% support. Time has reduced the data from 1,144 surveys to single data points, one for each candidate. They asked questions such as "If your choices for the Democratic nomination were just Hillary Clinton and Barack Obama, which one would you vote for if you had to decide today?". Averaging during business and technical decision-making is dangerous when the information is uncertain. Consider a very simple example where you are evaluating an alternative against a single criterion. You evaluate it as having high satisfaction based on your high knowledge about it. Looking at the Belief Map with the Isolines, an H-H (high-high) evaluation gives about 69% satisfaction with the proposal (it would be just below the curved line labeled 0.7). Say a colleague also evaluates the proposal the same way, and a third one does also, and a forth. The average evaluation is still 69%, but now you have four people, all agreeing that based on their "high" knowledge, the proposal rates high relative to the one and only criteria. Most decision-makers would consider this 69% average to be too conservative in this case. Accord calculates 69% - 83% - 92% - 96% as more people are added, better reflecting reality. So, Time's average may not be showing the correct results. Further, you may wonder, when looking at Time's plot whether or not people really have 95% knowledge about Hillary Clinton (for example). Do people actually know that much about her stand on important issues? What if Time had asked a few more questions? What more could they have learned? Going beyond what Time did, supposed they had identified the dominant issues of the day and used the Accord toolset to help analyze the choices. Here is what could happen. Say we agree that there are nine issues currently discussed when talking about the candidates:
If the electorate was asked to rank the four most important issues, there would be a wide variety of responses. Some would respond that terrorism is most important, followed by electability and so on. Others would have an entirely different list. For the poll here there may as many different lists as there are possibilities. In a business or technical situation the variation would be less, but no less important. Why only the top four, because the top four out of nine carry about 80% of the weight in a decision (see page 179 in Making Robust Decisions). Next, for the four that they each think are important, they are asked to respond to the following (for example):
These could also be asked relatively as in:
Through the use of Accord with this type of data, users can gain a better understanding of the following:
Robust Decisions is not in the polling business, but if someone wanted to use Accord to finish this problem, then it is an ideal tool to make the best possible use of polling information and deal effectively with uncertainty among a group of stakeholders. Even more important, Accord can help achieve robust decisions for technical and business decisions, the problems for which it was designed. Avoid the blunders that plague many decisions in today's business and engineering environments through the development of some simple "decision habits." The new book, Making Robust Decisions, presents a simplified approach to helping individuals and organizations establish better decision processes. Learn more by selecting the link below for more information. Are you ready to improve the decision-making in your organization? Robust Decisions offers a full range of products and services that allow you to benefit by making the best decision every time. Learn more about how to build a strong foundation that enables individuals, teams and entire organizations to consistently make the best possible decisions. Request more information or call 541.758.5088. 
|
SubscribeSubscribe to the monthly Decision Expert Newsletter™: Newsletter Archive
Learn MoreRecent Blog Posts
|
