Topic: Decision Making and Problem Solving
Authors: Herbert A. Simon et al.
Publication: Research Briefings 1986: Report of the Research Briefing Panel
on Decision Making and Problem Solving (1986)
What lies at the heart of everything that gets done is decision-making and problem-solving. Problem-solving includes fixing agendas, setting goals and designing actions and decision-making is evaluating and choosing the options thrown up by problem-solving actions. Both of these processes should happen effectively to address general and local problems.
In this age, it is not just humans but machines that hold the abilities and skills which make problem-solving and decision-making possible. How humans can use computers for enhancing how they make decisions and solve problems is one fertile avenue for further research and advances. In fact, much research has already been done and findings have been put to good use.
Subjective expected utility (SEU), a sophisticated mathematical model of choice, has informed much of our prescriptive knowledge on decision-making (not problem-solving). It is based on conditions of perfect utility-maximising rationality in a world of certainty. Empirical research, however, demonstrates that problem-solving is a selective and heuristic process given the limits on rationality and information. This is extremely crucial.
The real world of human-decisions is not a world of ideal-gases, frictionless planes, or vacuums. To bring it (decision-making) within the scope of human thinking powers, we must simplify our problem formulations drastically, even leaving out much or most of what is potentially relevant.
The growing relevance of descriptive theories is forcing prescriptive theories (SEU, for example) to amend their methods and assumptions. The elements of “unrealism” are being replaced by what is actually “attainable”. This alteration has strong implications for research in decision-making and problem-solving.
“Outline of current knowledge about decisison making and problem-solving”
SEU theory assumes a consistent utility function (a subjective ordering of preferences) and knowledge of the consequences of all the choices on that utility function. Based on these assumptions, it then seeks to determine how an actor would behave. This enables the marriage of subjective preferences and objective data.
The assumptions are very strong and they correspondingly lead to strong inferences. Most tools of modern operations research use SEU theory to determine the maximum that can be attained under certain given conditions.
The Limits of Rationality
SEU is extremely limitated when it comes to handling complex problems because complexity introduces uncertainty. It also makes enormous demands on information which is not forthcoming under most real world situations. The result is that study of actual decision-making processes have to substantially depart from the SEU framework.
Limited Rationality in Economic Theory
Predictions of economic behaviour based on the assumptions of perfect rationality and complete information give extremely different answers from those that assume limited rationality and incomplete information. The latter accounts for a bigger range of the behaviours that are seen in the economic arena. As such, while the assumption of profit maximisation is still acknowledged, what has changed is the understanding that profit maximisation is sought within the limits posed by incomplete and uncertain information.
The Theory of Games
SEU theory fails in situations where are conflicts of interests. Game theory is the most ambitious attempt to answers questions that are thrown up by conflicts of interests. The terms of the Prisoner’s Dilemma closely resemble those between organisations (nations, for example). And just as the game predicts, opposing parties tend to “satisfice” rather than to “optimise”.
Empirical Studies of Choice Under Uncertainty
- “When people are given information about the probabilities of certain events, and then are given some additional information as to which of the events has occurred, they tend to ignore the prior probabilities in favour of incomplete or even quite irrelevant information about the individual events.”
- “When asked to estimate the probability that 60 percent or more of the babies born in a hospital during a given week are male, people ignore information about the total number of births.”
- “There are instances in which people assess the frequency of a class by the ease with which instances can be brought to mind.”
- “When asked whether they would choose surgery in a hypothetical medical emergency, many more said they would when the chance of survival was given as 80 percent than when the chance of death was given as 20 percent.”
Methods of Empirical Research
All of these point to a need to improve research methodology. Some useful developments include the insistence on specific rather than general questions while keeping in mind the fact that how the question is phrased will have a significant bearing on the answer. Data obtained from the field is being supplemented by data obtained in the laboratory. Choice behaviour is studied as it happens not when it happens. Putting all these findings and techniques together in an empirically founded theory of decision making is what lies next.
Contemporary Problem-Solving Theory
Data gained from laboratories settings have been supplemented by field studies of professionals solving real-world problems in developing a problem-solving theory. The problem-solving process that has been understood from empirical studies can be described in the following manner.
Problem-solving involves a selective search through a wide range of possibilities using heuristics (or “rules of thumb”). This search is helped by procedures like “hill climbing” and “means-ends analysis” that allow the problem solver where to look next or what options to adopt as appropriate for the problem at hand. Problem-solving also depends on a large amount of information that the person doing it possesses.
Contemporary problem-solving theory thus accounts for “intuition and judgment” by locating them in the information and the inferential power that the researcher has. If they do not work, the researcher falls back to the processes of analysis.
Expert Systems in Artificial Intelligence
Research in artificial intelligence (AI) has benefited from and contributed to human problem solving. AI programs called expert systems have been built that resemble the typical human expert in terms of the information that they hold. While the computer programs are more analytic, the human experts will be more intuitive. The difference, however, is of quantity and not of kind.
Dealing with Ill-Structured problems
Complex ill-defined problems that have the capacity to be successively transformed in the course of the investigation are called ill-structured problems. An example is the problem facing an architect. Expert systems, in this area, have to not only know the design criteria but also know about them methods that will satisfy those criteria.
Setting the Agenda and Representing a Problem
Setting the agenda is important because resources are limited and not all problems can receive equal and sufficient attention. This first step in the problem-solving process remains poorly understood. The way a problem is represented depends a lot on the quality of solutions to be found. This is even less well understood.
Computation as Problem Solving
The use of computers for problem-solving has so far been substantial the domains of science and engineering. In fact, computation has become an object of explicit analysis itself along with the science it does. Computing power augmented by AI has successfully been deployed to chew through the incredible mass of data that is being produced by scientific instruments.
“A brief review of current research directions.”
Extensions of Theory
Decision Making Over Time
Tastes and priorities change over time. This makes the time dimension of decision extremely problematic.
The reality of varying societies or organisations makes it impossible to apply insights on problem solving and decision making across the board. How can this problem be resolved?
How does the behaviour of a person in his capacity as an individual differ from his behaviour as a member of an organisation? Also, while organisations tend to display a sophistication far beyond those of individuals, novelty situations lead to rather inappropriate responses.
From understanding how intelligent systems work, attention is now turning to how systems become intelligent. Learning is important for successful adaptation to an environment that is changing rapidly.
Current Research Programs
[This section outlines basic funding patterns as was current during the time of writing which is not germane to the current situation and, importantly, has little serious theoretical value.]
“Some of the principal research opportunities.”
Research Opportunities: Summary
- Empirical studies
- Decision making in organizational settings
- The resolution of conflicts of values (individual and group) and of inconsistencies in belief.
- Setting agendas and framing problems
 “In this game between two players, each has a choice between two actions, one trustful of the other player, the other mistrustful or exploitative. If both players choose the trustful alternative, both receive small rewards. If both choose the exploitative alternative, both are punished. If one chooses the trustful alternative and the other the exploitative alternative, the former is punished much more severely than in the previous case, while the latter receives a substantial reward. If the other player’s choice is fixed but unknown, it is advantageous for a player to choose the exploitative alternative, for this will give him the best outcome in either case. But if both adopt this reasoning, they will both be punished, whereas they could both receive rewards if they agreed upon the trustful choice (and did not welch on the agreement).”
 Hill-climbing and means-end problem solving are heuristic problem-solving strategies. In the hill-climbing heuristic, you simply choose the alternative that seem to lead most directly towards your goal state. In means-ends analysis, you divide the problem into a number of sub-problems (or sub-goals), and then you try to reduce the difference between the initial state and the goal state for each of the sub-problems. (For more: http://psychology.joelx.com/psych-355-hill-climbing/)