Risks of pooling resources with past & future selves
Reflecting on Adam Elga’s paper on the psychology of decision-makers pooling resources across e.g. fire stations (firemen), power plants (energy supply), or banks (liquidity); and how, given that pooling resources among nodes in a network like this will reduce wastage (saliently & naively) and reduce the probability of small-to-medium failures, decision-makers in pooled-resource networks will reduce the safety margins of nodes on the network, which significantly & non-obviously increases the small-but-meaningful probability of catastrophic failures (e.g. all of the fire stations are over capacity, all of the power plants fail, all of the banks don’t have enough liquidity).
Whew.
I was thinking about it in the context of past & future selves — you can think of your ‘self’ as a collection of nodes (past & future selves) on a network (the timeline of your life), each with some resources, some safety margin, and some demand for resources. The resources of the nodes on this network are pooled/coupled, not individual/decoupled: you sometimes ‘borrow’ from past/future versions of yourself, or ‘lend’ out to past/future versions of yourself. For examples:
- You might decide to borrow from past selves who saved for an emergency when your current self experiences one.
- Your current self in February might decide to do your taxes early, thereby lending time to your future self in April, when you know that a work project will become very demanding on your time/energy.
Elga’s paper characterizes the risk associated with this type of resource pooling, in particular underestimating catastrophic failures. What would the underestimate of catastrophic failures look like in this case?
An interesting property of this particular network compared to fire stations and financial institutions and power stations is that, with past/future selves, the ‘decision-maker’ can only directly affect one node in this network at a time. You can’t simultaneously coordinate lots of different resource-sharing agreements between lots of different nodes on the network; you can just (directly) affect your current-self’s choices. (Of course, you can indirectly affect half of other selves choices by, e.g., causing them to be in an environment in which their natural incentives point in particular directions: the future selves. For example, you can make particular friends who will push future selves in particular directions; you can get notifications for particular bank statements or investment implications; etc.)
Further:
- Acausal trade (LW overview, ACX overview)
- Hofstader’s writing on this & related topics
- Jacob Goldman-Wetzler, "Taking into account preferences of past selves”