Applied Decision Theory 1

No upcoming meetings

Welcome to the first of what will be a recurring decision-making workshop. Making good decisions is synonymous with being effective in the world. The art of accurate decision-making is at the intersection of epistemic rationality (understanding the world) and instrumental rationality (taking action).

There is no pre-workshop reading or assignment. The following instructions apply to the cohort meeting.

Non-facilitators should avoid reading blocks marked as Facilitator, such as the one below, since they will often contain spoilers about the problems.

This week you will first warm up with a relatively simple decision problem example, and then move on to solving your own personal decision problems.

Warm Up [20-30 minutes]

Make your own copy of the Radiation at Station Seven spreadsheet. It has been filled out with most of the relevant values and equations.

Utility Valuations [5 minutes]

Start with the Radiation Decision tab of the spreadsheet. Discuss the utility valuations in the Outcome Utilities column.

  1. What is the best outcome?
  2. What is the worst outcome?
  3. Does it make sense that they're both outcomes of the same choice?
  4. Which choice is best?

Oracle [10 minutes]

An Oracle AI is passing through your sector and offers to help you with your problem. Move to the "Oracle Decision" tab of the spreadsheet. The Oracle can simply tell you whether the radiation will be severe or not, for a fee. You perform a quick Value of Information calculation. Your choice is now not whether to bring the radiation shield, but whether to employ the Oracle. If you don't employ the oracle, your decision tree looks the same. List off the differences if you do employ the Oracle.

Payment [10 minutes]

The Oracle asks for 30 utility-equivalent in local currency for the information. Should you pay up?

Decision Workshop [50 minutes]

Choose a Decision [15 minutes]

Each cohort member should put forth a decision that they are thinking about. The decisions can be big or small. Whether you're pondering a change in career, waffling on whether to upgrade your PC, or deciding whether to dye your hair, if it's something you're having trouble making your mind up about, it's a valid decision to discuss. Everyone should put forth something.

Decomposition [20 minutes]

Set a timer for this section. Everyone will now individually break their decision problem into a small number of choices, outcomes, and distinctions, attempting to build the structure of a decision tree. I suggest using Google Sheets or Slides for sharing purposes. Do not worry about the probabilities or utilities at this point, only concern yourself with the structure.

Aim for no more than 3 possible choices, and no more than 2 levels of distinction, each distinction having no more than 3 possibilities each. (3 choices with 2 distinctions at 2 possibilities each already results in 12 outcomes!) In other words, try to write down your decision problem as simply as possible without losing anything important. Your object is to create a tree that communicates your decision problem clearly and effectively.

Sharing [15 minutes]

Return to the cohort session and share your decision tree. Describe your thought process in designing the tree. Remark upon whether you felt the exercise was helpful/clarifying in coming to a decision. The group should provide feedback, especially if something seems to be a mistake.

Wrap up

If there's time remaining, begin filling in your outcome utilities and branch probabilities. If not, then consider completing the trees on your own time, if you conclude that it would be valuable to do so.

Community Notes

This section contains links and information that Guildmembers found helpful.

Write a Note

You must be signed in to write a note.