DECISION SUPPORT SYSTEMS:GROUP AND ORGANIZATIONAL DECISION SUPPORT SYSTEMS

GROUP AND ORGANIZATIONAL DECISION SUPPORT SYSTEMS

A group decision support system (GDSS) is an information technology-based support system de- signed to provide decision making support to groups and / or organizations. This could refer to a group meeting at one physical location at which judgments and decisions are made that affect an organization or group. Alternatively, it could refer to a spectrum of meetings of one or more indi- viduals, distributed in location, time, or both. GDSSs are often called organizational decision support systems, and other terms are often used, including executive support systems (ESSs), which are information technology-based systems designed to support executives and managers, and command and control systems, which is a term often used in the military for a decision support system. We will generally use GDSS to describe all of these.

Managers and other knowledge workers spend much time in meetings. Much research into meeting effectiveness suggests that it is low, and proposals have been made to increase this through infor- mation technology support (Johansen 1988). Speciﬁc components of this information technology- based support might include computer hardware and software, audio and video technology, and communications media. There are three fundamental ingredients in this support concept: technological support facilities, the support processes provided, and the environment in which they are embedded. Kraemer and King (1988) provide a noteworthy commentary on the need for group efforts in their overview of GDSS efforts. They suggest that group activities are economically necessary, efﬁcient as a means of production, and reinforcing of democratic values.

There are a number of predecessors for group decision support technology. Decision rooms, or situation rooms, where managers and boards meet to select from alternative plans or courses of action, are very common. The ﬁrst computer-based decision support facility for group use is attributed to Douglas C. Engelbart, the inventor of the (computer) mouse, at Stanford in the 1960s. A discussion of this and other early support facilities is contained in Johansen (1988).

Engelbart’s electronic boardroom-type design is acknowledged to be the ﬁrst type of information technology-based GDSS. The electronic format was, however, preceded by a number of nonelectronic formats. The Cabinet war room of Winston Churchill is perhaps the most famous of these. Maps placed on the wall and tables for military decision makers were the primary ingredients of this room. The early 1970s saw the introduction of a number of simple computer-based support aids into situ- ation rooms. The ﬁrst system that resembles the GDSS in use today is often attributed to Gerald Wagner, the Chief Executive Ofﬁcer of Execucom, who implemented a planning laboratory made up of a U-shaped table around which people sat, a projection TV system for use as a public viewing screen, individual small terminals and keyboards available to participants, and a minicomputer to which the terminals and keyboards were connected. This enabled participants to vote and to conduct simple spreadsheet-like exercises. Figure 15 illustrates the essential features of this concept. Most present-day GDSS centralized facilities look much like the conceptual illustration of a support room, or situation room, shown in this Figure.

As with a single-user DSS, appropriate questions for a GDSS that have major implications for design concern the perceptions and insights that the group obtains through use of the GDSS and the

activities that can be carried out through its use. Also, additional concerns arise regarding the public screen, interactions between the public screen and individual screens, the characteristics of individual work screens, and contingency task structural variables associated with the individuals in the group using the GDSS (Gray and Olfman 1989).

Huber (1982) has indicated both the needs for GDSS and how an appropriately designed GDSS can meet these needs. He identiﬁes four interacting and complicating concerns:

1. Effective decision making requires not only obtaining an appropriate decision, but also ensuring that participants are happy with the process used to reach the decision and will be willing to meet and work cooperatively in the future.

2. Productivity losses occur because of dominant individuals and group pressures that lead to conformity of thought, or groupthink.

3. Miscommunications are common in group situations.

4. Insufﬁcient time is often spent in situation assessment, problem exploration, and generation of alternative courses of action.

Huber further indicates that a GDSS can help improve the unaided decision situation, which often suffers from imperfect information processing and suboptimal decision selection.

A GDSS is made up of:

1. Technological components, in terms of computer hardware and software, and communication equipment

2. Environmental components, in terms of the people involved, their locations in time and space, and their experiential familiarity with the task at hand

3. Process components, or variables made up of the conventions used to support task performance and enable the other components of decision making to function appropriately.

We have already described the technological design features of DSSs. Thus, our commentary on technological design features of a GDSS will be brief. We do need to provide a perspective on groups and organizations, and we will do this in our next two sections. Then we will turn to some architec- tural considerations speciﬁcally relevant to GDSS.

Information Needs for Group and Organizational Decision Making

The nature of the decisions and the type of information required differ across each of these four levels identiﬁed in Figure 1. Generally, operational activities occur much more frequently than stra- tegic planning activities. Also, there is a difference in the degree to which the knowledge required for each of these levels is structured. In 1960, Herbert Simon (Simon 1960) described decisions as structured or unstructured, depending upon whether the decision making process can be explicitly described prior to the time when it is necessary to make a decision. This taxonomy would seem to lead directly to that in which expert skills (holistic reasoning), rules (heuristics), or formal reasoning (holistic analysis) are normatively used for judgment. Generally, operational performance decisions are much more likely than strategic planning decisions to be prestructured. This gives rise to a number of questions concerning efﬁciency and effectiveness tradeoffs between training and aiding (Rouse 1991) that occur at these levels.

There are a number of human abilities that a GDSS should augment.

1. It should help the decision maker to formulate, frame, or assess the decision situation. This includes identifying the salient features of the environment, recognizing needs, identifying appropriate objectives by which we are able to measure successful resolution of an issue, and generating alternative courses of action that will resolve the needs and satisfy objectives.

2. It should provide support in enhancing the abilities of the decision maker to obtain and analyze the possible impacts of the alternative courses of action.

3. It should have the capability to enhance the decision maker’s ability to interpret these impacts in terms of objectives. This interpretation capability will lead to evaluation of the alternatives and selection of a preferred alternative option.

Associated with each of these three formal steps of formulation, analysis, and interpretation must be the ability to acquire, represent, and utilize information and associated knowledge and the ability to implement the chosen alternative course of action.

Many attributes will affect the quality and usefulness of the information that is obtained, or should be obtained, relative to any given decision situation. These variables are very clearly contingency task dependent. Among these attributes are the following (Keen and Scott Morton 1978).

• Inherent and required accuracy of available information: Operational control and performance situations will often deal with information that is relatively accurate. The information in strategic planning and management control situations is often inaccurate.

• Inherent precision of available information: Generally, information available for operational control and operational performance decisions is very imprecise.

• Inherent relevancy of available information: Operational control and performance situations will often deal with information that is fairly relevant to the task at hand because it has been prepared that way by management. The information in strategic planning and management control situ- ations is often obtained from the external environment and may be irrelevant to the strategic tasks at hand, although it may not initially appear this way.

• Inherent and required completeness of available information: Operational control and perform- ance situations will often deal with information that is relatively complete and sufﬁcient for operational performance. The information in strategic planning and management control situa- tions is often very incomplete and insufﬁcient to enable great conﬁdence in strategic planning and management control.

• Inherent and required veriﬁability of available information: Operational control and perform- ance situations will often deal with information that is relatively veriﬁable to determine cor- rectness for the intended purpose. The information in strategic planning and management control situations is often unveriﬁable, or relatively so, and this gives rise to a potential lack of conﬁ- dence in strategic planning and management control.

• Inherent and required consistency and coherency of available information: Operational control and performance situations will often deal with information that is relatively consistent and coherent. The information in strategic planning and management control situations is often inconsistent and perhaps even contradictory or incoherent, especially when it comes from mul- tiple external sources.

• Information scope: Generally, but not always, operational decisions are made on the basis of relatively narrow scope information related to well-deﬁned events that are internal to the or- ganization. Strategic decisions are generally based upon broad-scope information and a wide range of factors that often cannot be fully anticipated prior to the need for the decision.

• Information quantiﬁability: In strategic planning, information is very likely to be highly qual- itative, at least initially. For operational decisions, the available information is often highly quantiﬁed.

• Information currency: In strategic planning, information is often rather old, and it is often difﬁcult to obtain current information about the external environment. For operational control decisions, very current information is often needed and present.

• Needed level of detail: Often very detailed information is needed for operational decisions. Highly aggregated information is often desired for strategic decisions. There are many difﬁcul- ties associated with information summarization that need attention.

• Time horizon for information needed: Operational decisions are typically based on information over a short time horizon, and the nature of the control may change very frequently. Strategic decisions are based on information and predictions based on a long time horizon.

• Frequency of use: Strategic decisions are made infrequently, although they are perhaps reﬁned fairly often. Operational decisions are made quite frequently and are relatively easily changed.

• Internal or external information source: Operational decisions are often based upon information that is available internal to the organization, whereas strategic decisions are much more likely to be dependent upon information content that can only be obtained external to the organization.

These attributes, and others, could be used to form the basis for an evaluation of information quality in a decision support system.

Information is used in a DSS for a variety of purposes. In general, information is equivalent to, or may be used as, evidence in situations in which it is relevant. Often information is used directly as a basis for testing an hypothesis. Sometimes it is used indirectly for this purpose. There are three different conditions for describing hypotheses (Schum 1987, 1994):

1. Different alternative hypotheses or assessments are possible if evidence is imperfect in any way. A hypothesis may be imperfect if it is based on imperfect information. Imperfect infor- mation refers to information that is incomplete, inconclusive, unreliable, inconsistent, or un- certain. Any or all of these alternate hypotheses may or may not be true.

2. Hypotheses may refer to past, present, or future events.

3. Hypotheses may be sharp (speciﬁc) or diffuse (unspeciﬁed). Sharp hypotheses are usually based on speciﬁc evidence rather than earlier diffuse hypotheses. An overly sharp hypothesis may contain irrelevant detail and invite invalidation by disconﬁrming evidence on a single issue in the hypothesis. An overly diffuse hypothesis may be judged too vague and too uninteresting by those who must make a decision based upon the hypothesis, even though the hypothesis might have been described in a more cogent manner.

The support for any hypothesis can always be improved by either revising a portion of the hypothesis to accommodate new evidence or gathering more evidence that infers the hypothesis. Hypotheses can be potentially signiﬁcant for four uses:

1. Explanations: An explanation usually involves a model, which can be elaborate or simple. The explanation consists of the rationale for why certain events occurred.

2. Event predictions: In this case the hypothesis is proposed for a possible future event. It may include the date or period when the possible event will occur.

3. Forecasting and estimation: This involves the generation of a hypothesis based on data that does not exist or that is inaccessible.

4. Categorization: Sometimes it is useful to place persons, objects, or events into certain cate- gories based upon inconclusive evidence linking the persons, objects, or events to these cate- gories. In this case the categories represent hypotheses about category membership.

Assessment of the validity of a given hypothesis is inductive in nature. The generation of hy- potheses and determination of evidence relevant to these hypotheses involve deductive and abductive reasoning. Hypotheses may be generated on the basis of the experience and prior knowledge that leads to analogous representations and recognitional decision making, as noted by Klein (1990, 1998).

Although no theory has been widely accepted on how to quantify the value of evidence, it is important to be able to support a hypothesis in some logical manner. Usually there is a major hypothesis that is inferred by supporting hypotheses, and each of these supporting hypotheses is inferred by its supporting hypothesis, and so on. Evidence is relevant to the extent that it causes one to increase or decrease the likeliness of an existing hypothesis, or modify an existing hypothesis, or create a new hypothesis. Evidence is direct if it has a straightforward bearing on the validity of the main hypothesis. Evidence is indirect if its effect on the main hypothesis is inferred through at least one other level of supporting hypothesis.

In many cases, it is necessary to acquire, represent, use, and / or communicate knowledge that is imperfect. This is especially important in group decision situations. In describing the structure of the beliefs and the statements that people make about issues that are of importance to them, the nature of the environment that surrounds them, as well as the ways in which people reason and draw conclusions about the environment and issues that are embedded into the environment, especially when there are conﬂicting pieces of information and opinions concerning these, people often attempt to use one or more of the forms of logical reasoning. Many important works deal with this subject. Of particular interest here is the work of Toulmin and his colleagues (Toulmin et al. 1979), who have described an explicit model of logical reasoning that is subject to analytical inquiry and computer

implementation. The model is sufﬁciently general that it can be used to represent logical reasoning in a number of application areas.

Toulmin assumes that whenever we make a claim, there must be some ground on which to base our conclusion. He states that our thoughts are generally directed, in an inductive manner, from the grounds to the claim, each of which are statements that may be used to express both facts and values. As a means of explaining observed patterns of stating a claim, there must be some reason that can be identiﬁed with which to connect the grounds and the claim. This connection, called the warrant, gives the grounds–claim connection its logical validity.

We say that the grounds support the claim on the basis of the existence of a warrant that explains the connection between the grounds and the claim. It is easy to relate the structure of these basic elements with the process of inference, whether inductive or deductive, in classical logic. The warrants are the set of rules of inference, and the grounds and claim are the set of well-deﬁned propositions or statements. It will be only the sequence and procedures, as used to formulate the three basic elements and their structure in a logical fashion, that will determine the type of inference that is used.

Sometimes, in the course of reasoning about an issue, it is not enough that the warrant will be the absolute reason to believe the claim on the basis of the grounds. For that, there is a need for further backing to support the warrant. It is this backing that provides for reliability, in terms of truth, associated with the use of a warrant. The relationship here is analogous to the way in which the grounds support the claim. An argument will be valid and will give the claim solid support only if the warrant is relied upon and is relevant to the particular case under examination. The concept of logical validity of an argument seems to imply that we can make a claim only when both the warrant and the grounds are certain. However, imprecision and uncertainty in the form of exceptions to the rules or low degree of certainty in both the grounds and the warrant do not prevent us on occasion from making a hedge or, in other words, a vague claim. Often we must arrive at conclusions on the basis of something less than perfect evidence, and we put those claims forward not with absolute and irrefutable truth but with some doubt or degree of speculation.

To allow for these cases, modal qualiﬁers and possible rebuttals may be added to this framework for logical reasoning. Modal qualiﬁers refer to the strength or weakness with which a claim is made. In essence, every argument has a certain modality. Its place in the structure presented so far must reﬂect the generality of the warrants in connecting the grounds to the claim. Possible rebuttals, on the other hand, are exceptions to the rules. Although modal qualiﬁers serve the purpose of weakening or strengthening the validity of a claim, there may still be conditions that invalidate either the grounds or the warrants, and this will result in deactivating the link between the claim and the grounds. These cases are represented by the possible rebuttals.

The resulting structure of logical reasoning provides a very useful framework for the study of human information processing activities. The order in which the six elements of logical reasoning have been presented serves only the purpose of illustrating their function and interdependence in the structure of an argument about a speciﬁc issue. It does not represent any normative pattern of ar- gument formation. In fact, due to the dynamic nature of human reasoning, the concept formation and framing that result in a particular structure may occur in different ways. The six-element model of logical reasoning is shown in Figure 16.

Computer-based implementations of Figure 16 may assume a Bayesian inferential framework for processing information. Frameworks for Bayesian inference require probability values as primary inputs. Because most events of interest are unique or little is known about their relative frequencies of occurrence, the assessment of probability values usually requires human judgment. Substantial

psychological research has shown that people are unable to elicit probability values consistent with the rules of probabilities or to process information as they should, according to the laws of probability, in revising probability assessment when new information is obtained. For example, when people have both causal and diagnostic implications, they should weigh the causal and diagnostic impacts of the evidence. More often, however, unaided humans will reach judgments that suggest they apparently assess conditional probabilities primarily in terms of the direct causal effect of the impacts. If A is perceived to be the cause of B, for example, people will usually associate higher probabilities with P(BIA) than they will with P(AIB). Studies concerning the elicitation of probability values often report that individuals found it easier and showed more conﬁdence in assessing P(AIB) if B was causal to

A. This strongly suggests that the choice of which form of inference to invoke depends more on the level of familiarity of the observer with the task at hand and the frame adopted to initially represent the knowledge. Again, this supports the wisdom of appropriate structuring of decision situations.

In general, grounds can be categorized by the several means in which are warranted:

• Empirical observation

• Expert judgment

• Enumerative induction (statistics)

• Experiment (hypothesis test)

• Direct fact

A decision assessment system based on this concept is described in Janssen and Sage (2000) together with application to the issues in public policy decision making associated with agricultural pest control. This system can support several user groups, who can use the support system to state ar- guments for or against an important policy issue and to assist in identifying and evaluating alternatives for implementation as policy decisions. The resulting decision support system assists in improving the clarity of the lines of reasoning used in speciﬁc situations; the warrants, grounds, and backings that are used to support claims and speciﬁc lines of reasoning; and the contradictions, rebuttals, and arguments surrounding each step in the reasoning process associated with evaluating a claim or counterclaim. Thus, experts and decision makers with differing views and backgrounds can better understand each other’s thought processes in complex situations. The net effect is enhanced com- munications and understanding of the whole picture and, in many cases, consensus on decisions to be taken.

A number of human information processing capabilities and limitations interact with organiza- tional arrangements and task requirements to strongly inﬂuence resource allocations for organizational problem solving and decision making. These needs have provided much motivation for the devel- opment of group decision support systems (GDSSs). The purpose of these GDSSs as computerized aids to planning, problem solving, and decision making include:

1. Removing a number of common communication barriers in groups and organizations

2. Providing techniques for the formulation, analysis, and interpretation of decisions

3. Systematically directing the discussion process and associated problem solving and decision making in terms of the patterns, timing, and content of the information that inﬂuences the actions that follow from decisions that have been taken

A number of variations and permutations are possible in the provision of group and organizational decision support. These are associated with speciﬁc realization or architectural format for a GDSS to support a set of GDSS performance objectives for a particular task in a particular environment.

The same maladies that affect individual decision making and problem solving behavior, as well as many others, can result from group and organizational limitations. A considerable body of knowl- edge, generally qualitative, exists relative to organizational structure, effectiveness, and decision mak- ing in organizations. The majority of these studies suggest that a bounded rationality or satisﬁcing perspective, often heavily inﬂuenced by bureaucratic political considerations, will generally be the decision perspective adopted in actual decision making practice in organizations. To cope with this effectively requires the ability to deal concurrently with technological, environmental, and process concerns as they each, separately and collectively, motivate group and organizational problem solving issues.

The inﬂuencers of decision and decision process quality are particularly important in this. We should sound a note of caution regarding some possibly overly simplistic notions relative to this. Welch (1989) identiﬁes a number of potential imperfections in organizational decision making and discusses their relationship to decision process quality. In part, these are based on an application of seven symptoms identiﬁed in Herek et al. (1987) to the Cuban Missile Crisis of 1962. These potential imperfections include:

1. Omissions in surveying alternative courses of action

2. Omissions in surveying objectives

3. Failure to examine major costs and risks of the selected course of action (COA)

4. Poor information search, resulting in imperfect information

5. Selective bias in processing available information

6. Failure to reconsider alternatives initially rejected, potentially by discounting favorable infor- mation and overweighing unfavorable information

7. Failure to work out detailed implementation, monitoring, and contingency plans

The central thrust of this study is that the relationship between the quality of the decision making process and the quality of the outcome is difﬁcult to establish. This strongly suggests the usefulness of the contingency task structural model construct and the need for approaches that evaluate the quality of processes, as well as decisions and outcomes, and that consider the inherent embedding of outcomes and decisions within processes that lead to these.

Organizational ambiguity is a major reason why much of the observed ‘‘bounded rationality’’ behavior is so pervasive. March (1983) and March and Wessinger-Baylon (1986) show that this is very often the case, even in situations when formal rational thought or ‘‘vigilant information proc- essing’’ (Janis and Mann 1977) might be thought to be a preferred decision style. March (1983) indicates that there are at least four kinds of opaqueness or equivocality in organizations: ambiguity of intention, ambiguity of understanding, ambiguity of history, and ambiguity of human participation. These four ambiguities relate to an organization’s structure, function, and purpose, as well as to the perception of these decision making agents in an organization. They inﬂuence the information that is communicated in an organization and generally introduce one or more forms of information im- perfection. The notions of organizational management and organizational information processing are indeed inseparable. In the context of human information processing, it would not be incorrect to deﬁne the central purpose of management as development of a consensual grammar to ameliorate the effects of equivocality or ambiguity. This is the perspective taken by Karl Weick (1979, 1985) in his noteworthy efforts concerning organizations.

Starbuck (1985) notes that much direct action is a form of deliberation. He indicates that action should often be introduced earlier in the process of deliberation than it usually is and that action and thought should be integrated and interspersed with one another. The basis of support for this argument is that probative actions generate information and tangible results that modify potential thoughts. Of course, any approach that involves ‘‘act now, think later’’ behavior should be applied with consid- erable caution.

Much of the discussion to be found in the judgment, choice, and decision literature concentrates on what may be called formal reasoning and decision selection efforts that involve the issue resolution efforts that follow as part of the problem solving efforts of issue formulation, analysis, and interpre- tation that we have discussed here. There are other decision making activities, or decision-associated activities, as well. Very important among these are activities that allow perception, framing, editing and interpretation of the effects of actions upon the internal and external environments of a decision situation. These might be called information selection activities. There will also exist information retention activities that allow admission, rejection, and modiﬁcation of the set of selected information or knowledge such as to result in short-term learning and long-term learning. Short-term learning results from reduction of incongruities, and long-term learning results from acquisition of new in- formation that reﬂects enhanced understanding of an issue. Although the basic GDSS design effort may well be concerned with the short-term effects of various problem solving, decision making, and information presentation formats, the actual knowledge that a person brings to bear on a given problem is a function of the accumulated experience that the person possesses, and thus long-term effects need to be considered, at least as a matter of secondary importance.

It was remarked above that a major purpose of a GDSS is to enhance the value of information and, through this, to enhance group and organizational decision making. Three attributes of infor- mation appear dominant in the discussion thus far relative to value for problem solving purposes and in the literature in general:

1. Task relevance: Information must be relevant to the task at hand. It must allow the decision maker to know what needs to be known in order to make an effective and efﬁcient decision. This is not as trivial a statement as might initially be suspected. Relevance varies considerably across individuals, as a function of the contingency task structure, and in time as well.

2. Representational appropriateness: In addition to the need that information be relevant to the task at hand, the person who needs the information must receive it in a form that is appropriate for use.

3. Equivocality reduction: It is generally accepted that high-quality information may reduce the imperfection or equivocality that might otherwise be present. This equivocality generally takes the form of uncertainty, imprecision, inconsistency, or incompleteness. It is very important to note that it is neither necessary nor desirable to obtain decision information that is unequivocal or totally ‘‘perfect.’’ Information need only be sufﬁciently unequivocal or unambiguous for the task at hand. To make it better may well be a waste of resources!

Each of these top-level attributes may be decomposed into attributes at a lower level. Each is needed as fundamental metrics for valuation of information quality. We have indicated that some of the components of equivocality or imperfection are uncertainty, imprecision, inconsistency, and incom- pleteness. A few of the attributes of representational appropriateness include naturalness, transform- ability to naturalness, and concision. These attributes of information presentation system effectiveness relate strongly to overall value of information concerns and should be measured as a part of the DSS and GDSS evaluation effort even though any one of them may appear to be a secondary theme.

We can characterize information in many ways. Among attributes that we noted earlier and might use are accuracy, precision, completeness, sufﬁciency, understandability, relevancy, reliability, redun- dancy, veriﬁability, consistency, freedom from bias, frequency of use, age, timeliness, and uncertainty. Our concerns with information involve at least ﬁve desiderata (Sage 1987):

1. Information should be presented in very clear and very familiar ways, such as to enable rapid comprehension.

2. Information should be such as to improve the precision of understanding of the task situation.

3. Information that contains an advice or decision recommendation component should contain an explication facility that enables the user to determine how and why results and advice are obtained.

4. Information needs should be based upon identiﬁcation of the information requirements for the particular situation.

5. Information presentations and all other associated management control aspects of the support process should be such that the decision maker, rather than a computerized support system, guides the process of judgment and choice.

It will generally be necessary to evaluate a GDSS to determine the extent to which these information quality relevant characteristics are present.

The Engineering of Group Decision Support Systems

There are two fundamental types of decision making in an organization: individual decisions, made by a single person, and group or organizational decisions, made by a collection of two or more people. It is, of course, possible to disaggregate this still further. An individual decision may, for example, be based on the value system of one or more people and the individual making the decision may or may not have his or her values included. In a multistage decision process, different people may make the various decisions. Some authors differentiate between group and organizational deci- sions (King and Star 1992), but we see no need for this here, even though it may be warranted in some contexts. There can be no doubt at all, however, that a GDSS needs to be carefully matched to an organization that may use it.

Often groups make decisions differently from the way an individual does. Groups need protocols that allow effective inputs by individuals in the group or organization, a method for mediating a discussion of issues and inputs, and algorithms for resolving disagreements and reaching a group consensus. Acquisition and elicitation of inputs and the mediation of issues are usually local to the speciﬁc group, informed of personalities, status, and contingencies of the members of the group. Members of the group are usually desirous of cooperating in reaching a consensus on conﬂicting issues or preferences. The support for individual vs. group decisions is different and hence DSSs and GDSSs may require different designs. Because members of a group have different personalities, motivations, and experiential familiarities with the situation at hand, a GDSS must assist in supporting a wide range of judgment and choice perspectives.

It is important to note that the group of people may be centralized at one spot or decentralized in space and / or time. Also, the decision considered by each individual in a decision making group may or may not be the ultimate decision. The decision being considered may be sequential over time and may involve many component decisions. Alternatively, or in addition, many members in a de- cision making group may be formulating and / or analyzing options and preparing a short list of these for review by a person with greater authority or responsibility over a different portion of the decision making effort.

Thus, the number of possible types of GDSS may be relatively extensive. Johansen (1988) has identiﬁed no less than 17 approaches for computer support in groups in his discussion of groupware.

1. Face-to-face meeting facilitation services: This is little more than ofﬁce automation support in the preparation of reports, overheads, videos, and the like that will be used in a group meeting. The person making the presentation is called a ‘‘facilitator’’ or ‘‘chauffeur.’’

2. Group decision support systems: By this, Johansen essentially infers the GDSS structure shown in Figure 15 with the exception that there is but a single video monitor under the control of a facilitator or chauffeur.

3. Computer-based extensions of telephony for use by work groups: This involves use of either commercial telephone services or private branch exchanges (PBXs). These services exist now, and there are several present examples of conference calling services.

4. Presentation support software: This approach is not unlike that of approach 1, except that computer software is used to enable the presentation to be contained within a computer. Often those who will present it prepare the presentation material, and this may be done in an interactive manner to the group receiving the presentation.

5. Project management software: This is software that is receptive to presentation team input over time and that has capabilities to organize and structure the tasks associated with the group, often in the form of a Gantt chart. This is very specialized software and would be potentially useful for a team interested primarily in obtaining typical project management results in terms of PERT charts and the like.

6. Calendar management for groups: Often individuals in a group need to coordinate times with one another. They indicate times that are available, potentially with weights to indicate sched- ule adjustment ﬂexibility in the event that it is not possible to determine an acceptable meeting time.

7. Group authoring software: This allows members of a group to suggest changes in a document stored in the system without changing the original. A lead person can then make document revisions. It is also possible for the group to view alternative revisions to drafts. The overall objective is to encourage, and improve the quality and efﬁciency of, group writing. It seems very clear that there needs to be overall structuring and format guidance, which, while pos- sibly group determined, must be agreed upon prior to ﬁlling out the structure with report details.

8. Computer-supported face-to-face meetings: Here, individual members of the group work di- rectly with a workstation and monitor, rather than having just a single computer system and monitor. A large screen video may, however, be included. This is the sort of DSS envisioned in Figure 14. Although there are a number of such systems in existence, the Colab system at Xerox Palo Alto Research Center (Steﬁk et al. 1987) was one of the earliest and most sophisticated. A simple sketch of a generic facility for this purpose might appear somewhat as shown in Figure 17. Generally, both public and private information are contained in these systems. The public information is shared, and the private information, or a portion of it, may be converted to public programs. The private screens normally start with a menu screen from which participants can select activities in which they engage, potentially under the direction of a facilitator.

9. Screen sharing software: This software enables one member of a group to selectively share screens with other group members. There are clearly advantages and pitfalls in this. The primary advantage to this approach is that information can be shared with those who have a reason to know speciﬁc information without having to bother others who do not need it. The disadvantage is just this also, and it may lead to a feeling of ganging up by one subgroup on another subgroup.

10. Computer conferencing systems: This is the group version of electronic mail. Basically, what we have is a collection of DSSs with some means of communication among the individuals that comprise the group. This form of communication might be regarded as a product hier- archy in which people communicate.

11. Text ﬁltering software: This allows system users to search normal or semistructured text through the speciﬁcation of search criteria that are used by the ﬁltering software to select relevant portions of text. The original system to accomplish this was Electronic Mail Filter (Malone et al. 1987). A variety of alternative approaches are also being emphasized now.

12. Computer-supported audio or video conferences: This is simply the standard telephone or video conferencing, as augmented by each participant having access to a computer and ap- propriate software.

13. Conversational structuring: This involves identiﬁcation and use of a structure for conversa- tions that is presumably in close relationship to the task, environment, and experiential fa-

miliarity of the group with the issues under consideration (Winograd and Flores 1986). For these group participants, structured conversations should often provide for enhanced efﬁciency and effectiveness or there may be a perception of unwarranted intrusions that may defeat the possible advantages of conversational structuring.

14. Group memory management: This refers to the provision of support between group meetings such that individual members of a group can search a computer memory in personally pre- ferred ways through the use of very ﬂexible indexing structures. The term hypertext (Nielson 1989) is generally given to this ﬂexible information storage and retrieval. One potential dif- ﬁculty with hypertext is the need for a good theory of how to prepare the text and associated index such that it can be indexed and used as we now use a thesaurus. An extension of hypertext to include other than textual material is known as hypermedia.

15. Computer-supported spontaneous interaction: The purpose of these systems is to encourage the sort of impromptu and extemporaneous interaction that often occurs at unscheduled meet- ings between colleagues in informal setting, such as a hallway. The need for this could occur, for example, when it is necessary for two physically separated groups to communicate relative to some detailed design issue that needs to be resolved in order to continue product devel- opment.

16. Comprehensive work team support: This refers to integrated and comprehensive support, such as perhaps might be achieved through use of the comprehensive DSS design philosophy described above.

17. Nonhuman participants in team meetings: This refers to the use of unfacilitated DSS and expert systems that automate some aspects of the process of decision making.

According to Johansen (1988), the order in which these are described above also represents the order of increasing difﬁculty of implementation and successful use. These scenarios of support for decision making are also characterized in terms of support for face-to-face meetings (1, 2, 4, 8); support for electronic meetings (3, 9, 10, 11, 12); and support between meetings (5, 6, 7, 13, 14, 15, 16). There is much interest in groupware as a focused subject within modern information technology developments (Shapiro et al. 1996; Chaffey 1998; Smith 1999). A number of groupware products are available, Lotus Notes arguably being the best known.

A GDSS may and doubtless will inﬂuence the process of group decision making, perhaps strongly. A GDSS has the potential for changing the information-processing characteristics of individuals in the group. This is one reason why organizational structure and authority concerns are important ingredients in GDSS designs. In one study of the use of a GDSS to facilitate group consensus (Watson et al. 1988), it was found that:

1. GDSS use tended to reduce face-to-face interpersonal communication in the decision making group.

2. GDSS use posed an intellectual challenge to the group and made accomplishment of the purpose of their decision making activity more difﬁcult than for groups without the GDSS.

3. The groups using the GDSS became more process oriented and less speciﬁc-issue oriented than the groups not using the GDSS.

Support may be accomplished at any or all of the three levels for group decision support identiﬁed by DeSanctis and Gallupe (1987) in their deﬁnitive study of GDSS. A GDSS provides a mechanism for group interaction. It may impose any of various structured processes on individuals in the group, such as a particular voting scheme. A GDSS may impose any of several management control pro- cesses on the individuals on the group, such as that of imposing or removing the effects of a dominant personality. The design of the GDSS and the way in which it is used are the primary determinants of these.

DeSanctis and Gallupe have developed a taxonomy of GDSS. A Level I GDSS would simply be a medium for enhanced information interchange that might lead ultimately to a decision. Electronic mail, large video screen displays that can be viewed by a group, or a decision room that contains these features could represent a Level I GDSS. A Level I GDSS provides only a mechanism for group interaction. It might contain such facilities as a group scratchpad, support for meeting agenda development, and idea generation and voting software.

A Level II GDSS would provide various decision structuring and other analytic tools that could act to reduce information imperfection. A decision room that contained software that could be used for problem solution would represent a Level II GDSS. Thus, spreadsheets would primarily represent a Level II DSS. A Level II GDSS would also have to have some means of enabling group com- munication. Figure 17 represents a Level II GDSS. It is simply a communications medium that has been augmented with some tools for problem structuring and solution with no prescribed management control of the use of these tools.

A Level III GDSS also includes the notion of management control of the decision process. Thus, a notion of facilitation of the process is present, either through the direct intervention of a human in the process or through some rule-based speciﬁcations of the management control process that is inherent in Level III GDSS. Clearly, there is no sharp transition line between one level and the next, and it may not always be easy to identify at what level a GDSS is operating. The DSS generator, such as discussed in our preceding section, would generally appear to produce a form of Level III DSS. In fact, most of the DSSs that we have been discussing in this book are either Level II or Level III DSSs or GDSSs. The GDSS of Figure 17, for example, becomes a Level III GDSS if is supported by a facilitator.

DeSanctis and Gallupe (1987) identify four recommended approaches:

1. Decision room for small group face-to-face meetings

2. Legislative sessions for large group face-to-face meetings

3. Local area decision networks for small dispersed groups

4. Computer-mediated conferencing for large groups that are dispersed

DeSanctis and Gallupe discuss the design of facilities to enable this, as well as techniques whereby the quality of efforts such as generation of ideas and actions, choosing from among alternative courses of action, and negotiating conﬂicts may be enhanced. On the basis of this, they recommend six areas as very promising for additional study: GDSS design methodologies; patterns of information exchange; mediation of the effects of participation; effects of (the presence or absence of) physical proximity, interpersonal attraction, and group cohesion; effects on power and inﬂuence; and performance / satisfaction trade-offs. Each of these supports the purpose of computerized aids to plan- ning, problem solving, and decision making: removing a number of common communication barriers; providing techniques for structuring decisions; and systematically directing group discussion and associated problem-solving and decision making in terms of the patterns, timing, and content of the information that inﬂuences these actions.

We have mentioned the need for a GDSS model base management system (MBMS). There are a great variety of MBMS tools. Some of the least understood are group tools that aid in the issue formulation effort. Because these may represent an integral part of a GDSS effort, it is of interest to describe GDSS issue formation here as one component of a MBMS.

Other relevant efforts and interest areas involving GDSS include the group processes in computer mediated communications study; the computer support for collaboration and problem solving in meetings study of Steﬁk et al. (1987); the organizational planning study of Applegate et al. (1987), and the knowledge management and intelligent information sharing systems study of Malone et al. (1987). Particularly interesting current issues surround the extent to which cognitive science and engineering studies that involve potential human information processing ﬂaws can be effectively dealt with, in the sense of design of debiasing aids, in GDSS design. In a very major way, the purpose of a GDSS is to support enhanced organizational communications. This is a very important contem-

porary subject, as evidenced by recent efforts concerning virtual organizations (DeSanctis and Monge 1999; Jablin and Putnam 2000).

Distributed Group Decision Support Systems

A single-user DSS must provide single-user-single-model and single-user-multiple-model support, whereas a GDSS model base management system (MBMS) must support multiple-user-single-model and multiple-user-multiple-model support. A centralized GDSS induces three basic components:

1. A group model base management subsystem (GMBMS)

2. A group database management subsystem (GDBMS)

3. A group dialog generation and management subsystem (GDGMS)

These are precisely the components needed in a single-user DSS, except for the incorporation of group concerns. In the GDSS, all data are stored in the group database and models are stored in the group model base. The GMBMS controls all access to the individually owned and group-owned models in the model base.

A distributed GDSS allows each user to have an individual DSS and a GDSS. Each individual DSS consists of models and data for a particular user. The GDSS maintains the GMBMS and GDBMS, as well as controlling access to the GMBMS and GDBMS and coordination of the MBMSs of various individual DSSs (Liang 1988). During actual GDSS system use, the difference between the centralized and distributed GDSSs should be transparent to the users.

We generally use models to help deﬁne, understand, organize, study, and solve problems. These range from simple mental models to complex mathematical simulation models. An important mission of a GDSS is to assist in the use of formal and informal models by providing appropriate model management. An appropriate model base management system for a GDSS can provide the following four advantages (Hwang 1985):

1. Reduction of redundancy, since models can be shared

2. Increased consistency, since more than one decision maker will share the same model

3. Increased ﬂexibility, since models can be upgraded and made available to all members of the group

4. Improved control over the decision process, by controlling the quality of the models adopted

A model base management system provides for at least the following ﬁve basic functions (Blan- ning and King 1993): construction of new models, storage of existing and new models, access to and retrieval of existing models, execution of existing models, and maintenance of existing models. MBMSs should also provide for model integration and selection. Model integration by using the existing model base as building blocks in the construction of new or integrated models is very useful when ad hoc or prototype models are desired. Model integration is needed in the production of operational MBMSs.

In this section, we have provided a very broad overview of group decision support systems that potentially support group and organizational decision making functions. Rather than concentrate on one or two speciﬁc systems, we have painted a picture of the many requirements that must be satisﬁed in order to produce an acceptable architecture and design for these systems. This provides much fertile ground for research in many GDSS-relevant cognitive systems engineering areas (Rasmussen et al. 1995; Andriole and Adelman 1995).

Search This Blog

Operation Research course