Decision Analysis (DA) is the application of sciences such as philosophy, theory, methodology, and other faculties to address elements that constitute parts of a problem, and arrive at a decision in a formal manner. The process involves a set of procedures, methods, and tools to identify, represent, and asses the crucial aspects of a decision, and recommend a course of action.
First espoused by Ronald A. Howard, a professor of Stanford University in 1964, the concept has since then developed considerably and finds application in almost all areas including business management, dispute resolution, health care research, and more.
The most common method of representation is decision trees, with each branch of the “tree” representing one possible course of action. Uncertainty finds quantification as a probability percentage or probability distributions. Another form of representation is a logical flow chart, with each situation leading to various possible outcomes, and each outcome leading to further situations until the process reaches its logical end. Such graphical illustrations helps the decision maker cater to all; possible eventualities and reduce uncertainty. Such graphical representation allows the decision maker to select a course of action that provides the maximum expected utility, has the least uncertainties, or makes a trade-off between risk and utility.
A yes / no decision analysis is a simple and straightforward analysis where each possibility has two outcomes, “yes” or “no,” both leading to separate consequences, and each consequence in turn leading to other possibilities with “yes” or “no” outcomes until the process reaches a solution. Such analysis differs from the “If” type of decisions where eventualities or possibilities depend on some other variable realizing, or decisions which involves selecting one or more alternatives from a pool of possible options.
This analysis usually finds representation in a decision tree or a flowchart, and follows the same standard methodology of decision analysis. The only difference is each node representing an “is” type of question and the branches out of a node restricted to just two possibilities.
Assume a simple dilemma of hiring temporary workers or asking regular staff to work overtime. A decision analysis seeks to identify the best course of action. Using the yes / no type of analysis requires asking questions, and each “yes” or “no” leading to further questions.
For example, weighing the options, the first question is “will hiring seasonal workers reduce labor costs compared to providing overtime?”
If “yes,” the next question is “are seasonal workers available?.” If “no,” the logical flow ends.
If “yes,” the next question is “is hiring seasonal workers legit?”. If “no,” the logical flow ends.
If yes, the next question is “are seasonal workers competent to do the job?”. Here, “yes” means hiring seasonal workers remains the best course of action, as no further deliberation is required or possible. If “no” the onus is on the decision maker to consider whether the benefits up to this stage perceives the perceived benefits of another alternative. A “no” answer might lead to another yes / no analysis, such as “Is training seasonal workers more cost-effective than providing overtime to existing workers.” If “no,” then the decision is abandoned, and the decision maker considers the option of overtime. If “yes” the next question arises “is on-the job training the best option.” If “yes,” the node ends here, and if “no,” the next question arises, “Is simulated exercise the best training option,” and so on.
Pros and Cons
The advantage of a yes / no decision analysis is the possibility of considering the merits of each possible element. For instance, in the example quoted above, it required a specific analysis of each type of training method and a decision on the best method through a process of elimination. In contrast, if the decision analysis entails making a list of all possible training options and selecting one, the decision maker may overlook or remain oblivious to some finer points of a particular type of training method.
This yeah or nay analysis, however, requires making subjective assumptions. For instance, in the example above, the decision analysis requires framing the first question as “will hiring seasonal workers reduce labor cost compared to providing overtime” rather than a more all-encompassing “which is the most cost-effective option to process the extra seasonal workload?” Similarly, later on, the analysis requires making a guess or subjective assumption of the best type of training program, and listing it first for consideration. The decision maker answering “yes” to the question “is on-the job training the best option” precludes consideration of any other training option, even with other options may have their own merits.
Yes or no straitjackets all answers into “yes” or “no,” even when some questions will not have a direct “yes” or “no” answer. Again, this type of analysis remains highly specific, and even a small change in assumptions or other variables can led to the entire analysis chain blow away. One “yes” changing to a “no” or vice versa requires changing all the other decisions below in a chain reaction.
- Carnegie MellonUniversity. “Decision Tree Learning.” https://www.cs.cmu.edu/afs/cs/project/theo-20/www/mlbook/ch3.pdf. Retrieved 20 June 2011.
- Harris, Roberts. “Introduction to Decision Making.” https://www.virtualsalt.com/crebook5.htm. Retrieved 20 June 2011.
Image Credit: freedigitalimages.net/renjith krishnan