Decision Aiding
A process-based model for priority convergence in multi-period group decision-making

https://doi.org/10.1016/S0377-2217(02)00442-3Get rights and content

Abstract

Over the past three decades substantial research has been dedicated to the task of providing theoretical models for group decision processes. Unfortunately, many models remain grounded in single-period decision contexts. The current work investigates an extremely basic process approach to multi-criteria multi-period group decision-making. The time-series model considered incorporates factors relating to member influence over aggregated group decisions, the effect of past priorities on subsequent priorities and the cognizance/acceptance of performance optimizing task-criteria priorities over otherwise non-ideal ones. Quantitative values relating to criteria priorities are elicited for each period through paired comparisons. Empirical analysis of collected group task data is done through a three-stage evaluation of the model’s performance. Results suggest that the time-series approach is consistent in evaluating changes in criteria priorities over-time. The model can provide a strong auxiliary tool for multi-period extensions of existing approaches to decision support analysis.

Introduction

Extensive work has been devoted to modeling multi-criteria group decision (MCGD) problems. Work in recent years has been no exception. In general these problems have been associated with a number of features including the existence of conflict between alternatives, multiple competing performance criteria and a group decision environment (Hwang and Lin, 1986). Two of the primary points stressed in recent MCGD research have been the introduction of alternate techniques for group preference aggregation (Bolloju, 2001; Forman and Peniwati, 1998) and the consideration of the implications of incomplete information in the decision process (Kim and Ahn, 1997). A third point stressed in the literature has been the potential benefit of iterative techniques and applied decision support mechanisms for these processes, such that win–win solutions or other objectives may be structured for those involved (Teich et al., 1996; Teich et al., 1995). Yet the empirical evaluation of such issues and the appropriateness of alternate MCGD techniques remains relatively limited (Easly et al., 2000; Iz and Krajewski, 1992), particularly with regards to multi-period human decision processes. Ehtamo and Hamalainen (2001) provide a review of the literature that further emphasizes this gap.

Several traditional sources of influence can be considered within multi-period evaluations. One of the most intuitive is the relative power of group members on the iterative decision process. The use of decision weights as representations of member influence goes back to the early works of such authors as Harsanyi (1955), and Dyer and Sarin (1979), as well as the works of other more contemporary authors (Islei and Lockett, 1991). The classical tenets behind the application of such schemes stem from assumptions relating to complex social forces within the group including individual member indifference towards alternatives or alternate sets of task criteria priorities. Another key element is the presence of convincing group leaders, either by virtue of charisma or overt dominance, which are in turn partial to specific alternatives (Von Neumann and Morgenstern, 1944). Furthermore, the same issues represent secondary determinants, or structural “faults” of groups, as described in Janis’s groupthink framework (Janis, 1982).

The persistence of incomplete information is similarly cited by Janis (e.g., insulation from outside experts) as an antecedent of the group think phenomena, with the issue of group cohesion serving as the binding determinant. Of course, the bane of groupthink lays in the formation and maintenance of a premature assessment of a preferred set of priorities inferior to an otherwise ideal set (Janis and Mann, 1977; Aldag and Fuller, 1993). However, group cohesion in general should not necessarily be interpreted entirely as a phenomenon to be avoided, particularly when the group facilitates information pooling (Festinger, 1950; Stewart and Stasser, 1995; McCauley, 1998). Indeed, in the presence of incomplete information, cohesion may prove critical to efficient sequential improvements within decision processes, particularly in the absence of an overpowering leadership that misguides rather than improves the set of priorities effective for the group as a whole (Huber, 1989; Fuller and Aldag, 1998).

The efficient and repeated elicitation of group member views can also directly impact learning and convergence in group decision processes, approaching either an ideal or non-ideal group consensus (Salo, 1995). Its particular importance to group decision processes, in contrast to one-time group decisions, relates to the types of analysis needed to assess the nature of underlying effects and relationships. While decision-making can be studied from a structural-modeling standpoint, such efforts in their pure form are typically restricted to attempts at relating final alternative selections to single-stage explanatory inputs. Process-based approaches (e.g. time-series analysis, event analysis or sequence-focused content analysis) on the other hand attempt to trace processes through their duration, perhaps foregoing the breadth of independent factors under consideration in place of the depth of processual understanding gained with regards to a subset of such factors. In his recent work, Svenson (1996) indeed claims that process perspectives on human decision-making are essential in the explanation of human decision-making regularities. According to this author, existent theory seems to suggest the prevalence of a number of relevant levels of decision-making that may be observable throughout such process-based approaches.

At a base level are the effects of sub-conscious rules, analogous to Klein’s “recognition-primed” specifications, which emerge predominantly from prior experience (Klein, 1989; Shiffrin and Schneider, 1977; Svenson, 1992). Analytical analogies naturally arise with regard to autoregressive models, in which future indicator values of preferences are based largely on those prior to them. Another level cited by Svenson involves the specification of a preference within a range of options at any given point within the process, though again research to date has been predominantly restricted to the end-decision evaluation (Svenson, 1996). Nevertheless, such a level in group processes suggests the generation of some form of group consensus as well as the potential convergence of member preferences towards a more narrowly defined and accepted range of priorities. Lastly, the search for a promising set of priorities based on step-wise indications of idea performance-maximizing priority schemes, though almost entirely devoid of investigation in the academic realm, has been suggested as a particularly relevant issue with regards to the group process (Fischhoff, 1996).

One of the salient points with regard to the current work is the distinction between information-based and norm-based effects on both individual member priorities and group priorities regarding task criteria throughout the decision process. Keeney alludes to these two effects by considering the effects of alternative-focused as opposed to value-focused thinking (Keeney, 1988, Keeney, 1992). Essentially, the first can be interpreted as a disposition towards the consideration of any alternate set of priorities in general, as opposed to a strict dependence on recognition-primed alternatives, for example. Value-focused thinking however alludes to the ability of group members to favor the influences of performance-based indications of a more ideal set of priorities vis-à-vis the prior group consensus, as opposed to the group consensus alone.

While alternative-thinking is critical as a means to consider options that may or may not be ideal, value-thinking represents the channeling of appropriate resources in order to “make better decisions” (Keeney, 1994). It therefore encourages a progression towards ideal priorities during the process, rather than a convergence upon relatively non-ideal beliefs of mislead groupthink. Similar distinctions between informational and normative influences have also been made within the MCGD contexts (Bryson, 1996; Huang et al., 1993; Tan et al., 1993). One of the aims of the current work is to help reconcile the void present in the process literature, through the empirical testing of a process model based on these same tenets.

Section snippets

Model foundation and structure

The present investigative model has its foundations in a relatively simple and mechanical approach to individual decision processes, with extrapolations to group settings. Rather than deny the existence of a multitude of behavioral components active in this process, these foundations assume a priori that complex mechanisms are at work. The concentration here however is not on the dependency upon such antecedents themselves, an issue delegated to future research, but rather the interaction of

Consideration of model characteristics

Beyond its grounding in existent theory, the structure of the group priority convergence model in itself provides certain appealing characteristics. First, in its general form, it is simplistic as far as behavioral time-series models are concerned. This of course does not suggest that it is all encompassing at the same time. However the basic development of behavioral models necessarily begins with an understanding of simple underlying issues, prior to the imposition of more complex

Experimental design and validation

MCGD environments suitable for the examination of priority convergence phenomena are those in which {1} group decisions involve tradeoffs between alternatives, each characterized by a comparable yet distinct set of attributes, {2} overall task performance measures, and changes thereof, reflect group decisions and that these measures are unambiguously presented in immediate response to group decisions, and {3} individual priorities, can be assessed both prior and subsequent to the task (as well

Analysis and results

Each group’s convergence rate metrics were derived in accordance with their definitions in , , as was the back-calculation of average values for βCA, as described by Eq. (3). Ordinary least square fits were applied to the time-series data to estimate the decision weights (βDW) under consideration. In these fits, the individual priorities of group members served as independent variables whereas the group’s common (consensus-based) set of priorities served as the dependent variable. As an

Conclusion and future work

Empirical investigations into group priority aggregation and information usage in problem solving environments remains limited, particularly with regards to multi-period processes. To fill this gap, the current work has attempted to build upon past theory and research in this area to propose a simple and tractable approach to modeling multi-period group priority convergence. To add validity to the idea that simple time-series based models may be applicable in these complex settings, several

References (35)

  • J.E. Teich et al.

    Identifying Pareto-optimal settlements for two-party resource allocation negotiations

    European Journal of Operational Research

    (1996)
  • R.J. Aldag et al.

    Beyond fiasco: A reappraisal of the groupthink phenomena and a new model of group decision processes

    Psychological Bulletin

    (1993)
  • D.G. Bachrach et al.

    Attributions of the “causes” of group performance as an alternative explanation of the organizational citizenship behavior/organizational performance relationship

    Journal of Applied Psychology

    (2001)
  • N. Bryson

    Group decision-making and the analytic hierarchy process: Exploring the consensus-relevant information content

    Computers and Operations Research

    (1996)
  • J.S. Dyer et al.

    Group preference aggregation rules based on strength of preference

    Management Science

    (1979)
  • R.F. Easly et al.

    Capturing group preferences in a multi-criteria decision

    European Journal of Operational Research

    (2000)
  • H. Ehtamo et al.

    Interactive multi-criteria methods for reaching Pareto optimal agreements in negotiations

    Group Decision and Negotiation Journal

    (2001)
  • Cited by (14)

    View all citing articles on Scopus
    View full text