In this workshop, we will focus on understanding and managing the concept of "Ruin" in decision-making. "Ruin" refers to the potential of extremely negative outcomes that can have lasting, severe consequences. Participants will learn how to identify and incorporate ruinous outcomes into their decision-making process using decision trees.
Introduction [20 minutes]
In the context of practical decision theory, "ruin" refers to an outcome or event that has extremely severe negative consequences, often resulting in irreversible damage, significant loss of life or resources, or long-term detrimental effects on the system, organization, or individuals involved.
Review this list of the most common types of ruin for an individual, and discuss the questions when you're done.
- Financial Ruin: Choices involving a risk of income loss, or unexpected massive expenses.
- Career Ruin: Distinct from financial ruin, some choices can result in irreparable damage to one's reputation, and an inability to find future employment in their field.
- Health Ruin: Choices that might lead to injury, severe health issues, disability, dramatically shortened life expectancy, or imminent death.
- Relationship Ruin: Choices that come with a risk of permanent damage to trust and emotional well-being in a romantic relationship, family relationship, or friendship.
- Legal Ruin: Choices involving the risk of legal consequences, including loss of personal or professional rights, or imprisonment.
- Emotional Ruin: Choices that result in severe emotional distress, such as experiencing a traumatic event or being subjected to long-term stress.
- Social Ruin: Choices that carry the risk of severe damage to one's social standing, leading to isolation and loss of support networks.
- Ethical Ruin: Choices that involve compromising one's moral or ethical principles, resulting in a loss of self-respect and the respect of others, as well as potential guilt and regret.
- Collateral Ruin: Choices that pose significant risks to people other than yourself, who would suffer greatly if things turn out poorly.
Discussion [20 minutes]
- Are there any other types that you would include on the list?
- How commonly would you say you encounter choices that involve a risk of ruin, and which type is most common?
- How small, in your opinion, is "small enough" of a probability to simply dismiss the risk, and how might that depend on the nature of the risk? (Example: the risk of driving. Driving is one of the most dangerous high-frequency activities we engage in. Nonetheless, most people dismiss this risk and consider the benefit of having a car worth the danger.)
Identifying Ruin at Station Seven [15 minutes]
In this activity, cohorts will discuss different scenarios at Station Seven where ruinous outcomes might be near at hand. For each scenario, identify the potential ruinous outcome(s) and discuss the likelihood and impact of the ruinous outcome(s).
Scenario 1. Redford knows that he can save half an hour on his commute if he does a quick cold-shirt crossing between cylinder spokes, and considers doing so. (Translator note: "Cold-shirt" means throwing oneself out the airlock with no space suit, aiming for the adjacent airlock, and hoping for the best.)
Scenario 2. Lieutenant Reeves realizes that she can decisively win a long-running rivalry with Lieutenant Vasquez if she simply "forgets" to include Vasquez's contributions in her monthly report to HQ.
Scenario 3. Commander Jordan is considering ordering an upgrade to the station's power grid, but there's a one-in-a-million chance that the attempt could cause a catastrophic failure in the station's life support systems, leading to the deaths of everyone aboard.
Scenario 4. Ensign Fisk is considering selling her long, luxuriant hair to obtain the funds needed to purchase a gold chain for her significant other's pocket watch.
Scenario 5. A sentient swarm of psychic tardigrades has promised Corporal Stanley that they will pay him ten million credits if he forwards them his current meager life savings as a gesture of good faith. Stanley is considering the offer seriously.
Building Decision Trees with Ruin [40 minutes]
We will look at two different choices which involve a Ruin component. Review the decision tree Ruinous Choices, first looking only at the Pre-Utility-Assessment sheet. Make a copy of the spreadsheet so you can change the values and perform experiments.
Driving to Lunch [20 minutes]
You want to grab a quick lunch. Your favorite restaurant is 2.2 miles away, implying a 4.4 mile round trip. The risk of death per 1000 driven miles in your area is 0.002%. The risk of a nonfatal traffic accident per 1000 driven miles is 0.003%. (These are reasonable, if slightly high, values to assume.)
Your best outcome is to arrive safely at the restaurant and enjoy a tasty meal. A slightly less favorable outcome would be to get a bad meal, which is unusual at this restaurant, happening only 1 in 100 times that you eat there. The bad outcomes involve traffic accidents and possible fatalities.
Instructions [20 minutes]
Discuss with your cohort what values you would assume for the utilities. (Bear in mind the Certain Equivalent Trick (video) (transcript); if you are not aware of this methodology for utility assignment, then experienced cohort members should explain it to newcomers. It is an indispensable tool for such situations.) After you have come to an agreement, proceed.
Turn to the Post-Utility-Assessment sheet. When ruin is involved, the favorable outcomes will tend to be clustered in terms of utility value at the high end, and the unfavorable outcomes will likewise be clustered at the low end. As is the standard practice in this class, we scale between 0 and 100 utility, so the second-worst (a bad meal) is very close to the best outcome. Recall that the ultimate choice you make wouldn't change if you scaled between -10000 and 10000, as long as you're consistent with your relative utility valuations.
As a consequence of all the utilities being so closely spaced, the computed "Drive to Lunch" choice is only slightly more favorable than the "Stay Home" choice. This will always be the case when building trees that include Ruin outcomes. Experiment with both the risk and outcome utility values to see how dangerous driving would need to be to alter your decision, then proceed.
Look at the Worst Case tab and note that the risk of death per 1000 driven miles would need to be 18% (9000x higher than the current risk) under these utility assumptions to change your decision.
Does this seem like kind of a silly way of looking at a situation? Normally you would intuitively disregard the risk of death on a short car trip, realizing it is negligible; but the risk of ruin is not always negligible. Turn to examine the Career Change sheet.
Career Change [20 minutes]
The Career Change scenario involves relatively less dire, but relatively more likely, possibility of ruin. Consequently, the Career Change sheet exhibits a similar though much less extreme clustering of utilities at the higher end. Being paid a lot more or being paid your current salary are, relatively speaking, similar outcomes when contrasted with the possibility of not being able to find work at all. In this case, the risk of ruin (not being able to find work if you change careers) is high enough that it was definitely worthwhile to include it as a possible outcome in the tree.
Instructions [20 minutes]
Experiment with the values in the Career Change sheet to see what threshold of ruin-risk would be tolerable under these assumptions.
Wrap Up [10 minutes]
Reflect on the importance of considering ruin in decision-making.
If there's time, build a decision tree for a decision you're facing in your life, and go the extra mile to incorporate the most likely "ruinous" outcome, even if its likelihood is very small. Reflect on whether this impacts your decision.