Guiding Quote

“Learn from yesterday, live for today, hope for tomorrow. The important thing is not to stop questioning.” Einstein

Saturday, April 18, 2015

Models and Biases

One of the abiding myths in economics is that there is such a being as homo economicus, or a perfectly rational person when it comes to making decisions. Now that myth has been undermined by the work of behavioral economists, who have described a whole series of biases and proclivities that control how we think and make decisions.

Their work has been used by the "big number" crunchers to justify a process called business analytics; they claim that by using their models we can avoid these biases and thereby reach a rational decision. Models don't have emotions and they dont have biases, they're inorganic so they can't be swayed away from the facts. Sounds logical, right.  However, who creates the models? 

We fallible Homo sapiens create the models. us with all our pesky biases and non-rational behaviors.  People who build models do so with an end in mind. The model is built to reproduce a process. It is created from a set of requirements. It is formed from someone's belief in how a process works. That belief requires the creation of assumptions that underlie the models design. We know that assumptions are very susceptible to bias.

For example the financial models that were used in the mortgage industry pre 2008 did not allow for a decrease in property values, the widely held assumption was that property values always increase. Also the efficient market theory, much touted by economists, and its models assumed, no, asserted that the price of houses and shares was always right because the market "knew" all the facts about an asset and would price it correctly. The models did not, and still do not, allow for a price bubble. Cant happen, so in theory the worldwide real estate crash and the resultant financial crisis shouldnt have occurred, but it did. Oops. Guess the models are as flawed as the people who build them!

We project managers should always be careful when we are told that the model or the tool gives us a perfect answer. Unless we know the assumptions that went into building the application we should never take that statement on its face value. Do your own quick analysis to see if the answer is realistic, trust but verify. All PMs should construct a mental bullshit detector. Theres never a shortage of BS, or people to spread it around!

No comments:

Post a Comment