Epistemic status: My partner and I did 3 premortems for our romantic relationship and we found it surprisingly helpful.

Summary

1. My partner and I did 3 premortems for our romantic relationship.
2. The procedure is: define prompts; collect failures; independently estimate risk of each; align on estimates; brainstorm solutions.
3. In addition to usual benefits of a single person premortem, such session becomes a dedicated safe space to raise and discuss concerns in batch.

What is premortem?

Premortem (aka Murphyjitsu) is "a process for bulletproofing your strategies and plans". (CFAR handbook). The idea is to first think how your plans can fail, then brainstorm ways to prevent these failures. For a deeper introduction please see Murphyjitsu section in the CFAR handbook.

Procedure

My partner and I read the CFAR handbook together. We decided to do a premortem on our relationship. This might have sounded awkward ("Let's brainstorm how our relationship can fail"), but keeping the end goal - improving likelihood of success - helped to avoid this pitfall. Since then we did 3 premortems and converged to a following procedure:

1. Define prompt(s). The goal is to make sure that we brainstorm in the same direction. Examples:
   1. it is 6 month from now, we broke up, how did this happen?
   2. we came back from a vacation, it was terrible, why?
2. Independently collect failure scenarios. Separately collect ideas why the failure could occur. I personally do this across multiple days. This way I just notice potential failures during conversations and other activities. My partner prefers to brainstorm right before the discussion. Doing this separately helps to improve our coverage, since we don't bias each other.
3. Explain scenarios. Add your scenarios to a shared spreadsheet. Make sure that both of you understand all scenarios. If not, give examples or ask questions.
4. Independently rank scenarios. For each scenario separately estimate expected impact from 1 (negligible) to 10 (crisis). The expected impact already includes both probability and magnitude. It is a product of how likely a scenarios is by how bad it is.
5. Discuss scenarios with large differences in expected impact; update estimates until converged. Sort scenarios in decreasing order of absolute difference between your estimates. Discuss all scenarios which have difference at least 2. Usually the person whose estimate is larger presents their concern first. Present each other arguments until you align on the estimate. That is update the estimates until the difference becomes at most 1. The goal is to agree on potential risk of each scenario and its derivation.
6. Discuss top scenarios with largest sum of estimates, brainstorm mitigations. Sort scenarios in decreasing order of the sum of your estimates. Discuss top items, brainstorm ways to decrease their likelihood.

Why this is useful?

In the contrast to the original single person Murphyjitsu technique, a group premortem seems to bring value even before mitigations are discussed. Just aligning on potential risk of each scenario already brings value. In each of our 3 sessions there were concerns, which one of us had, but then they were completely defused by the information other party had. In other words, the session becomes a dedicated safe space to express your concerns and exchange available related information. This is useful in two ways:

1. Normally one does not think about failure scenarios at all
2. This batches discussing the concerns. To have a productive discussion one needs to be in a special mental state. Entering this state for each concern separately is too much overhead. Also raising concerns one by one in ordinary conversations just seems odd.

The brainstorming for mitigations is also useful. First, there are usual benefits of a single person premorten - this creates a list of potential issues and readily available actions. This helps to notice the problems and makes it easier to actually perform the mitigations. Second, there are advantages due to communication and alignment:

• the solutions have been approved by both parties. Thus, when one executes a particular solution, they can be sure that this is ok with the other party and they will have their support. This decreases friction to execute.
• this creates a communication shortcut - "I think that top issue from the premortem is happening now". This reduces friction to bring the issue up in the future.

Potential pitfalls

It is not clear whether premortem actually changed our relationship or we would have discussed these issues anyway.

I tried prompting other people, who are not into rationality, to do premortems outside of romantic context. This didn't work. People either didn't try hard enough or kept refuting the prompt.