Decision-making in uncertainty
Decision-making is about risk management, and making good, informed decisions. These good, informed decisions should be based upon facts and probability. More specifically, decisions should be based upon the probability that an event will occur, the cost of the occurrence, and the impact of the occurrence. In this article we will deal with the issue of probability.
Decisions made in uncertainty are decisions made based upon conjecture, prejudice, or perception. Lacking quality information, we imbue our decision charges with anthropomorphic qualities, much like the gambler who tosses a coin. After the coin has landed heads 10 times in a row, the gambler imbues it with the anthropomorphic quality of memory, and thus “next time it must come up tails.” But the reality is that the coin has no memory and so there is still only a 50/50 chance for either outcome.
When people are presented with the following scenario, we learn something about risk versus the perception of risk.
You are going into battle. Your soldiers are your most valuable asset. You must protect them, and to do so you will have to make a decision. If you move forward and capture a hill, you will loose fully one fifth of your men. If you retreat or do nothing, you will only stand a 20% chance that you will lose any soldier.
Of those queried:
• 58% retreat or do nothing
• 35% advance
• 7% correctly realize it is a break-even for losses.
What does the above tell us? That only 7% of those presented with this problem realized that no matter which of these two decisions they take, the number of losses will be approximately the same, which may be an indication that some novel third approach — the difference between a good general and a great general — needs to be developed!
Another example is based on a study on policing in a southwestern community. (This is not a reflection on law enforcement, but rather a statement about our human perceptions, and it could apply equally to engineers.)
The police department covered a sizable city with many very recognizable communities, each with its own ethnic makeup and particular sets of policing challenges. Blind surveys were taken of the police officers who operated in the different communities. The survey questions dealt with the personal attitudes and prejudices of the police officers.
The surveyors somehow expected to see a more or less equal distribution of stereotyping and prejudicial views across the department. What they found was that police officers were most poorly disposed toward the given ethnic population of that portion of the community they were policing. Hence, the officers who possessed generally negative feelings toward Slavs were policing Slavic neighborhoods, while those who had generally negative feelings toward Hispanics were policing Hispanic neighborhoods.
Could the police department knowingly, or unconsciously, have placed the officers in those neighborhoods with an ethnic makeup they couldn’t stand? No. What had happened was that when community-oriented policing was implemented, officers were no longer rotated out of neighborhoods. Hence, those officers patrolling a Slavic neighborhood were arresting mostly Slavs and those patrolling a mostly Hispanic neighborhood were arresting mostly Hispanics. Based on the statistical nature of their regular negative contact with these populations, the police officers developed substantial, continually reinforced, and not-invalid-within-their-neighborhood negative attitudes toward these populations.
Note that there is a big difference between saying “within this Slavic neighborhood most criminals are Slavs” and saying “most Slavs in this Slavic neighborhood are criminals” or “most criminals in non-Slavic neighborhoods are Slavic.”
To test this theory, some officers were rotated out of their assigned neighborhoods to a new neighborhood with a different ethnic makeup. These officers were found to be more likely to give the benefit of the doubt to a person in their newly assigned ethnic neighborhood than they were to a person in their previously assigned ethnic neighborhood. Of course, as their statistical sampling changed with time, and with their increased experience with the new population, their views changed to reflect their new reality (though it is not clear whether the new group was added to the old population or replaced it).
This is a powerful statement about human decision-making based on inaccurate or obsolete data, and on carrying data valid in one environment into a new environment. It is also a powerful tool for professionals to use in understanding our perception of risks and how we deal with those risks, and the need to be able to quantify risk in some valid manner.