Projected Biases

Hope and Lies

“Behavioral economics suffer from a "psychology bias... Cognitive bias is therefore only half the story in behavioral science. Political bias is the other half.” Bent Flyvbjerg

The improbably-named Bent Flyvbjerg has spent many years looking at why projects go off the rails. His research includes some disturbing work on how projects get approved - “It is not the best projects that get implemented in this manner, but the projects that look best on paper” and how much they go over budget - “Our research shows that IT projects have far greater cost risk than is commonly assumed”.

His work is truly the stuff of nightmares for project managers. And, lets face it, we are a pretty miserable bunch anyway - with our risk registers and contingencies and mitigations. We are professional pessimists. Although Flyvbjerg would probably say that we are not pessimistic enough.

However his aim is truly noble: How do we as a species get better at running projects? Because badly run projects make our lives worse over all. One of my favorite papers by him concerns behavioural biases in project management. A Top Ten in fact. The word “behavioural” has been chosen carefully. As the quote at the start indicates, Flyvbjerg sees biases emerging from the politics and money that can be garnered by projects as well as the psychological biases that we are all prone to, In fact: “total bias -- political plus cognitive -- escalates, but not in a simple linear manner where total bias equals the sum of political and cognitive biases, but instead in a complex, convex way where political bias amplifies cognitive bias, leading to convex risk”. If you don’t understand what that means then let me reassure that it is bad. Perhaps a crude way is to say that political greed and personal blindspots multiply each other.

One of Flyvbjerg’s solutioms is reference class forecasting:

1. Identify a reference class of past, similar projects.

2. Establish a probability distribution for the selected reference class for the parameter that is being forecast.

3. Compare the specific project with the reference class distribution, in order to establish the most likely outcome for the specific project.

So if you are implementing an ERP system or building a metro line, what are the probability distributions of cost and duration?

Now obviously it is possible to do this badly and to negate any benefit such a method might have. And you need reference data - but the good news here is that you probably have more reference data than you think. You are not the special snowflake that you think you are.

As with alcoholism, the first step is to admit that we were powerless over forecasting and that our lives had become unmanageable. A searching and fearless moral inventory of ourselves is probably not required.

Just a willingness to learn from others.