Approaches to project management have focused on the systems, procedures, and software put in place to determine progress and likely outcomes. These outcomes are usually expressed in terms of cost, schedule, and technical achievement against the project requirements and framing assumptions—the oft-cited three-legged stool of project management.
These analytics, and the methodologies used to derive them, are effective in recording progress as it occurs. Analytics are then compared against an expected outcome over time based on an estimate that has been converted into a time-phased baseline plan. The variances against that plan are used as the basis for measuring success.
Progress is discussed among the members of the project management team. This team, consisting of individuals of different levels of experience and from different disciplines, proceeds to individually socialize the results and react in assorted possible ways: acceptance, rejection, sadness, joy, anger, curiosity, apathy, etc. Once this emotional phase plays out, healthy acceptance and critical thinking might follow, or additional behaviors like selection bias, denial, avoidance, evasion, diversion, scapegoating, and groupthink might occur instead.
Thus, the behavior of people in the project management team, individually and as a whole, will adapt to the information and feedback given to them by the selected feedback system—in this case cost, schedule, and technical achievement. Do these feedbacks really provide a predictive measure of future outcome given that the information itself will affect the behavior of the system? For taken together, the project management team is a complex adaptive system.
What Is a Complex Adaptive System?
I was first introduced to complexity theory by the Nobel physicist Murray Gell-Mann through his popular work The Quark and the Jaguar. The insight provided by complexity theory has revolutionized and provided insights in many fields, including evolutionary biology, economics, social organization, political science, and business systems.
At its core, a complex adaptive system (CAS) is a learning system, which is what distinguishes it from inanimate complex systems, such as weather patterns, currents, and other physical phenomena. CAS adapts to even small changes that cause modifications to the initial condition, sometimes in very profound ways. The agents in CAS have some “free will” within certain physical and probabilistic constraints or sets of internal rules—as to their reaction to changes in the environment. The total result is these interactions and adaptations of the agents are seen as being self-organizing. It is also seen as emergent, in that the resultant effect is non-linear and often exhibits the pattern of a power law.
Note however that there is no magic by proclaiming something to be a CAS. The use of the concept of CAS provides us with the basis for a model to understand what is happening within the organization, and what its likely trajectory will be.
A Four-Legged Chair Is More Stable than a Three-Legged Stool
As such, understanding that a project management team is a CAS gives us the basis for a fourth leg in our assessment of project performance. It is derived from both sociological and psychological factors (SPF, for short) of the group. SPF stems from two models of behavior.
The first is behavior and insights derived from the Nash Equilibrium. This is often called “game theory” and introduced to popular culture through the example of the “prisoner’s dilemma.” In the model of the prisoner’s dilemma, we have two people who collaborated to commit a crime and are subsequently arrested and held in separate rooms. The police propose to make a deal to both prisoners: The first one to admit to the crime and implicate the other will go free. If they hold out and their partner confesses, then they will be given no deal and, by implication, a long prison sentence.
What the police do not say is that if both stay firm and refuse to implicate the other, both will go free. What should each prisoner do?
The outcome is based on the perception of how the other will act. If Prisoner #1 has a high level of confidence in Prisoner #2’s ability to continue to deny the crime, then Prisoner #1 will also stay firm. What Nash predicts, however, is that in most cases both prisoners will eventually rat each other out because they find themselves in an unstable situation—the dilemma—and so each seeks equilibrium in the situation, which results in each serving a reduced sentence. The pressure of incomplete and unequal access to information—sometimes referred to as “information asymmetry”—forces them to seek their self-interest to their mutual disadvantage.
There will be permutations of this simplistic example, but the “prisoner” for our own purposes is the project management team. As with all cases in real life, the team finds itself in a situation where it must act on imperfect information. Oftentimes individual members within the team or other external entities that interact with the project team create a situation of information asymmetry. As a unit this forces the team to create systems that provide an approximation of the reality so that analytics can be derived that have overcome complete subjectivity. But individual actors within and external to the CAS may act in their own self-interest, which achieves a suboptimal result. In these cases we apply “mix-strategy equilibrium” to inform our model. This involves a set of possible actions based on imperfect information that provides us with a probability distribution of anticipated outcomes.
In these cases we introduce what are called Bayes-Nash equilibrium models, since, even with imperfect information, our choices and the possible outcomes are not unlimited. The universe is bounded, and any outcome is thus bounded by probabilities. At A Fine Theorem blog, Kevin Bryan of the University of Toronto summarizes the work of Reinhard Selten in determining that one’s perception of other actors and their motivations—regardless of one’s own (ir)rational assessment of a situation—determines the probability of any set of outcomes.
Having worked as a contract negotiator myself, this understanding is more intuitive than mathematical. A negotiator will try to find the overlapping areas in this relationship to strike a deal, with the purpose of satisfying all parties given an understanding of the acceptable positive outcomes defined by the environment. The understanding being sought not only applies to the parties to any transaction, but also within the context of the decision environment, which may consist of other interests and parties affected by the transaction that determine whether it is realistic. This understanding might take the conceptual form below:
This two-dimensional representation is further complicated by the design space within which the decision environment operates. I would posit “design space” as the conditions of possibility that underlie any effort that is undertaken. In this way I am borrowing from and modifying Daniel Dennett from his book Darwin’s Dangerous Idea, regarding biological evolution. The conditions for possibility follow as quoted from the book:
- logical possibility – “a matter of being describable without contradiction.” (p. 104)
- physical possibility – the effort or outcome is consistent with the laws of physics.
- biological possibility – a subset of physical possibility; what biology can support in living things. In Dennett’s sense this means what can occur from the perspective of the evolution or adaptation of the organism. For our purposes, I would modify this subset to be what is possible based on human physical, psychological, neurological, and/or ecological constraints.
- historical possibility – what might have been possible or impossible in the past but may no longer be possible.
These a priori conditions create our design space upon which the decision environment exists. What this means is that the cards may be stacked against the parties to any transaction, regardless of their efforts. There is also a probability grid across the design space that will predetermine the possible level of success in executing the work associated with the transaction.
The above conceptualization of design space would contain peaks and valleys that would bound the probability distributions possible in any portion of the vector.
Squaring the Circle
The Nash Equilibrium, even when informed by Bayesian methods, approaches what is called “intractability” as the number of variables increases. Something is intractable when a problem can be solved in theory, but in practice the computational time to solve it takes too much time to be useful. That is, they cannot be solved in polynomial time. Given Moore’s Law in which computing power not only doubles but the cost associated with each additional bit has the effect of nearing zero as posited by Landauer’s Principle, I believe that this is only a temporary problem, but it is still the current state.
The core of our issue is that in any set of problems there are categories of them that can be guessed at with a high degree of confidence, but then when a computation must be made to calculate that problem it cannot be solved within polynomial time. Thus, in shorthand, for the class of NP (non-polynomial problems), there are those that are solved in polynomial time (P) and those that can be guessed (NP-complete), but that P never equals NP.
The problems regarding applying the Nash Equilibrium practically in real time were effectively documented by Daskalakis, Goldberg, and Papadimitriou in this MIT paper. Still, Daskalakis in particular has managed to solve the Nash Equation’s intractability in specific cases, by breaking down the complexity of a system into constituent components.
For example, in 2012 he and his graduate students built on the Nobel work of the University of Chicago’s Roger Myerson on the economics of auctions. Daskalakis and his students were able to solve the complexity of auctions where bidders are competing for a multiple number of items. The solution was to view the auction as made up of a probabilistic combination of simple auctions, reducing the problem to what he termed a “geometric problem.”
This approach suggests that seeking optimum organizational success in balancing the many variables of complex project management can be achieved by identifying and properly integrating the subsystems that make up the whole. But doing so requires that we also properly identify the foundations upon which to assess the relative impact of these contributors.
Keep in mind, however, that this challenge still does not have a solution.
Bringing It All Back Home
Getting back to SPF, the second necessary addition to the insights provided by the Nash Equilibrium are those provided by contract theory. This work garnered both Oliver Hart and Bengt Holmstrӧm the 2016 Nobel in economic sciences.
To the project manager, the basis of the project is the contract that defines the work and conditions of the work. Furthermore, a project manager will usually enter into subcontracts and have to manage a number of employees, each of whom either individually or collectively operates under contract.
In neo-classical economics and business, many of us have been taught that contracts are built on the conditions of supply and demand, and that the needs sorted out by the market create ideal structures and conditions for business. This is known as equilibrium theory. It was a highly popular and influential theory, and it has influenced both business thinking and public policy.
It also has had the effect of not being true in the real world. It’s simply not how things works, and the damage from those who use it as a basis for practical decisions is felt at all levels of society. The research that has led to contract theory pushes equilibrium theory to the side.
One of the best non-technical definitions is found at Business Insider, where the writer explains that contract theory “studies the design of formal and informal agreements that motivate people with conflicting interests to take mutually beneficial actions.” It looks to human behavior and the fact that, unlike equilibrium theory, it does not rely on ideal circumstances where information is symmetrical and contracts (written or perceived) are sufficiently detailed.
So far, three aspects underlie the observed behavior that composes contract theory. The first two were largely developed by Holstrom and the last by Hart. These are the following:
- Tests of moral hazard and how information asymmetry affects the behavior of people where full insurance of an outcome is not possible. For example, party A contracts with party B for development of an end-item application C. Party B is required to report on progress using particular metrics of performance to party C. Given that party B will have information that party A cannot observe or verify, information asymmetry will create a contract where full insurance of the result is not guaranteed to party A. Moral hazard in the effort falls on party A as a result of this condition, which also provides incentives for party B to withhold adverse information.
- Tests regarding adverse selection in which parties engage in a contract or agreement where asymmetry in information exists. In this principal-agent relationship, the agent will have information about its condition that is not easily known by the principal. This is known as the agent’s type. A good example in this case is when someone issuing a contract must assess the responsiveness, reliability, and capability of a possible supplier all within the constraints of cost. The agent’s type to be assessed that will not be known to the principal is more than likely cost, since during the contracting process the responsiveness, reliability, and capability of the supplier can usually be determined as part of the procurement process. But cost, even where the supplier’s accounting system and cost contributors and rates are exposed, often cannot be fully determined. Once again, information asymmetry and competing incentives can result in a condition where not all costs have been identified.
- Tests regarding incomplete contracts where, quoting again from the cited Business Insider article, the “idea is that it is impossible to write a contract that anticipates every potentially relevant future contingency. Consequently, the allocation of control rights becomes a powerful tool for creating incentives. This perspective enables the analysis of fundamental questions such as whether companies should outsource or integrate production, which assets they should own and how they should choose between equity and debt financing.” This conflict in incentives also extends to issues of public-private ventures, in which public goods are outsourced to private firms where the incentives for, say, cost reduction are outweighed by reductions in quality. Most recently, this research was influential in the decision by the Department of Justice to eliminate private prisons.
What these measures of behavior tell us is that, for all of our reliance on predictive analytics and key performance indicators, the human factor is the element that has the ability to defeat all of the management control systems in place. This is true particularly if the relative control rights and incentives built into the contractual relationship are in conflict with the goals of the organization.
Back when I was a young U.S. Navy officer, a wise old Navy Captain advised me that “people behave the way they are managed.” What he meant is that there are a plethora of unwritten and unspoken incentives that exist in the workplace, combined and altered by both the motivations and perceptions of the people that make up the organization. It is the wise leader to ferret these hidden factors, and bend them to encourage team building and normative behavior.
Thus, given empirical observed behaviors that increase our understanding of these system dynamics, it is essential that practical models be established and applied based on this knowledge. Applying these models will determine the types of incentive packages, team structures, technical approaches, and contract types that are most appropriate in establishing an environment that at least pushes the bias toward success.
As organizations rush to implement new project management models like agile as a means of addressing organizational improvement, we must keep in mind that these new prescriptions are largely based on untested assumptions of human behavior. These methods, pushed by cultural or industry gurus, based on manifestos that are characteristic of revealed religion or astrology, may possess a grain of insight or truth. But they most often represent efforts that artificially prescribe a panacea outside of human agency, when the real answer is simply to understand ourselves.
For more brilliant insights, check out Nick’s blog: Life, Project Management, and Everything