Goal
Local minima
Consultant
- Bounded rational agents
- Cost of backtracking
- Decision space is divergent
How to prune the decision space?
We can do better than random or exhaustive search
Multi-Armed bandit
Maximize some reward function through balancing exploration & exploitation
... failure and invention are inseparable twins. To invent you have to experiment, and if you know in advance that it's going to work, it's not an experiment. Most large organizations embrace the idea of invention, but are not willing to suffer the string of failed experiments necessary to get there.
Jeff Bezos
Uncertainty
Aleatoric & Epistemic
Uncertainty
natural randomness vs incomplete model/map of situation
How do we improve this map?
Epistemic uncertainty
Uncertainty
Aleatory variability
P(tiger | friend saw tiger) ?
P(terrorist attack | terrorism attack happened yesterday) ?
Local vs Global knowledge
Loss aversion due to global knowledge hampers exploration within decision space !
P(terrorist attack | terrorism attack happened yesterday) ?
-
-
-
-
+
Continuous negative feedback of exploration encourages quick wins to create artificial, short-term positive feedback
- use two-way door decisions as exploration tool
- This reduces epistemic uncertainty in your environment
- A more precise map allows for long-term planning
- Beware loss aversion
- Backtracking is good!
Key Takeaways
exploration vs exploitation
By laurenstc
