Life Lessons from Theory

1. Time-consistency: Postponing is delivering more pain to the next self. Time-consistency is just respect to your future self. You are doing Game Theory with the future, and there is no reason to deviate.
2. Markov-Decision Processes + Sunk-cost Fallacy: No problems in life you should think about time (stationary), only think about the measures that matter (consequentialism). Do not think about the past, only at the point you are standing now (Markov). It is deterministic, a clear solution. There is many ways to be optimal, but their value is the same. A plan is an action for every contingency.
3. Backward Induction: Look at possible outcomes, find what you like, act backwards.
4. Incomplete Preferences: Not all things are easy to compare, but still you need to make a decision (revealed preferences). If you are indecisive, you either are indifferent or you need more information. In either case, the action needed to take is clear.
5. Exploration vs Exploitation: There is an optimal amount of exploring and understanding the environment before exploiting.
6. Mental Accounting and Combinatorial Optimization: Choosing among discrete choices is NP-hard. If you do Mental Accounting, you can use your money stress-free without complex comparisons. It may not be optimal, but is efficient.
7. Value Function Approximation (Reinforcement Learning): You cannot learn the exact value of every possible life situation through experience alone - there are too many possibilities. Instead, you learn general principles and patterns that help you estimate values in new situations. Seek experiences that teach you transferable lessons about what brings satisfaction, what relationships work, and what goals are worth pursuing. The goal isn’t to catalog every specific situation, but to build good intuition for evaluating new opportunities.
8. Gradient Descent: In order to solve life’s problems, SGD could be nice. You make little local adjustments. If the problem is very high-dimensional, we won’t be trapped in a local minima but go into a global one. 3B1B video on this is key.
9. Decision under Uncertainty: When dealing with uncertainty in life, weight how probable each outcome is and how much utility you have on each. Act accordingly. If ambiguous, just make a generous overprecaution and choose with that in mind. In any case, you will be not embarassed (definition of rationality by Gilboa) of your choice.
10. Equilibrium Resilience: An action (profile) is resilient if it is still a best-response even if few people still deviate. Make sure few people deviate, or make resilient plans.
11. Entropy and Following Rules: Again on group actions. People do not follow rules as well as we wanted them to. There is “entropy” and loss of adherence. Accomodate that. If there is no oversight, then people stop doing bothersome tasks. Make sure it is an habit before you go loose on oversight.