author: niplav, created: 2025-02-27, modified: 2025-03-21, language: english, status: notes, importance: 1, confidence: joke
New philosophical position/Effective Altruism cause area dropped. Benefits of pursuing the creation of libertarian free will likely not worth the costs, based on a Monte-Carlo estimate.
The philosophical literature is rich in discussions about free will: Whether beings (usually humans1) can make choices that are not, in some sense, "completely determined" by the causal forces of nature. I won't take any position in that debate here, but instead will ask a far less explored question:
If libertarian free will does not exist, would it be good to bring it into existence?
Clearly, humans value freedom in various: Libertarianism makes it the central tenet of its political philosophy, mundane forms of unfreedom (slavery, imprisonement, social pressure &c) are usually regarded as undesirable, at least in WEIRD morality; the capability approach and the will to power both take as central object the value of expanded actions and some theories of life and artificial intelligence take expanded empowerment/control over the environment as the feature defining successful existence.
Therefore, it's just natural to extend the desire for an expanded action space to the ontological level; if we could, would we not want to be able to act counter to the laws of physics, peskily constraining each of us to a single future? (Many worlds notwithstanding, since it also produces a pre-defined multiverse, and quantum randomness (if it exists) also does not count, as outlined in the debate around the existence of libertarian free will.)
One can call this the constructive axiological free will hypothesis (CAFWH): "If libertarian free will doesn't (yet) exist, it would be good to create it and imbue humans with it".
The ITN framework is ready-to-hand for evaluating whether working on the CAWFH is a good idea.
Starting with the weakest point. It is currently not clear how to bring about new ontological entities. One can weakly estimate the cost from taking the number of basic ontological categories in existence and dividing them by the work that was involved in bringing them about.
As for basic ontological categories, my estimate ranges from 0 (ontological nihilism) to ~10 (list: mathematical objects, God and/or gods, qualia, matterenergy, spacetime, abstract objects, souls, moral facts and likely others I have forgotten to consider).
The total work needed to bring those about is difficult to estimate, as it may range from no work (in the case of ontological nihilism) up to absolute infinity if supernatural beings created existence. I'll take as a mean the matter-energy content of the observable universe multiplied by four (since matter-energy and spacetime are only two of the eight basic ontological categories listed above).
The amount of baryonic
matter
in the observable universe is estimated
at ~$10^{53}$ kg,
but since baryonic matter only makes up 4.5% of the total mass-energy
in the universe, I'll adjust the estimate (not leaving out dark energy &
dark matter).
Using squigglepy:
import squigglepy as sq
num_ontological_categories=sq.to(0, 10)
cost_new_entity=sq.lognorm(lognorm_mean=(1/0.045)*10**53, lognorm_sd=20)
prop_universe_entities=num_ontological_categories/sq.to(1, 3)
new_entity_cost=cost_new_entity/prop_universe_entities
Usually, a new ontological entity costs the mass-energy of $10^{53}$ kg:
>>> sq.sample(new_entity_cost, 10)
array([2.96996124e+53, 7.18535888e+53, 2.98417566e+54, 1.11103257e+54,
6.09266737e+53, 8.39602018e+53, 4.08215136e+53, 5.01485048e+53,
4.63852243e+53, 3.93717114e+53])
It would likely be valuable to create libertarian free will. (Though see the list of possible risks below.)
As a proxy, one can try to estimate how much time and energy humans expand on broadening the list of possible choices available to them; examples include education, migration to more democratic and liberal countries, buying transport, many health interventions &c. Philosophical intuition points me to ~10% of human effort being spent on pure expansion of mundane freedoms.
Assuming that libertarian free will would be ~5× more valuable, and a ~20% chance that humans already have libertarian free will, together with a gross world product of ~$100T, we can arrive at the value humanity should be willing to pay for libertarian free will:
gwp=sq.norm(mean=10**14, sd=1)
prop_spent_on_freedom=sq.beta(a=2, b=8)
real_freewill_mult=sq.to(4, 20)
chance_freewill_exists=sq.beta(a=1, b=4)
total_value=prop_spent_on_freedom*gwp*real_freewill_mult*(1-chance_freewill_exists)
Sampling, we usually get a few tens of trillions of dollars in value:
>>> sq.sample(total_value, 10)
array([9.96797785e+13, 4.23572584e+13, 8.24106658e+13, 4.32935165e+14,
2.43818602e+14, 1.26017517e+14, 1.86109674e+14, 1.43918708e+14,
4.01545018e+13, 2.40841431e+14])
As far as I know, no being is working on creating libertarian free will de novo, and philosophers have not yet discussed the possibility and desirability2. It is possible that hidden supernatural entities are engaged in the process, but that seems unlikely4%, and it's not clear they would imbue humans specifically with the capability if they created it.
Unfortunately, it seems like the cost for new ontological basic entities
is too high: It is implausible that we will be able to bring up the
equivalent of ~$10^{53}-10^{54}$ kg of massenergy with an investment
of ~$$10^{13}-10^{14}$.
However, due to the small amount of thought that has gone into ontological engineering, there is the potential that creating new ontological categories is much cheaper than estimated here, and the neglectedness leaves space for a few experimental philosophers.
Drastic changes to the structure of existence are not without their risks. Creating libertarian free will may induce multiple hazards:
Though even here we may consider possibilities similar to the ones Shalizi discusses: What if free will is possible, but only a few species in the past had it, or only a few species in the future will have it, but not homo sapiens? What if there are inanimate objects that have free will, but choose not to exercise it? What if some humans have free will, but others don't, or if all humans have free will as a latent ability, but have failed to notice and deploy it? How about the hypothesis that figuring out whether something/someone has free will is undecidable, or at least EXPTIME-complete? Humans may not have free will themselves, but be theoretically able to create beings with free will… it's just that, by the way the universe is structured, we never will. ↩
Based on a short websearch using Google Scholar & Perplexity, and a conversation with ChatGPT with web search enabled. ↩