Scope and focus
What do I want as the main outcome
Focusing on systems behavior, not on behavior of individuals (as opposed to the Stanford study, J.S. Park), although individual behavior might influence or determine the behavior of a system (which would potentially vary based on system size, function and other factors).
Use-cases include:
- Predicting the evolution of a given metric based on non-political changes (like abortion rights) introduced in the system
- Decision making in resource allocation
- E.g. Predicting what impact on human happiness and in what time horizon would building a new huge reactor in central Africa VS new space program with the goal of mining the asteroid belt have
- This problem could also be viewed as political: people in the central Africa region might feel they’re being forgotten about.
- Option 1: The system should predict political uncertainty in the region as one of the possibilities
- Option 2: A paradigm shift within the human conception will happen and people will understand that the decision is not political, but it’s only a metric optimization technique. Mechanisms would be put in place to ensure that a specific region (or any other subset of people) wouldn’t be “too much” worse off than the rest / other subset of people. Nevertheless, that’s a question of the moral and ethical views of the system which should be embedded within the prediction system.
- Both options should probably be combined within the prediction system
- Decision making in how to change the everyday life (Irwin, Notes 7 or page 7)
- Would people be happier in community housing in some period of their life?
- What would be necessary to make people feel satisfied with their life if they didn’t have to work if they didn’t want to?*
- How should a personal AI be structured to be the most helpful?*
-
- the prediction system probably should suggest the structure but only test the different structures input in the model
But don’t include:
- Predicting what impact would a political decision (like abortion rights) have on a given metric (like human happiness)
Should be included?:
Possible use-cases for predicting system behavior:
Include:
- A/B testing prediction, user testing
- Market’s reaction to a new product (demand and willingness-to-buy prediction) – testing phase
- Evolution of key factors in a system after introducing a change (e.g. “How is happiness of a society affected by introduction of a UBI (universal basic income)”) and general forecasting