Decision Theory
Decision making under risk and uncertainty is treated in many statistical writings, but they are not used in day-to-day decision making. In fact, most decision makers are uninformed regarding these "basic and standard" concepts.
Decision theory depends on our belief in the validity of deductive and inductive reasoning, which is covered by epistemology.
References
Kahneman, Daniel; Slovic, Paul; Tversky, Amos, editors. Judgment under Uncertainty: Heuristics and Biases. Cambridge University Press; 1982. 544 p. ISBN: 0521284147.
Taleb, Nassim Nicholas. Fooled by Randomness: The Hidden Role of Chance in the Markets and in Life. 2nd ed. Texere: 2004. 312 p. ISBN: 158799190X.
Bush, Vannevar. As we may think. The Atlantic Monthly 176, 1 (July 1945), 641-649.
J^T Thoughts
Intuition is a great decision basis for some decisions, but demonstrated to be horrible in other areas. Decision makers should learn when to use it and when to call for more rigorous methods.
Standard debate techniques, or basic rules of logic, should be applied throughout an organization's thought process. Unfortunately, most people are poorly trained in how to use their mind.
Familiarity with the basics behavior of samples from populations (i.e. statistics) is needed to draw valid conclusions from observations, understand confidence levels, and spot liars.
This vital knowledge should be ingrained to everyone from middle school on, but is missing from nearly everyone. How do we fix this critical learning deficiency?
Decision Making Augmentation
How can decision making be strengthened and clarified by technology?
Humans’ capability to make sound decisions is critical to our individual survival and survival as a species. Despite its criticality, our decision faculty is inconsistent. It is capable of correct “snap” judgments in many circumstances, but in other cases it is subject to biases and fallacies. Yet, there is scant technological support for the actual decision making act.
Computing technology has provided dramatically useful abilities to perform computational work and to capture and find information. For example, Google has far exceeded the capabilities of Vannevar Bush’s memex envisioned in his renowned 1945 essay.
In the same essay, Bush mentions “We may some day click off arguments on a machine with the same assurance that we now enter sales on a cash register.” That day has not yet arrived.
Suppose that we relax the goal to create systems that reason completely automatically, and instead consider systems that augment humans’ decision capabilities. These systems would operate in a supporting role, helping users to structure and validate assertions.
The function of such systems would focus on avoiding known common fallacies. For example, intuitive probability estimates of individual events may be accurate in areas of a user’s experience. However, intuition often is misleading for probability of combinations of events and for conditional probabilities. These systems could aid their users by computing these type of composite probabilities.
Arguments processed in such a system would be composed of a mix of user-supplied assertions along with system generated or validated conclusions. Components of an argument could be marked as rigorous when system-validated and other components as non-rigorous if the user determines validation is not desirable presently. Representing the entire structure of an argument, even parts not rigorously validated, would help maintain the integrity of the matter under consideration.
For the user, the system could be presented the metaphor of a workshop in which one crafts a valid argument using tools such as the composite probability calculator mentioned above. Tools would be constructed based on areas of human weaknesses, as identified by decision science and psychology research.
...[continue? or, is this enough?]...
Related work has been ongoing in a number of fields, for example:
- Knowledge representation efforts such as the OWL Web Ontology Language provide a means to represent assertions in a machine-processable form.
- Automated reasoning work has produced methods and systems to reach logical conclusions without human assistance.
- Modal logic and probabilistic logic enable formal reasoning about a wider range of situations then standard first order logic.
- Decision science, behavioral economics, and judgement psychology have documented human foibles and biases in many decision situations.
...[Develop a sketch of the research work program here]...
...[Specifically, the intent is to integrate with humans better, not to develop new decision theory]...
Obviously, improving decision making has broad applicability. Here are a few examples that fit this approach:
- Business strategy is full of “hunch” and anecdote based choices, when millions of dollars are at stake. Management fads are symptomatic of the desire for improved structure to the strategic decision process.
- It’s amusing to consider applying some intellectual rigor to politics. Certainly, the candidates wouldn’t be interested in these tools, but they would be helpful in the discussions among the citizenry.
- Security....
- In medicine, reasoning about diagnoses and treatment is a mix of doctors’ experienced intuition and [probability/consequences/risk -- needs help]....
- ...[others?]...