You are in a relationship, someone made some objection about communication, you don’t seem to understand what’s going on. Many years later you find yourself looking back at the relationship and reflecting with friends. That’s when someone brings up The Five Love Languages. Oh deep and great and meaningful secrets encoded into a book.
The 5 languages are:
- Quality time
- Words of affirmation
- Acts of service (devotion)
- Physical touch (intimacy)
Oooooh if only you had spent more energy trying to get quality time, and less effort on gifts that relationship could have been saved. Or the other way – the relationship was doomed because you wanted quality time and they wanted gifts as a show of love.
You start seeing the world in 5 languages, your coworker offering to get you a coffee is a gift. Your boss praising your good work is words of affirmation. You start thinking like a Man with a hammer. Strictly speaking I enjoy man with a hammer syndrome. I like to use a model to death, and then pick a new model and do it all again.
What I want you to do now is imagine you didn’t do that. Imagine we cloned the universe. In one universe we gave you the love-languages book and locked you in a room to read it. In the second universe we offered to run you through a new relationship-training exercise. “It’s no guide book on how to communicate with your partner, but it’s a pretty good process”, we lock you in a room with a chair, a desk, some paper, pens (few distractions) and order you to derive some theory and idea about how to communicate with your partner.
Which one do you predict will yield the best result?
When I ask my system 2, it is fairly happy with the idea that using someone else’s model is a shortcut to finding the answers. After all they pre-derived the model. No need to spend hours working on it myself when it’s all in a book.
When I ask my system 1, it thinks that the self-derived system is about a billion times better than the one I found in a book. It’s going to be personally suited, it’s going to be sharp and accurate, and bend to my needs.
Which is going to yield the best result for the problem? Self-derived solutions to all future problems? Book-derived solutions for all problems?
I propose that the specific strategy used to answer the problem, depending on the problem (obviously sometimes 1+1 will only be solved with addition, and solving it with subtraction is going to be difficult), is mostly irrelevant compared to having the meta-strategy.
In the original example:
My relationship has bad communication, so we end the relationship.
The meta-strategy for this case:
My relationship has bad communication, how do we find more information about that and solve that problem.
In the general case:
I have a problem, I will fix the problem.
the meta strategy for the general case:
I have a problem, what is the best way to solve the problem?
Or the meta-meta strategy:
I have a problem, how will I go about finding what is the best way to solve the problem?
I propose that having the meta strategy, and the meta-meta strategy is almost as powerful as the true strategy. On the object level for the problem example, instead of searching for the book in the problem field that is the five love languages you could instead search for any book about the problem area. Any book is better than no book. In fact I would make a hierarchy:
The best strategy > a good strategy > any strategy > no strategy
The best book > a good book > any book on the topic > no book on the topic
You encounter a problem in the wild – what should you do?
- Try just solve the problem
- Try any strategy (with a small amount of thinking – a few seconds or minutes)
- search for a better strategy
Depending on the problem, the time, the real factors – the best path forward may be to just “think of what to do then do that”, or it may be to “stop and write out a 10 page plan before executing 10 pages worth of instructions”.
Should you read the five love languages book? That depends. What is the problem? and have you tried solving the problem on your own first?
Meta: this took an hour to write.
Cross posted to lesswrong: lesswrong.com/r/discussion/lw/nta