Why Decision Making Matters
Biases affecting decision-making. I mean, not only the ML Models decision making, but ours.
As the number of options increases, we will incur in decision fatige as this is a conginitively demanding.
The Go with the flow ones: Sunk cost, endowment effect, status-quo bias,
Common Traps
Sunk Cost
Why to stop now when we have invested so much in this? -> if we were not invested, how much would we invest NOW?
The endownment effect
We allocate much more value to what we already have (remember about loss aversion), this is applicable for objects, but also processes and activities.
Status-Quo Bias
The tendency to do something, simply because we have been doing it before. The preference for the current situation.
We are neurologically wired to favour the default option even it it brings sub-optimal results:
- Cost of Change
- Stability Preference
- Selection Difficulty - More options, more ‘computing’ cost.
- Anticipated regret and change
It has always been done this way.
Starting from a blank paper can help here to efficiently take decision.
The reversal Test: given a proposal to increase one parameter:
- If someone states that increasing P will have a bad outcome, then
- we could then try and make that parameter P, decrease. If that will be confronted, then
- That person should be able to explain why the current P value is the optimal one
- If not, we can suspect to be falling in Status-Quo Bias.