10 guiding principles

1. 10 Guiding Principles for the Design of Computer-Based Decision-Support Systems Over the past decade the work of the CAD Research Center has focused on ‘complex problems’ and computer-based decision systems that are designed to assist, not replace, human decision makers. The relative level of complexity of a problem is primarily a function of the number and strengths of the inter and intra relationships that exist among internal and external components of the problem, and the degree of uncertainty that surrounds the definition of these components. Today, such complex problems are commonly found in many planning, operational, and management areas in a diverse range of fields such as transportation, economics, engineering design and manufacturing, military command and control, medicine, and law. A successful decision-support system is one that assists rather than replaces the human decision maker. Human beings and computers are complementary in many respects. The strengths of human decision makers in the areas of conceptualization, intuition and creativi- ty are the weaknesses of the computer. Conversely, the strengths of the computer in computation speed, parallelism, accuracy and the persistent storage of almost unlimited detailed information are human weaknesses. It therefore makes a great deal of sense to view a decision-support system as a partnership between human and computer-based resources and capabilities. Automation should be restricted to the monitoring of problem solving activities, the detection of conflicts, and the execution of evaluation, search and planning sequences. In this partnership a high level of interaction between the user and the computer is a necessary feature of the decision-support environment. It provides opportunities for the user to guide the computer in those areas of the decision making process, such as conceptualization and intuition, where the skills of the user are likely to be far superior to those of the computer. Particularly prominent among these areas are conflict resolution and risk assessment. CADRCCAD RESEARCH CENTER relationships that appear to have precipitated the conflict. Tracing these relationships may produce more progress toward a solution than the resolution of the conflict itself. The importance of a high degree of interaction between the user(s) and the various components of the decision-support system is integral to most of the principles described here. This interaction is facilitated by two system characteristics: a high level object representation; and, an intuitive user interface. The user interface should be graphical in nature. The human cognitive system excels in pattern matching. Words and numbers require the performance of a translation task that is relatively time consuming, subject to information loss, and carries with it the potential for confusion and misinterpretation. An on-line help system should be available to both assist the user in the execution of operational sequences and provide explanations of system activities. The latter should include exploration of the recommenda- tions, evaluation results and proposals contributed by the various components (e.g., agents) of the system. In the past it has been considered helpful, as a means of simplifying complex problems, to treat planning and execution as distinct activities. Under this school of thought the purpose of planning is to clearly define and analyze the problem, and then develop a solution as a course of action that can be implemented during the execution stage. However, as the complexity and tempo of problem solving situations increases, these apparently distinct functional areas can no longer be categorized as discrete operational spheres of activity. They tend to merge into a single integrated functional pool of capabilities from which the human decision maker draws assistance as necessary. In such problem solving situations continuous information changes require constant replanning, even during those phases when the need for action and execution overshadows all other activities. This is particularly apparent in the military field, but equally relevant in management, marketing and manufacturing situations where changing conditions require the most thorough and carefully laid out plans to be spontaneously reformulated. For example, in military missions the impacts of enemy actions dictate the need for continuous replanning and training during execution. Under these conditions functional integration is essential. Not only must the planning functions be accessible from the same computer system, but they must be able to operate on the same information that applies to the execution functions. Similarly, in the manufacturing fields changing production conditions such as equipment failures and material supply delays may require significant modification of the original design that may border on a complete redesign. These design modifications have to be accomplished while production operations are in progress. In a distributed, cooperative decision-support system architecture the necessary level of integration has the potential to be achieved, since functional modules and information resources are treated as sharable components. In such a shared environment distributed databases may be accessed by any of the functional components whenever the need arises and the necessary authorizations are available. The ability to switch from one functional mode to another then becomes largely a function of the user interface and does not require the user to move out of the current application environment. In other words, the physical separation of individual computer-based components need not exist at the logical level of the user interface. DECISION-SUPPORT SYSTEMS Principle 1: Emphasis on Partnership Principle 9: The Computer-User Interface Principle 10: Functional Integration

2. Complex problem environments normally involve many parties that collaborate from widely distributed geographical locations and utilize information resources that are equally dispersed. The decision-support system can take advantage of the distributed participation by itself assuming a distributed architecture. Such an architecture typically consists of several components that can execute on more than one computer. Both the information flow between these components and the computing power required to support the system as a whole can be decentralized. This greatly reduces the potential for communication bottlenecks and increases the computation speed through parallelism. Another advantage of the distributed approach is the ability to modify some components of the system while the system as a whole continues to operate with the remaining components. Similarly, the malfunc- tion or complete failure of one component does not necessarily jeopardize the entire system. This is not so much a matter of redundancy, although the distributed architecture lends itself to the provision of a high degree of redundancy, but rather a direct result of the physical independence of the components. While the components may be closely integrated from a logical point of view they can operate in their own autonomous physical environment. The high degree of uncertainty that pervades complex problem environments extends beyond the decision making activity of the collaborating problem solvers to the configuration of the decision-support system itself. The components of the system are likely to change over time, through modification, replacement, deletion and extension. It should be possible to implement these changes in a seamless fashion through common application programming interfaces and shared databases. The decision-support systems should be designed as a set of tools rather than as solutions to a predeter- mined set of problems. The indeterminate nature of complex problems does not allow us to predict, with any degree of certainty, either the specific circumstances of a future problem situation or the precise terms of the solution. Under these circumstances it is far more constructive to provide tools that will extend the capabilities of the human decision maker in a highly interactive problem solving environment. In this sense a tool is defined more broadly than a sequence of algorithms, heuristics or procedures that are applied largely on the direction of a user. Tools can be self-activating, be capable of at least semi- autonomous behavior, and cooperate with each other and users in requesting and providing services. The ability of a decision-support system to have some level of understanding of the meaning of the information it processes is the single most important prerequisite for a cooperative and collaborative problem solving environment. A high level representation of the real world objects that define the problem system forms the basis of the interactions between the users and the system and, also, the degree of intelligence that can be embedded in its components. For example, it is virtually impossible to build a useful computer-based tool that can provide meaningful assistance to a military commander in the analysis of the physical battlefield if the battlefield terrain is represented in the computer in terms of ‘x,y’ coordinates and pix- els. To the commander the battlefield consists of real world objects, such as mountains, roads, rivers, trees, observation posts, buildings, and so on. Each of these objects has attributes that determine its behavior under certain conditions. These semantic descriptors form the basis of collaboration among human problem solvers, and are likewise the fundamental unit of communication in a computer-based decision- support environment. The decision-support system should be a knowledge-based system. In this context knowledge can be described as experience derived from observation and interpretation of past events or phenomena, and the application of methods to past situations. Knowledge bases capture this experience in the form of rules, case studies, standard practices, and typical descriptions of objects and object systems that can serve as prototypes. Problem solvers typically manipulate these prototypes through adaptation, refinement, muta- tion, analogy, and combination, as they apply them to the solution of current problems. The decision-support system need not, and should not, exercise centralized control over the decision making environment. Much of the decision making activity can be localized. For example, components of the system (e.g., mentor-agents) that are responsible for pursuing the interests of real world objects, such as soldiers in military applications and technical and management personnel in commercial and industrial applications, can achieve many of their objectives through service requests and negotiations that involve only a few nodes of the problem system. This greatly reduces the propensity for the formation of communication bottlenecks and at the same time increases the amount of parallel activity in the system. The ability to combine in a computer-based decision-support system many types of semi-autonomous and autonomous components (i.e., agents), representing a wide range of interests and incorporating different kinds of knowledge and capabilities, provides the system with a great deal of versatility and potential for problem solving to occur simultaneously at several levels of granularity. This is similar to human problem solving teams in which individual team members work concurrently on different aspects of the problem and communicate in pairs and small groups as they gather information and explore sub-problems. The decision-support system should focus on the identification rather than the automatic resolution of conflicts. This notion gains in importance as the level of complexity of the problem system increases. The resolution of even mundane conflicts can provide subtle opportunities for advancing toward solution objectives. These opportunities are more likely to be recognized by a human decision maker than a computer- based agent. The identification of conflicts is by no means a trivial undertaking. It includes not only the ability to recognize that a conflict actually exists, but also the determination of the kind of conflict and the Principle 2: Cooperative and Distributed Principle 3: An Open Architecture Principle 4: Tools, not Solutions Principle 5: High Level Internal Representation Principle 7: Decentralized Decision Making Principle 8: Emphasis on Conflict Identification Principle 6: Embedded Knowledge

3. Complex problem environments normally involve many parties that collaborate from widely distributed geographical locations and utilize information resources that are equally dispersed. The decision-support system can take advantage of the distributed participation by itself assuming a distributed architecture. Such an architecture typically consists of several components that can execute on more than one computer. Both the information flow between these components and the computing power required to support the system as a whole can be decentralized. This greatly reduces the potential for communication bottlenecks and increases the computation speed through parallelism. Another advantage of the distributed approach is the ability to modify some components of the system while the system as a whole continues to operate with the remaining components. Similarly, the malfunc- tion or complete failure of one component does not necessarily jeopardize the entire system. This is not so much a matter of redundancy, although the distributed architecture lends itself to the provision of a high degree of redundancy, but rather a direct result of the physical independence of the components. While the components may be closely integrated from a logical point of view they can operate in their own autonomous physical environment. The high degree of uncertainty that pervades complex problem environments extends beyond the decision making activity of the collaborating problem solvers to the configuration of the decision-support system itself. The components of the system are likely to change over time, through modification, replacement, deletion and extension. It should be possible to implement these changes in a seamless fashion through common application programming interfaces and shared databases. The decision-support systems should be designed as a set of tools rather than as solutions to a predeter- mined set of problems. The indeterminate nature of complex problems does not allow us to predict, with any degree of certainty, either the specific circumstances of a future problem situation or the precise terms of the solution. Under these circumstances it is far more constructive to provide tools that will extend the capabilities of the human decision maker in a highly interactive problem solving environment. In this sense a tool is defined more broadly than a sequence of algorithms, heuristics or procedures that are applied largely on the direction of a user. Tools can be self-activating, be capable of at least semi- autonomous behavior, and cooperate with each other and users in requesting and providing services. The ability of a decision-support system to have some level of understanding of the meaning of the information it processes is the single most important prerequisite for a cooperative and collaborative problem solving environment. A high level representation of the real world objects that define the problem system forms the basis of the interactions between the users and the system and, also, the degree of intelligence that can be embedded in its components. For example, it is virtually impossible to build a useful computer-based tool that can provide meaningful assistance to a military commander in the analysis of the physical battlefield if the battlefield terrain is represented in the computer in terms of ‘x,y’ coordinates and pix- els. To the commander the battlefield consists of real world objects, such as mountains, roads, rivers, trees, observation posts, buildings, and so on. Each of these objects has attributes that determine its behavior under certain conditions. These semantic descriptors form the basis of collaboration among human problem solvers, and are likewise the fundamental unit of communication in a computer-based decision- support environment. The decision-support system should be a knowledge-based system. In this context knowledge can be described as experience derived from observation and interpretation of past events or phenomena, and the application of methods to past situations. Knowledge bases capture this experience in the form of rules, case studies, standard practices, and typical descriptions of objects and object systems that can serve as prototypes. Problem solvers typically manipulate these prototypes through adaptation, refinement, muta- tion, analogy, and combination, as they apply them to the solution of current problems. The decision-support system need not, and should not, exercise centralized control over the decision making environment. Much of the decision making activity can be localized. For example, components of the system (e.g., mentor-agents) that are responsible for pursuing the interests of real world objects, such as soldiers in military applications and technical and management personnel in commercial and industrial applications, can achieve many of their objectives through service requests and negotiations that involve only a few nodes of the problem system. This greatly reduces the propensity for the formation of communication bottlenecks and at the same time increases the amount of parallel activity in the system. The ability to combine in a computer-based decision-support system many types of semi-autonomous and autonomous components (i.e., agents), representing a wide range of interests and incorporating different kinds of knowledge and capabilities, provides the system with a great deal of versatility and potential for problem solving to occur simultaneously at several levels of granularity. This is similar to human problem solving teams in which individual team members work concurrently on different aspects of the problem and communicate in pairs and small groups as they gather information and explore sub-problems. The decision-support system should focus on the identification rather than the automatic resolution of conflicts. This notion gains in importance as the level of complexity of the problem system increases. The resolution of even mundane conflicts can provide subtle opportunities for advancing toward solution objectives. These opportunities are more likely to be recognized by a human decision maker than a computer- based agent. The identification of conflicts is by no means a trivial undertaking. It includes not only the ability to recognize that a conflict actually exists, but also the determination of the kind of conflict and the Principle 2: Cooperative and Distributed Principle 3: An Open Architecture Principle 4: Tools, not Solutions Principle 5: High Level Internal Representation Principle 7: Decentralized Decision Making Principle 8: Emphasis on Conflict Identification Principle 6: Embedded Knowledge

4. 10 Guiding Principles for the Design of Computer-Based Decision-Support Systems Over the past decade the work of the CAD Research Center has focused on ‘complex problems’ and computer-based decision systems that are designed to assist, not replace, human decision makers. The relative level of complexity of a problem is primarily a function of the number and strengths of the inter and intra relationships that exist among internal and external components of the problem, and the degree of uncertainty that surrounds the definition of these components. Today, such complex problems are commonly found in many planning, operational, and management areas in a diverse range of fields such as transportation, economics, engineering design and manufacturing, military command and control, medicine, and law. A successful decision-support system is one that assists rather than replaces the human decision maker. Human beings and computers are complementary in many respects. The strengths of human decision makers in the areas of conceptualization, intuition and creativi- ty are the weaknesses of the computer. Conversely, the strengths of the computer in computation speed, parallelism, accuracy and the persistent storage of almost unlimited detailed information are human weaknesses. It therefore makes a great deal of sense to view a decision-support system as a partnership between human and computer-based resources and capabilities. Automation should be restricted to the monitoring of problem solving activities, the detection of conflicts, and the execution of evaluation, search and planning sequences. In this partnership a high level of interaction between the user and the computer is a necessary feature of the decision-support environment. It provides opportunities for the user to guide the computer in those areas of the decision making process, such as conceptualization and intuition, where the skills of the user are likely to be far superior to those of the computer. Particularly prominent among these areas are conflict resolution and risk assessment. CADRCCAD RESEARCH CENTER relationships that appear to have precipitated the conflict. Tracing these relationships may produce more progress toward a solution than the resolution of the conflict itself. The importance of a high degree of interaction between the user(s) and the various components of the decision-support system is integral to most of the principles described here. This interaction is facilitated by two system characteristics: a high level object representation; and, an intuitive user interface. The user interface should be graphical in nature. The human cognitive system excels in pattern matching. Words and numbers require the performance of a translation task that is relatively time consuming, subject to information loss, and carries with it the potential for confusion and misinterpretation. An on-line help system should be available to both assist the user in the execution of operational sequences and provide explanations of system activities. The latter should include exploration of the recommenda- tions, evaluation results and proposals contributed by the various components (e.g., agents) of the system. In the past it has been considered helpful, as a means of simplifying complex problems, to treat planning and execution as distinct activities. Under this school of thought the purpose of planning is to clearly define and analyze the problem, and then develop a solution as a course of action that can be implemented during the execution stage. However, as the complexity and tempo of problem solving situations increases, these apparently distinct functional areas can no longer be categorized as discrete operational spheres of activity. They tend to merge into a single integrated functional pool of capabilities from which the human decision maker draws assistance as necessary. In such problem solving situations continuous information changes require constant replanning, even during those phases when the need for action and execution overshadows all other activities. This is particularly apparent in the military field, but equally relevant in management, marketing and manufacturing situations where changing conditions require the most thorough and carefully laid out plans to be spontaneously reformulated. For example, in military missions the impacts of enemy actions dictate the need for continuous replanning and training during execution. Under these conditions functional integration is essential. Not only must the planning functions be accessible from the same computer system, but they must be able to operate on the same information that applies to the execution functions. Similarly, in the manufacturing fields changing production conditions such as equipment failures and material supply delays may require significant modification of the original design that may border on a complete redesign. These design modifications have to be accomplished while production operations are in progress. In a distributed, cooperative decision-support system architecture the necessary level of integration has the potential to be achieved, since functional modules and information resources are treated as sharable components. In such a shared environment distributed databases may be accessed by any of the functional components whenever the need arises and the necessary authorizations are available. The ability to switch from one functional mode to another then becomes largely a function of the user interface and does not require the user to move out of the current application environment. In other words, the physical separation of individual computer-based components need not exist at the logical level of the user interface. DECISION-SUPPORT SYSTEMS Principle 1: Emphasis on Partnership Principle 9: The Computer-User Interface Principle 10: Functional Integration

10 guiding principles

More Related Content

What's hot (16)

Viewers also liked (14)

Similar to 10 guiding principles (20)

10 guiding principles