Structural Change and Homeostasis in Organizations: A Decision-Theoretic Approach

Abstract

We present here a decision-theoretic framework for the analysis of organizational change and homeostasis under risk. An algorithm is demonstrated that identifies optimal change paths given uncertainty involving execution time, intervention cost, and payoffs resulting from particular structural configurations. An elaboration of the basic framework to accommodate external structural perturbations is shown, and is applied to the problem of organizational homeostasis. Finally, an extension of the decision model is provided that admits multiple decision makers with divergent preferences and capacities for inducing organizational change.

Keywords:

• decision theory
• dynamic network analysis
• homeostasis
• network evolution
• organizational adaptation
• organizational design
• structural change

This work was supported in part by the Office of Naval Research (ONR), United States Navy Grant No. 1681-1-1001944, by the Army Research Labs on the grant “Personnel Turnover and Team Performance,” by the National Science Foundation under Grants No. ITR/IM IIS-0081219, ITR IIS-0331707, and IGERT Grant No. 9972762, and by the center for Computational Analysis of Social and Organizational Systems (CASOS) at Carnegie Mellon University. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Office of Naval Research, the Army Research Labs, the National Science Foundation, or the U.S. government.

The authors would like to thank David Rode for his helpful comments regarding this article.

Notes

¹In fact, even this assumption is not required for any of the formal results shown here; these depend only on the properties of the constructs , , and π (see below). Our discussion of potential structural payoff functions and our examples, however, do assume this.

²Vertices intrinsically uninvolved in a given relation (e.g., knowledge elements in a task assignment network) can simply be regarded as isolates for that relation.

³Note that we assume here that implementation time is a function of the structural transformation, and not of the structure on which this transformation operates.

⁴The astute reader will note that these representations are not precisely equivalent, in the sense that there may be more than one transformation connecting each pair of adjacent structures. That this discrepancy is of no consequence for the decision problem is shown below.

⁵E.g., it may become optimal to travel in endless loops.

⁶The performance of Dijkstra's algorithm is close to optimal for dense graphs (Ahuja et al., 1993), and thus it serves as an obvious starting point for the present application.

⁷For instance, applying a transformation which removes personnel may increase the probability that Chance will apply a transformation resulting in further personnel loss, thus providing a model for the “snowballing” of turnover processes (Krackhardt and Porter, 1986).

⁸The exception being when there are sufficient numbers of optimal forms that the organization may constantly shift between them; this seems unlikely in practice, however.

⁹Other than the identity transformation; since including this within  _C would be equivalent to a reduction in λ, we can safely ignore this possibility.

¹⁰The Lambert-W function is defined as the principal solution to the inverse of f(W) = W e ^W .

¹¹We are indebted to David Rode for this suggestion.

¹²We omit the case in which commitments are enforceable and all Agents have the same capabilities, since this is merely the standard optimal change problem.

¹³We further assume that the Principal's preferences over all such potential transformations are well-formed.