Risk Decision-Making: Where Models and Cognitive Bias Collide

Why do seemingly effective management teams at times make horrible risk decisions under uncertainty? Analytical models can only take us so far. A combination of decision-maker behaviors coupled with underdeveloped decision-making processes lies at the heart of the problem. However, there are ways to significantly improve the quality of risk decisions. 

Although we call it risk management, oftentimes most of the decisions we make as risk managers are under uncertainty. Over the years, the quantitative tools and data to measure risk generally have improved. However, many of the core risks of organizations are not easily quantifiable, particularly when the present and future no longer look like the past.

Our instincts are to gravitate toward developing and refining analytical tools that measure and predict outcomes that pose risk, however, doing so without taking great care over the applicability of those tools can lead to poor decisions. At the same time, our decision support structures have not kept pace with developments on the analytics side of risk management and can further diminish our ability to detect emerging risks or identify potential opportunities due to cognitive biases and other behaviors of key decision makers. Improving the way risk managers make decisions under uncertainty is of paramount importance to the organization.

Coming to Grips with Uncertainty

Peter L. Bernstein’s bestseller Against the Gods: The Remarkable Story of Risk highlights the struggles over the millennia that mankind has faced in a quest to measure events that appeared unpredictable. Whether it was some natural disaster destroying a harvest, a flood event wiping out communities, or the loss of a shipment of goods along a sea route, risk and uncertainty have been part of the human condition.

Over time, as ways to measure some of these events developed, we could start to determine the likelihood of an event and its severity. Once we had those two pieces of information, it became possible to reasonably measure potential loss. That is the very notion of risk management and yet, in many instances, our ability to measure loss from an event is limited or even nonexistent. This is where we are forced to confront the problem of making decisions under uncertainty.

According to Marchau et al, outcomes follow a continuum between full knowledge with precise measurement to the complete unknown. In between we find a progression of information and data levels to develop point estimates (Level 1 uncertainty) and entire distributions (Level 2 uncertainty) of outcomes. Levels 3 and 4 (deep uncertainty) are where things for risk managers get tricky. At these levels, the ability to measure and predict events comes effectively to a screeching halt. Lacking reliable probabilities of potential outcomes limits analysis to the use of a set of scenarios. But anything beyond a certain outcome, i.e., fully deterministic decision-making, becomes more difficult as cognitive bias and underdeveloped decision-making processes greatly impact the quality of the decision.

Other factors such as time horizon can further undermine a decision process. Climate risk is a good example of deep uncertainty. Reliable predictions of the future are difficult to come by, particularly for predictions that extend out to the year 2100. Making hard money decisions on such uncertain outcomes is suboptimal and exhibits what I call shiny object bias exhibited by decision-makers overly enamored with seemingly elegant and complex analytical models that have not been properly pressure-tested.

The possibility of a range of outcomes, even if they can be reliably measured, provides an environment for decision-maker cognitive bias to flourish and potentially hijack otherwise good decisions. We’ve seen this story before prior to the 2008 Global Financial Crisis where bank management and boards were overly optimistic regarding mounting credit risk in their mortgage portfolios, and in March 2023 with banks once again misunderstanding the impact of interest rate risk on their balance sheets. Cognitive bias comes in many forms, any of which can fatally impact a decision. These biases can include choosing to emphasize more recent performance over other periods, following the herd, or making decisions anchored on information that is easy to come by but in fact is irrelevant.

Decision outcomes can be significantly influenced by cognitive bias especially when those biases are exhibited by influential decision makers where influence is reflected by some combination of authority or knowledge of the problem. In such cases, a countervailing force coming from another authoritative source can provide balance and perspective that promotes consideration of all sides of an issue. In some organizations, this is codified by an independent risk management function engaging management in effective challenges.

Decisions are likely to be suboptimal for processes lacking such decision support structures.

How to Improve Decision-Making Under Uncertainty

While the technical apparatus to make informed decisions about various risks has improved over time, less time has been spent in organizations about how such decisions should be made. Understanding elements of group dynamics and decision-maker behaviors is key to having a chance of making good decisions when facing uncertainty.

There are a number of proactive steps risk managers can take to strengthen their decision-making processes. One approach is to consider adopting one of several frameworks for decision-making under uncertainty. These could include Robust Decision Making, Decision Making Under Deep Uncertainty, among others. Such frameworks establish a clear structure and criteria for how decisions should be made, establishing the expectations of decision-makers and the process for how information and analysis should be presented based on the nature of uncertainty presented in the problem.

It is essential for organizations to establish a decision-making structure that promotes competing views and healthy, robust debate based on clear-eyed judgment leveraging relevant information, data and analysis. A decision-making structure, whether by a board or management risk committee or other mechanism must assure balance in views at the table and avoid situations where a dominant individual can steer toward a decision devoid of any effective challenge. Having experts able to clearly articulate alternative views puts the organization in the best position to make a sound decision. Likewise, leveraging data and models in the decision-making process is imperative but must be viewed with a healthy dose of skepticism. Use models as guideposts rather than as decision-making crutches. Where uncertainty prevails, an incremental test and expanded approach avoids exposing the organization to excessive risk while providing the flexibility to adapt to changing conditions.

Parting Thoughts

Despite the term “risk” in our titles, risk managers operate daily in a world of uncertainty. Uncertainty is difficult to navigate and our tendency to lean heavily on analytical models must be tempered with the realization that such tools are fallible, acutely so when we lack solid data to inform them. Unfortunately, risk decision-making processes remain underdeveloped to properly address potential problems with decision-maker cognitive bias and authoritative influence. Understanding the limitations of decision-makers and models is critical to building strong decision-making support structures that can ward off making bad decisions under uncertainty.

Clifford Rossi, PhD, is the academic director of the Smith Enterprise Risk Consortium at the University of Maryland and a professor of the practice and executive in residence at UMD’s Robert H. Smith School of Business. Before joining academia, he spent more than 25 years in the financial sector as both a C-level risk executive at several top financial institutions and a federal banking regulator. He is the former managing director and CRO of Citigroup’s Consumer Lending Group.